Modeling and performance evaluation of ground source heat pump systems论文翻译

Book reviewModeling,Simulation,and Control of Flexible Manufacturing Systems ±A Petri Net Approach;Meng Chu Zhou;Kurapati Venkatesh;Yushun Fan;World Scienti®c,Singapore,19991.IntroductionA ¯exible manufacturing system (FMS)is an automated,mid-volume,mid-va-riety,central computer-controlled manufacturing system.It can be used to produce a variety of products with virtually no time lost for changeover from one product to the next.FMS is a capital-investment intensive and complex system.In order to get the best economic bene®ts,the design,implementation and operation of FMS should be carefully made.A lot of researches have been done regarding the modeling,simulation,scheduling and control of FMS [1±6].From time to time,Petri net (PN)method has also been used as a tool by di erent researcher in studying the problems regarding the modeling,simulation,scheduling and control of FMS.A lot of papers and books have been published in this area [7±14].``Modeling,Simulation,and Control of Flexible Manufacturing Systems ±A PN Approach''is a new book written by Zhou and Venkatesh which is focused on studying FMS using PN as a systematic method and integrated tool.The book's contents can be classi®ed into four parts.The four parts are introduction part (Chapter 1to Chapter 4),PNs application part (Chapter 5to Chapter 8),new research results part (Chapter 9to Chapter 13),and future development trend part (Chapter 14).In the introduction part,the background,motivation and objectives of the book are described in Chapter 1.The brief history of manufacturing systems and PNs is also presented in Chapter 1.The basic de®nitions and problems in FMS design and implementation are introduced in Chapter 2.The authors divide FMS related problems into two major areas ±managerial and technical.In Chapter 4,basic de®nitions,properties,and analysis techniques of PNs are presented,Chapter 4can be used as the fundamentals of PNs for those who are not familiar with PN method.In Chapter 3,the authors presented their approach to studying FMS related prob-lems,the approach uses PNs as an integrated tool and methodology in FMS design and implementation.In Chapter 3,various applications in modeling,analysis,sim-ulation,performance evaluation,discrete event control,planning and scheduling of FMS using PNs are presented.Through reading the introduction part,the readers can obtain basic concepts and methods about FMS and PNs.The readers can also get a clear picture about the relationshipbetween FMS and PNs.Mechatronics 11(2001)947±9500957-4158/01/$-see front matter Ó2001Elsevier Science Ltd.All rights reserved.PII:S 0957-4158(00)00057-X948Book review/Mechatronics11(2001)947±950The second part of the book is about PNs applications.In this part,various applications of using PNs in solving FMS related problems are introduced.FMS modeling is the basis for simulation,analysis,planning and scheduling.In Chapter5, after introduction of several kinds of PNs,a general modeling method of FMS using PNs is given.The systematic bottom-up and top-down modeling method is pre-sented.The presented method is demonstrated by modeling a real FMS cell in New Jersey Institute of Technology.The application of PNs in FMS performance analysis is introduced in Chapter 6.The stochastic PNs and the time distributions are introduced in this Chapter. The analysis of a¯exible workstation performance using the PN tool called SPNP developed at Duke University is given in Section6.4.In Chapter7,the procedures and steps involved for discrete event simulation using PNs are discussed.The use of various modeling techniques such as queuing network models,state-transition models,high-level PNs,object-oriented models for simulations are brie¯y explained.A software package that is used to simulate PN models is introduced.Several CASE tools for PNs simulations are brie¯y intro-duced.In Chapter8,PNs application in studying the di erent e ects between push and pull paradigms is shown.The presented application method is useful for the selection of suitable management paradigm for manufacturing systems.A manufacturing system is modeled considering both push and pull paradigms in Section8.3which is used as a practical example.The general procedures for performance evaluation of FMS with pull paradigm are given in Section8.4.The third part of the book is mainly the research results of the authors in the area of PNs applications.In Chapter9,an augmented-timed PN is put forward. The proposed method is used to model the manufacturing systems with break-down handling.It is demonstrated using a¯exible assembly system in Section9.3. In Chapter10,a new class of PNs called Real-time PN is proposed.The pro-posed PN method is used to model and control the discrete event control sys-tems.The comparison of the proposed method and ladder logic diagrams is given in Chapter11.Due to the signi®cant advantages of Object-oriented method,it has been used in PNs to de®ne a new kind of PNs.In Chapter12,the authors propose an Object-oriented design methodology for the development of FMS control software.The OMT and PNs are integrated in order to developreusable, modi®able,and extendible control software.The proposed methodology is used in a FMS.The OMT is used to®nd the static relationshipamong di erent objects.The PN models are formulated to study the performance of the FMS.In Chapter12,the scheduling methods of FMS using PNs are introduced.Some examples are presented for automated manufacturing system and semiconductor test facility.In the last Chapter,the future research directions of PNs are pointed out.The contents include CASE tool environment,scheduling of large production system,su-pervisory control,multi-lifecycle engineering and benchmark studies.Book review/Mechatronics11(2001)947±950949 mentsAs a monograph in PNs and its applications in FMS,the book is abundant in contents.Besides the rich knowledge of PNs,the book covers almost every aspects regarding FMS design and analysis,such as modeling,simulation,performance evaluation,planning and scheduling,break down handling,real-time control,con-trol software development,etc.So,the reader can obtain much knowledge in PN, FMS,discrete event system control,system simulation,scheduling,as well as in software development.The book is a very good book in the combinations of PNs theory and prac-tical applications.Throughout the book,the integrated style is demonstrated.It is very well suited for the graduate students and beginners who are interested in using PN methods in studying their speci®c problems.The book is especially suited for the researchers working in the areas of FMS,CIMS,advanced man-ufacturing technologies.The feedback messages from our graduate students show that compared with other books about PNs,this book is more interested and easy to learn.It is easy to get a clear picture about what is PNs method and how it can be used in the FMS design and analysis.So,the book is a very good textbook for the graduate students whose majors are manufacturing systems, industrial engineering,factory automation,enterprise management,and computer applications.Both PNs and FMS are complex and research intensive areas.Due to the deep understanding for PNs,FMS,and the writing skills of the authors,the book has good advantages in describing complex problems and theories in a very easy read and understandable fashion.The easy understanding and abundant contents enable the book to be a good reference book both for the students and researchers. Through reading the book,the readers can also learn the new research results in PNs and its applications in FMS that do not contained in other books.Because the most new results given in the book are the study achievements of the authors,the readers can better know not only the results,but also the background,history,and research methodology of the related areas.This would helpthe researchers who are going to do the study to know the state-of-art of relevant areas,thus the researchers can begin the study in less preparing time and to get new results more earlier.As compared to other books,the organization of the book is very application oriented.The aims are to present new research results in FMS applications using PNs method,the organization of the book is cohesive to the topics.A lot of live examples have reinforced the presented methods.These advantages make the book to be a very good practical guide for the students and beginners to start their re-search in the related areas.The history and reference of related research given in this book provides the reader a good way to better know PNs methods and its applications in FMS.It is especially suited for the Ph.D.candidates who are determined to choose PNs as their thesis topics.950Book review/Mechatronics11(2001)947±9503.ConclusionsDue to the signi®cant importance of PNs and its applications,PNs have become a common background and basic method for the students and researchers to do re-search in modeling,planning and scheduling,performance analysis,discrete event system control,and shop-¯oor control software development.The book under re-view provides us a good approach to learn as well as to begin the research in PNs and its application in manufacturing systems.The integrated and application oriented style of book enables the book to be a very good book both for graduate students and researchers.The easy understanding and step-by-step deeper introduction of the contents makes it to be a good textbook for the graduate students.It is suited to the graduated students whose majors are manufacturing system,industrial engineering, enterprise management,computer application,and automation.References[1]Talavage J,Hannam RG.Flexible manufacturing systems in practice:application,design,andsimulation.New York:Marcel Dekker Inc.;1988.[2]Tetzla UAW.Optimal design of¯exible manufacturing systems.New York:Springer;1990.[3]Jha NK,editor.Handbook of¯exible manufacturing systems.San Diego:Academic Press,1991.[4]Carrie C.Simulation of manufacturing.New York:John Wiley&Sons;1988.[5]Gupta YP,Goyal S.Flexibility of manufacturing systems:concepts and measurements.EuropeanJournal of Operational Research1989;43:119±35.[6]Carter MF.Designing¯exibility into automated manufacturing systems.In:Stecke KE,Suri R,editors.Proceedings of the Second ORSA/TIMS Conference on FMS:Operations Research Models and Applications.New York:Elsevier;1986.p.107±18.[7]David R,Alla H.Petri nets and grafcet.New York:Prentice Hall;1992.[8]Zhou MC,DiCesare F.Petri net synthesis for discrete event control of manufacturing systems.Norwell,MA:Kluwer Academic Publishers;1993.[9]Desrochers AA,Al-Jaar RY.Applications of petri nets in manufacturing systems.New York:IEEEPress;1995.[10]Zhou MC,editor.Petri nets in¯exible and agile automation.Boston:Kluwer Academic Publishers,1995.[11]Lin C.Stochastic petri nets and system performance evaluations.Beijing:Tsinghua University Press;1999.[12]Peterson JL.Petri net theory and the modeling of systems.Englewood Cli s,NJ:Prentice-Hall;1981.[13]Resig W.Petri nets.New York:Springer;1985.[14]Jensen K.Coloured Petri Nets.Berlin:Springer;1992.Yushun FanDepartment of Automation,Tsinghua UniversityBeijing100084,People's Republic of ChinaE-mail address:*****************。

performance evaluation理工英语4 Performance EvaluationIntroductionPerformance evaluation is a crucial process in assessing the effectiveness and efficiency of individuals, teams, or organizations. It involves the systematic assessment and measurement of performance against predetermined goals, objectives, and standards. This article aims to explore the concept of performance evaluation, its significance in various contexts, and the different methods used for evaluation.Defining Performance EvaluationPerformance evaluation is defined as the systematic process of assessing and reviewing an individual's or organization's performance in relation to established goals and objectives. It involves analyzing the quality, quantity, and timeliness of work, as well as the overall contribution towards achieving desired outcomes.Significance of Performance EvaluationPerformance evaluation plays a critical role in various contexts, including:1. Employee Performance Evaluation: In organizations, performance evaluation helps assess employees' job performance, identify areas for improvement, and determine reward and promotion opportunities. It provides valuable feedback and helps create a performance-driven culture.2. Team Performance Evaluation: Evaluating team performance is essential for identifying strengths and weaknesses, enhancing collaboration, and optimizing resources. It enables organizations to allocate tasks effectively, promote teamwork, and achieve collective goals.3. Organizational Performance Evaluation: Assessing the overall performance of an organization is essential for strategic planning, decision-making, and performance improvement. It helps identify areas requiring attention and enables organizations to align their objectives with key performance indicators (KPIs).Methods of Performance EvaluationThere are several methods used for performance evaluation, depending on the nature and context of evaluation:1. Rating Scales: This method involves using predefined scales to rate employees' performance against specific criteria. It provides a structured approach and simplifies the evaluation process. However, it can be subjective and may not capture the full extent of performance.2. 360-Degree Feedback: This method involves obtaining feedback from multiple sources, including supervisors, subordinates, peers, and customers. It provides a holistic view of an individual's performance and promotes a comprehensive understanding of strengths and areas for improvement.3. Objective Measurements: Objective measurements involve quantifying performance based on quantifiable data, such as sales figures, production output, or customer satisfaction ratings. This method provides a precise assessment of performance but may not capture qualitative aspects.4. Self-Assessment: Self-assessment encourages individuals to reflect on their performance and identify areas for improvement. It promotes self-awareness, accountability, and personal development. However, it may be biased and influenced by individuals' perceptions.5. Behavioral Observation: This method involves directly observing individuals' behavior in specific work-related situations. It provides valuable insights into work habits, interpersonal skills, and adherence to organizational values. However, it can be time-consuming and may not capture performance in all areas.ConclusionPerformance evaluation is a vital process for assessing and improving individual, team, and organizational performance. It helps organizations align their objectives, motivate employees, and ensure efficient resource allocation. By using appropriate evaluation methods, organizations can drive continuous improvement and achieve long-term success. It is essential for organizations to establish clear evaluation criteria, provide constructive feedback, and support employee development to maximize the benefits of performance evaluation.。

Geometric ModelingGeometric modeling is a fundamental aspect of computer-aided design (CAD) and computer graphics, playing a crucial role in the creation of virtual 3D models and simulations. This technology has revolutionized various industries, including architecture, engineering, animation, and gaming, by enabling designers and developers to visualize and manipulate complex geometric shapes with precision and efficiency. In this response, we will explore the historical background, different perspectives, case studies, and critical evaluation of geometric modeling, as well as its future implications and recommendations. The development of geometric modeling can be traced back to the early 1960s when Ivan Sutherland created Sketchpad, a revolutionary computer program that allowed users to draw and manipulate basic geometric shapes on a screen. This marked the beginning of computer-aided design (CAD) and laid the foundation for the development of geometric modeling as we know it today. Over the years, geometric modeling has evolved significantly, with the introduction of advanced algorithms, rendering techniques, and modeling tools that have expanded its applications across various industries. From a historical perspective, the evolution of geometric modeling has been driven by the increasing demand for more sophisticated and realistic 3D models in fields such as architecture, automotive design, industrial engineering, and entertainment. As technology has advanced, so too has the complexity and realism of geometric models, leading to a greater emphasis on precision, detail, and interactivity in the design and visualization process. From a technological perspective, geometric modeling has undergone a paradigm shift with the advent of parametric and non-parametric modeling techniques. Parametric modeling allows designers to create models based on a set of parameters, enabling them to make changes and updates to the design easily. On the other hand, non-parametric modeling focuses on creating freeform shapes and surfaces, providing greater flexibility and creativity in the design process. These different perspectives have led to debates within the industry about which approach is more effective for specific applications and design requirements. One example of the impact of geometric modeling can be seen in the field of architecture. With the use of CAD software and advanced geometric modeling tools, architects can create highlydetailed and realistic 3D models of buildings and structures, allowing them to visualize the final product and make necessary adjustments before construction begins. This not only improves the design process but also helps in conveying the design intent to clients and stakeholders, leading to better communication and decision-making. In the automotive industry, geometric modeling hasrevolutionized the design and manufacturing process, enabling engineers and designers to create complex 3D models of vehicles and their components with precision and accuracy. This has led to significant improvements in vehicle safety, performance, and aesthetics, as well as streamlined production processes and reduced time-to-market for new vehicle models. While geometric modeling has brought about numerous benefits, it also poses certain challenges and drawbacks. One of the main drawbacks is the steep learning curve associated with advanced modeling tools and techniques, which can be a barrier for newcomers to the field. Additionally, the complexity of geometric models can lead to performance issuesand computational challenges, especially when dealing with large-scale models or simulations. Looking ahead, the future implications of geometric modeling arevast and promising. As technology continues to advance, we can expect to see even more realistic and interactive 3D models that push the boundaries of visualfidelity and immersion. Furthermore, the integration of geometric modeling with other emerging technologies such as virtual reality (VR) and augmented reality (AR) holds great potential for creating new and innovative applications in fields such as education, training, and entertainment. In conclusion, geometric modeling has played a pivotal role in shaping the way we design, visualize, and interact with3D models across various industries. While it has brought about significant advancements and benefits, there are also challenges and considerations that need to be addressed. By understanding the historical background, different perspectives, and case studies related to geometric modeling, we can better appreciate its impact and potential for the future. As technology continues to evolve, it is essential to stay informed and adaptable to the latest trends and developments in geometric modeling, in order to harness its full potential and drive innovation in design and visualization.。

SPECIFICA TIONSNI cDAQ™-91844-Slot, Ethernet CompactDAQ ChassisDefinitionsWarranted specifications describe the performance of a model under stated operating conditions and are covered by the model warranty.Characteristics describe values that are relevant to the use of the model under stated operating conditions but are not covered by the model warranty.•Typical specifications describe the expected performance met by a majority of the models.•Nominal specifications describe parameters and attributes that may be useful in operation. Specifications are Typical unless otherwise noted.ConditionsSpecifications are valid at 25 °C unless otherwise noted.Analog InputInput FIFO size127 samples per slotMaximum sample rate1Determined by the C Series module or modules Timing accuracy250 ppm of sample rateTiming resolution212.5 nsNumber of channels supported Determined by the C Series module or modules 1Performance dependent on type of installed C Series module and number of channels in the task.2Does not include group delay. For more information, refer to the documentation for each C Series module.Analog OutputNumber of channels supportedHardware-timed taskOnboard regeneration16Non-regeneration Determined by the C Series module or modules Non-hardware-timed task Determined by the C Series module or modules Maximum update rateOnboard regeneration 1.6 MS/s (multi-channel, aggregate)Non-regeneration Determined by the C Series module or modules Timing accuracy50 ppm of sample rateTiming resolution12.5 nsOutput FIFO sizeOnboard regeneration8,191 samples shared among channels used Non-regeneration127 samples per slotAO waveform modes Non-periodic waveform,periodic waveform regeneration mode fromonboard memory,periodic waveform regeneration from hostbuffer including dynamic updateDigital Waveform CharacteristicsWaveform acquisition (DI) FIFOParallel modules511 samples per slotSerial modules63 samples per slotWaveform generation (DO) FIFOParallel modules2,047 samples per slotSerial modules63 samples per slotDigital input sample clock frequencyStreaming to application memory System-dependentFinite0 MHz to 10 MHz2| | NI cDAQ-9184 SpecificationsDigital output sample clock frequencyStreaming from application memory System-dependentRegeneration from FIFO0 MHz to 10 MHzFinite0 MHz to 10 MHzTiming accuracy50 ppmGeneral-Purpose Counters/TimersNumber of counters/timers4Resolution32 bitsCounter measurements Edge counting, pulse, semi-period, period,two-edge separation, pulse widthPosition measurements X1, X2, X4 quadrature encoding withChannel Z reloading; two-pulse encoding Output applications Pulse, pulse train with dynamic updates,frequency division, equivalent time sampling Internal base clocks80 MHz, 20 MHz, 100 kHzExternal base clock frequency0 MHz to 20 MHzBase clock accuracy50 ppmOutput frequency0 MHz to 20 MHzInputs Gate, Source, HW_Arm, Aux, A, B, Z,Up_DownRouting options for inputs Any module PFI, analog trigger, many internalsignalsFIFO Dedicated 127-sample FIFOFrequency GeneratorNumber of channels1Base clocks20 MHz, 10 MHz, 100 kHzDivisors 1 to 16 (integers)Base clock accuracy50 ppmOutput Any module PFI terminalNI cDAQ-9184 Specifications| © National Instruments| 3Module PFI CharacteristicsFunctionality Static digital input, static digital output, timinginput, and timing outputTiming output sources3Many analog input, analog output, counter,digital input, and digital output timing signals Timing input frequency0 MHz to 20 MHzTiming output frequency0 MHz to 20 MHzDigital TriggersSource Any module PFI terminalPolarity Software-selectable for most signalsAnalog input function Start Trigger, Reference Trigger,Pause Trigger, Sample Clock,Sample Clock TimebaseAnalog output function Start Trigger, Pause Trigger, Sample Clock,Sample Clock TimebaseCounter/timer function Gate, Source, HW_Arm, Aux, A, B, Z,Up_DownModule I/O StatesAt power-on Module-dependent. Refer to the documentationfor each C Series module.Network InterfaceNetwork protocols TCP/IP, UDPNetwork ports used HTTP:80 (configuration only), TCP:3580;UDP:5353 (configuration only), TCP:5353(configuration only); TCP:31415; UDP:7865(configuration only), UDP:8473 (configurationonly)Network IP configuration DHCP + Link-Local, DHCP, Static,Link-Local3Actual available signals are dependent on type of installed C Series module.4| | NI cDAQ-9184 SpecificationsHigh-performance data streams7Data stream types available Analog input, analog output, digital input,digital output, counter/timer input,counter/timer output, NI-XNET4Default MTU size1500 bytesJumbo frame support Up to 9000 bytesEthernetNetwork interface1000 Base-TX, full-duplex; 1000 Base-TX,half-duplex; 100 Base-TX, full-duplex;100 Base-TX, half-duplex; 10 Base-T,full-duplex; 10 Base-T, half-duplex Communication rates10/100/1000 Mbps, auto-negotiated Maximum cabling distance100 m/segmentPower RequirementsCaution The protection provided by the NI cDAQ-9184 chassis can be impaired ifit is used in a manner not described in the NI cDAQ-9181/9184/9188/9191 UserManual.Note Some C Series modules have additional power requirements. For moreinformation about C Series module power requirements, refer to the documentationfor each C Series module.Note Sleep mode for C Series modules is not supported in the NI cDAQ-9184.V oltage input range9 V to 30 VMaximum power consumption515 W4When a session is active, CAN or LIN (NI-XNET) C Series modules use a total of two data streams regardless of the number of NI-XNET modules in the chassis.5Includes maximum 1 W module load per slot across rated temperature and product variations.NI cDAQ-9184 Specifications| © National Instruments| 5Note The maximum power consumption specification is based on a fully populatedsystem running a high-stress application at elevated ambient temperature and withall C Series modules consuming the maximum allowed power.Power input connector 2 positions 3.5 mm pitch mini-combicon screwterminal with screw flanges, SauroCTMH020F8-0N001Power input mating connector Sauro CTF020V8, Phoenix Contact 1714977,or equivalentPhysical CharacteristicsWeight (unloaded)Approximately 643 g (22.7 oz)Dimensions (unloaded)178.1 mm × 88.1 mm × 64.3 mm(7.01 in. × 3.47 in. × 2.53 in.) Refer to thefollowing figure.Screw-terminal wiringGauge0.5 mm 2 to 2.1 mm2 (20 AWG to 14 AWG)copper conductor wireWire strip length 6 mm (0.24 in.) of insulation stripped from theendTemperature rating85 °CTorque for screw terminals0.20 N · m to 0.25 N · m (1.8 lb · in. to2.2 lb · in.)Wires per screw terminal One wire per screw terminalConnector securementSecurement type Screw flanges providedTorque for screw flanges0.20 N · m to 0.25 N · m (1.8 lb · in. to2.2 lb · in.)If you need to clean the chassis, wipe it with a dry towel.6| | NI cDAQ-9184 SpecificationsFigure 1. NI cDAQ-9184 Dimensions30.6 mm 47.2 mm Safety VoltagesConnect only voltages that are within these limits.V terminal to C terminal30 V maximum, Measurement Category IMeasurement Category I is for measurements performed on circuits not directly connected to the electrical distribution system referred to as MAINS voltage. MAINS is a hazardous liveNI cDAQ-9184 Specifications | © National Instruments | 7electrical supply system that powers equipment. This category is for measurements of voltages from specially protected secondary circuits. Such voltage measurements include signal levels, special equipment, limited-energy parts of equipment, circuits powered by regulatedlow-voltage sources, and electronics.Caution Do not connect the system to signals or use for measurements withinMeasurement Categories II, III, or IV.Note Measurement Categories CAT I and CAT O (Other) are equivalent. These testand measurement circuits are not intended for direct connection to the MAINsbuilding installations of Measurement Categories CAT II, CAT III, or CAT IV.Environmental-20 °C to 55 °C6Operating temperature (IEC 60068-2-1and IEC 60068-2-2)Caution To maintain product performance and accuracy specifications when theambient temperature is between 45 and 55 °C, you must mount the chassishorizontally to a metal panel or surface using the screw holes or the panel mount kit.Measure the ambient temperature at each side of the CompactDAQ system 63.5 mm(2.5 in.) from the side and 25.4 mm (1.0 in.) from the rear cover of the system. Forfurther information about mounting configurations, go to /info and enterthe Info Code cdaqmounting.-40 °C to 85 °CStorage temperature (IEC 60068-2-1 andIEC 60068-2-2)Ingress protection IP 30Operating humidity (IEC 60068-2-56)10% to 90% RH, noncondensingStorage humidity (IEC 60068-2-56)5% to 95% RH, noncondensingPollution Degree (IEC 60664)2Maximum altitude5,000 mIndoor use only.6When operating the NI cDAQ-9184 in temperatures below 0 °C, you must use the PS-15 powersupply or another power supply rated for below 0 °C.8| | NI cDAQ-9184 SpecificationsHazardous LocationsU.S. (UL)Class I, Division 2, Groups A, B, C, D, T4;Class I, Zone 2, AEx nA IIC T4Canada (C-UL)Class I, Division 2, Groups A, B, C, D, T4;Class I, Zone 2, Ex nA IIC T4Europe (ATEX) and International (IECEx)Ex nA IIC T4 GcShock and VibrationTo meet these specifications, you must direct mount the NI cDAQ-9184 system and affix ferrules to the ends of the terminal lines.Operational shock30 g peak, half-sine, 11 ms pulse (Tested inaccordance with IEC 60068-2-27. Test profiledeveloped in accordance withMIL-PRF-28800F.)Random vibrationOperating 5 Hz to 500 Hz, 0.3 g rmsNon-operating 5 Hz to 500 Hz, 2.4 g rms (Tested in accordancewith IEC 60068-2-64. Non-operating testprofile exceeds the requirements ofMIL PRF-28800F, Class 3.)Safety and Hazardous Locations StandardsThis product is designed to meet the requirements of the following electrical equipment safety standards for measurement, control, and laboratory use:•IEC 61010-1, EN 61010-1•UL 61010-1, CSA C22.2 No. 61010-1•EN 60079-0:2012, EN 60079-15:2010•IEC 60079-0: Ed 6, IEC 60079-15; Ed 4•UL 60079-0; Ed 6, UL 60079-15; Ed 4•CSA 60079-0:2011, CSA 60079-15:2012Note For UL and other safety certifications, refer to the product label or the OnlineProduct Certification section.NI cDAQ-9184 Specifications| © National Instruments| 9Electromagnetic CompatibilityThis product meets the requirements of the following EMC standards for electrical equipment for measurement, control, and laboratory use:•EN 61326-1 (IEC 61326-1): Class A emissions; Basic immunity•EN 55011 (CISPR 11): Group 1, Class A emissions•EN 55022 (CISPR 22): Class A emissions•EN 55024 (CISPR 24): Immunity•AS/NZS CISPR 11: Group 1, Class A emissions•AS/NZS CISPR 22: Class A emissions•FCC 47 CFR Part 15B: Class A emissions•ICES-001: Class A emissionsNote In the United States (per FCC 47 CFR), Class A equipment is intended foruse in commercial, light-industrial, and heavy-industrial locations. In Europe,Canada, Australia and New Zealand (per CISPR 11) Class A equipment is intendedfor use only in heavy-industrial locations.Note Group 1 equipment (per CISPR 11) is any industrial, scientific, or medicalequipment that does not intentionally generate radio frequency energy for thetreatment of material or inspection/analysis purposes.Note For EMC declarations and certifications, and additional information, refer tothe Online Product Certification section.CE ComplianceThis product meets the essential requirements of applicable European Directives, as follows:•2014/35/EU; Low-V oltage Directive (safety)•2014/30/EU; Electromagnetic Compatibility Directive (EMC)•2014/34/EU; Potentially Explosive Atmospheres (ATEX)Online Product CertificationRefer to the product Declaration of Conformity (DoC) for additional regulatory compliance information. To obtain product certifications and the DoC for this product, visit / certification, search by model number or product line, and click the appropriate link in the Certification column.10| | NI cDAQ-9184 SpecificationsEnvironmental ManagementNI is committed to designing and manufacturing products in an environmentally responsible manner. NI recognizes that eliminating certain hazardous substances from our products is beneficial to the environment and to NI customers.For additional environmental information, refer to the Minimize Our Environmental Impact web page at /environment. This page contains the environmental regulations and directives with which NI complies, as well as other environmental information not included in this document.Waste Electrical and Electronic Equipment (WEEE) EU Customers At the end of the product life cycle, all NI products must bedisposed of according to local laws and regulations. For more information abouthow to recycle NI products in your region, visit /environment/weee.电子信息产品污染控制管理办法（中国RoHS）中国客户National Instruments符合中国电子信息产品中限制使用某些有害物质指令(RoHS)。

效能评估英语Performance evaluation is an important process for any organization as it allows for the measurement and improvement of employee effectiveness and productivity. This evaluation often includes various methods such as feedback, goal setting, and performance appraisals. In this essay, I will discuss the importance of performance evaluation and some effective methods for conducting it.Firstly, performance evaluation provides feedback to employees, which is crucial for their growth and development. By providing constructive criticism and highlighting areas for improvement, employees can learn from their mistakes and strive to enhance their performance. Feedback also serves as a means of recognition and reward for a job well done, boosting employee morale and motivation.Secondly, performance evaluation helps in setting clear and measurable goals for employees. By defining their objectives and expectations, employees can prioritize their tasks and work towards achieving them. Setting goals also allows managers to monitor progress and provide necessary support and guidance, ensuring that employees are on track.Thirdly, performance evaluation facilitates the identification of training and development needs. By assessing employee performance, managers can identify areas where additional training or skill development is required. This can help in creating targeted training programs and workshops to bridge the skill gap and improve overall employee performance.There are several effective methods for conducting performance evaluation. One method is the 360-degree feedback, where feedback is gathered from multiple sources including peers, subordinates, and customers. This comprehensive feedback provides a holistic view of an employee's performance and helps in identifying areas for improvement. Another method is the Management by Objectives (MBO), where employees and managers jointly set goals and objectives that are specific, measurable, achievable, relevant, and time-bound (SMART). This method ensures that employees are aligned with the organization's objectives and allows for regular monitoring of progress.Performance appraisals are also commonly used to evaluate employees' performance. This method involves reviewing an employee's performance over a specific period and assessing their strengths and weaknesses. It provides an opportunity for managers to discuss career aspirations, provide feedback, and set development plans for the future.In conclusion, performance evaluation is crucial for organizational success as it helps in measuring and improving employee effectiveness and productivity. It provides feedback, sets clear goals, and identifies training needs. By using effective methods such as 360-degree feedback, MBO, and performance appraisals, organizations can ensure a fair and comprehensive evaluation of employee performance.。

Bayesian Group-Sparse Modeling and Variational InferenceS.Derin Babacan,Member,IEEE,Shinichi Nakajima,and Minh N.Do,Fellow,IEEEAbstract—In this paper,we present a general class of multi-variate priors for group-sparse modeling within the Bayesian framework.We show that special cases of this class correspond to multivariate versions of several classical priors used for sparse modeling.Hence,this general prior formulation is helpful in analyzing the properties of different modeling approaches and their connections.We derive the estimation procedures with these priors using variational inference for fully Bayesian estimation. In addition,we discuss the differences between the proposed in-ference and deterministic inference approaches with these priors. Finally,we show theflexibility of this modeling by considering several extensions such as multiple measurements,within-group correlations,and overlapping groups.Index Terms—Bayes methods,group-sparsity,variational inference.I.I NTRODUCTIONW E consider the general linear model given by(1) where observations of the original unknown signal are taken with an measurement matrix(or dictionary),and represents the noise.This paper is concerned with the problem offinding an estimate of the un-known signal from the observations.Generally,the case of interest is the regime,which makes the problem challenging and requires appropriate modeling of the unknown signal.Manuscript received May20,2012;revised January07,2014;accepted March28,2014.Date of publication April22,2014;date of current version May09,2014.The associate editor coordinating the review of this manuscript and approving it for publication was Dr.Sukeyman S.Kozat.S.D.Babacan acknowledges the Beckman Postdoctoral fellowship from University of Illinois at Urbana-Champaign.S.Nakajima thanks the support from MEXT Kakenhi 23120004.M.N.Do acknowledges the support of the National Science Foundation grant CCF09-64215.S.D.Babacan was with the Beckman Institute for Advanced Science and Technology,University of Illinois at Urbana-Champaign,Urbana,IL,USA.He is now with Google,Incorporated,Mountain View,CA94043USA(e-mail: dbabacan@).S.Nakajima is with the Optical Research Laboratory,Nikon Corporation, Tokyo140-8601,Japan(e-mail:nakajima.s@nikon.co.jp).M.N.Do is with the Department of Electrical and Computer Engineering and the Beckman Institute for Advanced Science and Technology,Uni-versity of Illinois at Urbana-Champaign,Urbana,IL61801USA(e-mail: minhdo@).Color versions of one or more of thefigures in this paper are available online at .Digital Object Identifier10.1109/TSP.2014.2319775Problems of the general form(1)are very common in signal processing,statistics,neuroscience and machine learning. Typical applications include compressive sensing[1],sparse representation[2]–[4],super resolution[5],source localization [6],variable/model selection and prediction[7],among many others.A general design principle in these approaches is spar-sity,which amounts tofinding the most important components of and suppressing the elements with relatively lower im-portance.In this design,the unknown vector is assumed to contain a small number of nonzero elements,while the majority of the components are zero.This assumption is translated into the optimization problem forfinding using sparsity-pro-moting penalty functions,of which the most common example is the-norm based formulation given by(2) This formulation is commonly referred to as basis pursuit de-noising[4]and is related to lasso[8].It implicitly models the noise as zero-mean white Gaussian distributed with variance ,and is the regularization parameter controlling the strength of the enforced sparsity.A large number of optimiza-tion methods have been developed for solving(2)[4],[8]–[11]. In addition,different sparse signal models have been proposed extending the-norm to the more general-norm with[9],[11].In the traditional sparse modeling,the sparsity constraint is imposed on individual components of.Recently,a different modeling approach has emerged where sparsity is enforced on groups instead of the individual components.This group-sparse (also called block-sparse)approach is a natural generalization of the traditional sparse modeling methods.It effectively models the structural properties of the signal by clustering relevant signal components together,such that dependencies between signal components are taken into account.It is also shown to lead to higher performance in pruning out irrelevant components compared to independent modeling of the coeffi-cients[12].Group-sparsity has recently been considered in compressive sensing[12]–[16]and machine learning[17]–[21],and is also closely related to signal modeling within union of subspaces [12],[22]–[24].It has rapidly found applications in, e.g., imaging[25],[26]and network analysis[27],demonstrating promising performance.A general optimization formulation for group-sparse regular-ization is(3)1053-587X©2014IEEE.Personal use is permitted,but republication/redistribution requires IEEE permission.See /publications_standards/publications/rights/index.html for more information.where denotes the combined-norm with(4) where denotes the th group,and is the number of groups. Each group contains elements,such that if the groups are not overlapping.It is clear that this formulation includes the traditional-based formulation as a special case (when).The optimization problem(3)is similar to the-based opti-mization,and thus some-based approaches can be applied to this problem with some modifications.Deterministic methods directly addressing the problem(3)have been developed in [3],[28],[29],and in group-lasso methods[21],[20].Several Bayesian approaches have been developed for group-sparse modeling:the Bayesian group-lasso[30]proposed to use multivariate Laplace priors on separate groups,and provided a sampling scheme for inference.A similar group-sparse prior is used in covariance estimation problem in[18].In[17],[31], Laplacian scale mixtures have been used for the construction of the group-sparse prior,and the inference is performed using expectation-maximization(EM).An important issue in all sparse reconstruction problems is choosing the regularization parameters and.Clearly, optimizing(3)jointly with respect to them is not suitable since it results in the trivial solution.A similar problem is encountered when the problem is converted to weighted least squares problems,as in iteratively reweighted least squares (IRLS)with-priors[9],[11],[32].Deterministic heuristic methods are devised for parameter estimation,such as L-curves [33]or penalizing the trivial solution[9].A more systematic approach can be obtained using Bayesian inference,as shown in this article.In this paper,we present a Bayesian approach for group-sparse modeling and ing a normal variance mix-ture formulation,we present the hierarchical construction of a general signal prior suitable for modeling group-sparse signals. This general signal prior contains a large class of distributions as special cases,obtained via different selections of distributions in the hierarchical ing this general formulation, we explore different options for group-sparse modeling,ana-lyze their connections,and their sparsity-enforcing properties. We show that some of the special cases of this generalized prior correspond to several standard models used in the sparse and group-sparse reconstruction literature.For estimation using this class of priors,we provide the hierarchical inference rules using the variational Bayesian(VB)approach for a fully-Bayesian es-timation(i.e.,including algorithmic parameters).We compare the proposed inference with deterministic inference approaches, and show the thresholding properties of different priors both in deterministic and Bayesian frameworks.Finally,we consider several extended modeling possibilities within Bayesian group-sparse modeling,such as within-group correlations and over-lapping groups,and consider the multiple measurement vector case.The rest of this paper is organized as follows.Section II pro-vides the hierarchical construction of the generalized group-sparse prior using normal variance mixtures.We also derive its special cases and show their properties.In Section III,we de-velop fully-Bayesian inference methods using these priors via variational Bayesian approximation.Properties of the modeling and inference in comparison with deterministic approaches are discussed in Section IV.Several extensions to Bayesian group-sparse modeling are provided in Section V.Empirical evalua-tion of different aspects of the group-sparse modeling are pre-sented in Section VI,and conclusions are drawn in Section VII.II.B AYESIAN G ROUP-S PARSE M ODELINGThe Bayesian modeling of(1)requires the definition of a joint distribution of all unknown and observed quantities.This joint distribution typically includes the conditional distribution for the observations,and a prior that models the characteristics of the unknown signal.In the following,wefirst present a class of distributions suitable for group-sparse modeling of using variance mixtures of Gaussian distributions.We then de-rive its special cases and show the connections between them and models proposed in the literature.Finally,we complete the Bayesian model by specifying the observation model and hy-perpriors assigned to the parameters of all distributions.We use the following notation throughout this paper.Vectors are denoted by small-case bold letters,while matrices are in capital bold letters.is a diagonal matrix with vector as its diagonal,and denotes the expectation with respect to the corresponding distribution.A.Signal ModelsFor modeling the unknown signal,wefirst define groups of coefficients such that the vector contains signal coef-ficients assigned to group.The case with, corresponds to independent sparse modeling of the coefficients. Assuming a priori independence between groups,we express the signal prior as(5) where is the vector containing all.Independent groups arise naturally in many applications(e.g.,multi band signal recon-struction[34],sampling signals that lie in a union of subspaces [22],microarray analysis[30]),and group independence is a good approximation in many others.Note also that the infer-ence scheme adopted in this work makes the groups dependent a posteriori(see Section III-A).Sparsity is enforced on each group via the conditional priors.For their representa-tion,we use the normal variance mixture model[35](also called scale mixtures of Gaussians[36],[37]).Specifically,we repre-sent each group as(6) where and is a standard multivariate Gaussian variable, i.e.,with a zero vector of length andthe identity matrix.It is clear that given,isa multivariate Gaussian variable with zero mean and variance,that is(7) Notice that the coefficients within each group are not indepen-dent.The marginal probability distribution of can be found by integrating out the latent variables as(8) Here,is called the mixing distribution and determines the form of the marginal distribution.Normal variance mixtures have been extensively used in the literature for representing a large number of distributions,and for deriving efficient inference procedures for parameter esti-mation[36]–[40].A variety of distributions can be represented in this fashion by different selection of the mixing distribution .In this paper,for the mixing distribution we consider the generalized inverse Gaussian(GIG)distribution(9) where is the modified Bessel function of the second kind. The moments of this distribution are given by[41](10) With this mixing density,the marginal distribution of is found from(8)as the generalized hyperbolic(GH)distribution [35](11)In this paper,we chose the GIG distribution as the mixing dis-tribution as it includes a fairly broad class of distributions com-monly used as hyperpriors,and the resulting marginal distribu-tion,the GH distribution,again covers a large number of distri-butions as special cases.Due to this generalization,we are able to analyze the connections between different modeling strate-gies.As we shall see in the following,several special cases cor-respond to standard priors commonly used in sparse modeling. To see the rich family of distributions that can be obtained from the GH distribution,distributions obtained with varying values of,and are depicted in Figs.1–3.It can be seen that both the central and tail behavior can be varied using dif-ferent parameter values,and as will be shown later,the resulting distributions have different estimation characteristics.In the fol-lowing,we consider the special cases of the GH distribution at the limit parameter values,along with the mixing distributions.Let usfirst give some expressions on asymptotic approxima-tions of the modified Bessel function that will be useful:(12)(13)(14)(15) and for integer,(16) 1)McKay’s Bessel Function Distribution:Whenwith,the mixing GIG distribution reduces to the gamma distribution,given by(17) The corresponding marginal distribution is(18) which is McKay’s Bessel function distribution[42]–[44](also called multivariate variance-gamma[38],multivariate general-ized Laplace[43],or multivariate K distribution[39],[40]). We now consider two special cases of(18)that are related to the Laplace distribution.In the case,the mixing distri-bution becomes the exponential distribution(19) such that the marginal becomes(20) To see the relation with the univariate Laplace distribution,we can use(16)and rewrite(20)for odd as(21) The similarity to the univariate case can be seen from the ex-ponential term,and noticing that all other terms vanish with .Note,however,that there are additional terms that are power functions of.A more directly related case can beFig.1.Generalized hyperbolic distributions(a)and log-distributions(b)with varying,when,(,the cross-section is shown).Fig.2.Generalized hyperbolic distributions(a)and log-distributions(b)with varying,when,(,the cross-section is shown).Fig.3.Generalized hyperbolic distributions(a)and log-distributions(b)with varying,when,(,the cross-section is shown).TABLE IS UMMARY OF D ISTRIBUTIONS AND P ARAMETER ESTIMATESfound by the selection ,which simpli fies (18)using (16)as(22)in which case the mixing distribution is a gamma distribution given by(23)Both distributions (20)and (22)were termed as multivariate Laplace distributions in the literature:the form in (20)is used in [39],[45]due to the similarity of the hierarchical structure to the univariate case,and (22)is used in the Bayesian group-lasso method [30]due to the similarity of the marginal distributions.Here we will use the term multivariate Laplace for the distri-bution in (22)since it has an estimation behavior similar to the univariate case (see Section IV-A).The distribution in (20)will be referred to as McKay .Notice that both distributions reduce to the univariate Laplace distribution when .It is also possible to integrate out from by as-signing a gamma hyperprior on .When ,the corresponding marginal has a closed form and is given by(24)which is the multivariate version of the generalized double Pareto distribution [46],[47].2)Multivariate Student’s :Whenwith ,we have the inverse gamma distribution as the mixing density(25)The corresponding marginal is given by(26)which is a multivariate Student’s t distribution withde-grees of freedom.Finally,when ,and ,we have the Jeffrey’s non-informative prior .In this case,the marginaldistribution becomes(27)Fig.4.McKay,Laplace,Student’s t,Jeffrey’s and Gaussian log-distributions(,the cross-section is shown).In summary,the variance mixture model with the GIG mix-ture distribution includes a number of classical distributions as special cases at the limiting values of its parameters.In the fol-lowing,we mainly limit our discussion to the four distributions described above,i.e.,multivariate McKay ,Laplace,Student’s t distributions and Jeffrey’s prior.These distributions along with the corresponding parameter selections are summa-rized in Table I.The log-distributions for all cases are shown in Fig.4,along with the Gaussian distribution.It is evident that all distributions have heavy-tails,which is generally considered to be a desirable property for enforcing sparsity and variable plete ModelAfter the signal model is de fined,we complete the Bayesian model characterization by modeling the observations in (1).Assuming independent Gaussian noise with zero mean and vari-ance equal to ,the conditional distribution is expressed as(28)with a conjugate gamma prior placed onas(29)A prior is called conjugate if it leads to a posterior distribution that has the same functional form as the prior [7].The use of con-jugate priors signi ficantly simplify the form of posterior distri-bining(28),(29)and the hierarchical signal prior (7)and(9),we define the joint probability distribution as(30) where,,,are vectors containing,,,and respectively.The hyperprior is used to model the pa-rameters and for their estimation,and will be discussed in Section III-C.III.V ARIATIONAL I NFERENCEBayesian inference is based on the posterior distribution,where denotes the set of all un-knowns such that.However,as in many multidimensional problems,the Bayesian model defined with the joint distribution in(30)does not allow for exact inference as the marginal distribution is intractable.Therefore, approximation methods must be used for the inference.In the following,we use the variational Bayesian(VB)approximation [48],[49],which has attractive computational properties along with high estimation performance.With the definition of the joint distribution in(30),the variational Bayes method provides a distribution that approximates the posterior. Specifically,is found by minimizing the Kullback-Leibler (KL)divergence between the approximation and the unknown posterior as[48],[49](31)(32) where is the joint probability distribution given in(30). To solve this optimization,the only assumption needed is an appropriate factorization of.Here we use the mean-field approximation[48]with(33)Using this factorization in(32),the distributions of each variable is found as[48],[49](34)(35)where denotes the set with removed.Individual dis-tributions are updated by(35)at each iteration byfixing the remaining distributions,which corresponds to an alternating minimization of the KL divergence in(32).This it-erative procedure is repeated until the KL distance converges. The VB method is a generalization of the maximum a pos-teriori(MAP)and expectation-maximization(EM)methods. The EM estimates can be found by restricting some distribu-tions to be degenerate,i.e.,delta distributions at a par-ticular value.On the other hand,MAP solutions can be found by restricting all of the distributions to be degenerate.When a distribution is degenerate,it can be shown from(32)that its corresponding estimation amounts to minimizing the negative expected log joint distribution,which re-duces to the log joint distribution in the case We will discuss the MAP estimation in more detail in Section IV.In the following subsections,we provide the explicit forms of the update rules for all unknowns.For notational simplicity, the optimal distributions are denoted by instead of.A.Signal EstimateFrom(35),the posterior approximation of is found as a multivariate Gaussian(36) with parameters(37)(38)(39) with,with each repeated times.1It can be seen from(38)that except when,the groups are a posteriori dependent,despite the a priori independence assumption in(5).Sparsity in the groups occur when particular variables,in which case the th group is pruned out from the signal estimate.2Notice also the estimation of requires the inversion of an matrix using(38),and an matrix using(39).B.Estimation of the Variance ParametersThe crucial part of(37)is the estimates of,which control the sparsity and hence the structure of the signal estimate.Here we derive the estimation rules for the general case with the GIG hyperprior,from which the special cases can easily be obtained. First,with some algebra,it can be derived from(35)in combi-nation with(33)that the distribution factorizes over, such that(40) Therefore,in the following we provide the update rules for each ing(35),wefind the approximate posterior from(7)and(9)as a GIG distribution(41) with the expectation computed as(42) 1Notice that this assumes non-overlapping groups;overlapping groups willbe discussed later.2The modeling used in this paper does not allow for exact sparsity.However, sparsity occurs in practice when estimates become very large such that the coefficients in the th group are from zero.where denotes the submatrix of corresponding to the th group.The posterior estimate of can be calculated by the moments of this distribution in(10)as(43) The update rules for the limiting cases can be found from this general form,and are shown in the third column of Table I. C.Estimation of the Hyperparameters andNotice that in the general case(43),the posterior estimate of contains the hyperparameters,,and,which deter-mine the shape of the enforced distribution on.With the vari-ational approximation,their posterior distributions can be esti-mated using(35)as well,with the appropriate selection of the hyperpriors,and(or with a joint hyperprior ).However,in the general case with GIG mixing distribution,the joint estimation of all,and is chal-lenging:the estimation of requires numerical solutions(in-stead of analytical closed form updates),and when all parame-ters are jointly estimated,the accuracy greatly depends on the initial estimates.Therefore,we instead provide hyperparameter estimates of and in the special cases,and leave as a free parameter.1)McKay’s Bessel Function Distribution:Recall that withand,we have the gamma distribution(17)as the mixing density.As the corresponding hyperprior for,we choose the conjugate gamma distribution(44) with the shape parameter and the inverse scale parameter. The posterior becomes(45) with the corresponding update(46) The moment can be found from(41)using(10).2)Multivariate Student’s:When with, the mixing distribution(25)is an inverse gamma distribution in terms of,but it is a gamma distribution with respect to the parameter.Hence we choose the gamma distribution that is conjugate for(47) The posterior distribution is found as a gamma distribution(48)with mean(49)D.Estimation of the Noise VarianceThe Bayesian methodology allows for the estimation of the noise variance as ing the prior in(29),the posterior of becomes a gamma distribution,and can be estimated using its mean as(50)with the expectation given by(51)E.SummaryThe signal priors presented in the previous sections,along with the corresponding mixing distributions and variational es-timation rules are summarized in Table I.The algorithm alter-nates between estimating the signal using(37),and the vari-ances and hyperparameters,using the equations shown in Table I,according to the selected signal distribution.The normal variance mixture with the GIG mixing distribu-tion is extremelyflexible,and encompasses a large family of distributions some of which can be used for modeling group-sparse signals.Other,non-standard,distributions can also be obtained by further extending the hierarchical construction and marginalization.The advantages of using the variance mixture formulation are the tractable properties of the Gaussian distri-bution obtained for the signal estimate in(36)and the conjugate prior mechanism that allows for closed-form estimation of the parameters.In this work,we used a three-level hierarchical estimation procedure,involving the estimation of,,and in alternating fashion.Instead,two-level hierarchical estimation procedures can be devised using the marginal distributionsand appropriate hyperpriors on and (therefore bypassing the estimation of).This approach is a generalization of Laplacian scale mixtures[17].However, this approach brings some difficulties:First,the marginal distributions have complicated forms and the corresponding conjugate hyperpriors on and are hard tofind.Second,the marginal distributions generally do not allow for closed form updates of the posterior mean.Finally,the posterior mean updates of and in general require expectations that do not have closed forms.Hence,fully-Bayesian inference with this two-level hierarchy is generally hard.Note,however,that if parameter estimation is not desired,deterministic approaches can be used(see Section IV)with relative ease with some forms of the marginal distributions,e.g.,the Laplace distribution. This approach is closely related to reweighted-minimization schemes[11],[32]and the EM approach presented in[17].IV.C OMPARISON W ITH D ETERMINISTIC E STIMATION The signal priors considered in Section II-A can also be used in a deterministic maximum a posteriori(MAP)framework,which is commonly encountered in the ing a de-terministic framework allows us to show some interesting con-nections between different signal priors and also compare and demonstrate some properties of the variational Bayesian esti-mation described before.For the MAP optimization with the Bayesian model in this paper,two approaches can be considered.A.MAP Estimation Using Marginal DistributionsBy forming the joint probability distributionusing the observation model in (28)and the generalized hyperbolic distribution in(11)as the signal prior,and applying a log-transform,we obtain the MAP estimate as(52)(53) Note that the mode of the posterior distribution is sought within this formulation.In the general case with nonzero,,and, closed form updates for cannot be found and numerical so-lutions are required.However,closed-form updates can easily be found in the case of multivariate Laplace(22)and t-distribu-tions(26),and Jeffrey’s prior(27).In the case of multivariate Laplace priors,the optimization problem becomes(54) which is equivalent to the-norm formulation in(3).With the multivariate t-distributions,we have(55) Although the connection between this problem and the -norm formulation in(3)is not immediately clear,they are in fact related.Consider the following-norm based group-sparse estimation problem(56) with.Notice that recovers the-norm minimization in(3).Using the formula(57) it can be seen that the multivariate t prior is a limiting case of the -norm based group-sparse estimation procedure.In addition, in the case of Jeffrey’s priors,the penalty function is the limiting case of as.In this regard,the Laplace and t-distri-butions can be thought to be at the opposite ends of the-norm penalties;while Laplace prior leads to an-based method,t-dis-tributions enforce sparsity similar to-norms.The generalized -norm based formulation with can be constructed using Gaussian variance mixtures as well,but the mixing dis-tribution is an alpha-stable distribution without a closed-form, which makes the inference very hard.Using the MAP formulation in(53),we can also analyze the thresholding properties of different distributions when is or-thonormal,i.e.,.In this case,the problem decouples into optimization problems(the groups become independent), and can be solved for each group separately as(58) The thresholding functions for different distributions forfixed, and are shown in Fig.5(a).The multivariate Laplace distri-bution has a soft-thresholding behavior(similar to the univariate case),while the behavior of all other distributions is similar to hard-thresholding,including the McKay distribution. In addition,the multivariate Laplace and McKay priors have a constant bias independent of the signal value.Student’s t and Jeffrey’s priors do not have this disadvantage;the bias con-verges to zero as the signal magnitude increases.On the other hand,the Laplace prior is continuous at the thresholding value, whereas the others have discontinuities,which is generally con-sidered as a disadvantage since small changes in the data might lead to large changes in the estimation[50].In comparison,the thresholding functions obtained by the variational Bayesian inference described in the previous sections is shown in Fig.5(b).It can be observed that all thresh-olding curves become smoother,and in fact,none of the priors lead to a thresholding rule:the estimates are only“almost”sparse,i.e.,they have very small values in an interval but are never exactly zero.Interestingly,the thresholding function of the Jeffrey’s prior now exhibits a soft-thresholding behavior while the bias is again converging to zero as the signal mag-nitude increases.On the other hand,the thresholding property of the Laplace and McKay is decreased.However,it should be emphasized that when,and are not constant but also estimated,all priors lead to exact thresholding rules. An important remark is that simultaneous estimation of the parameters and cannot in general be performed using the MAP formulation if the hyperpriors and are not suitably chosen.The objective(53)becomes unbounded from below for some values of parameters,,,,in which case the global minimum is obtained at the trivial solution ,and.Therefore,other methods should be employed,such as cross-validation or L-curves[33].B.Hierarchical EstimationA second method is to use the hierarchical representations of the distributions,and consider the joint minimization problem as(59)。

以下是为⼤家整理的关于殷保群教授个⼈简历范⽂的⽂章，希望⼤家能够喜欢！殷保群，男，教授，博⼠⽣导师。

中国科学技术⼤学教授。

1962年2⽉⽣，1985年7⽉毕业于四川⼤学数学系基础数学专业，随后考⼊中国科学技术⼤学基础数学研究⽣班，1987年7⽉毕业，并留校任教。

1993年5⽉在中国科学技术⼤学数学系应⽤数学专业获得理学硕⼠学位，1998年12⽉在中国科学技术⼤学⾃动化系模式识别与智能系统专业获得⼯学博⼠学位，现在中国科学技术⼤学⾃动化系任教。

长期从事随机系统、系统优化以及信息络系统理论及其应⽤等⽅⾯的研究⼯作，⽬前感兴趣的主要⽅向为Markov决策过程、络建模与优化、络流量分析、媒体服务系统的接⼊控制以及云计算等。

在国内外主要学术刊物上发表学术论⽂100余篇，其中SCI收录10余篇，EI收录30余篇，出版学术专著1部。

曾于2004年4⽉⾄12⽉在⾹港科技⼤学做访问学者。

第xx届（2006年）何潘清漪优秀论⽂获奖者。

⽬前感兴趣的主要研究⽅向：1、离散事件动态系统； 2、Markov决策过程； 3、排队系统； 4、信息络论⽂著作主要著作殷保群，奚宏⽣，周亚平，排队系统性能分析与Markov控制过程,合肥：中国科学技术⼤学出版社，2004.期刊论⽂Yin, B. Q., Guo, D., Huang, J., Wu, X. M., Modeling and analysis for the P2P-based media delivery network, Mathematical and Computer Modelling (2011), doi:10.1016/j.mcm.2011.10.043. (SCI 收录, JCR II 区) Yin, B. Q., Lu, S., Guo, D., Analysis of Admission Control in P2P-Based Media Delivery Network Based on POMDP, International Journal of Innovative Computing, Information and Control, 2011, 7(7B): 4411-4422. (SCI收录, JCR II 区) Kang, Yu, Yin, Baoqun, Shang, Weike, Xi, Hongsheng, Performance sensitivity analysis and optimization for a class of countable semi-Markov decision processes, Proceedings of the World Congress on Intelligent Control and Automation (WCICA2011), June 21, 2011 - June 25, 2011, Taipei, Taiwan. (EI收录20113614311870) Li, Y. J., Yin, B. Q., Xi, H. S., Finding Optimal Memoryless Policies of POMDPs under the Expected Average Reward Criterion, European Journal of Operational Research, 2011, 211(2011): 556-567. (SCI 收录, JCR II 区) 江琦，奚宏⽣，殷保群，事件驱动的动态服务组合策略在线⾃适应优化，控制理论与应⽤，2011, 28(8): 1049-1055. (EI收录20114214431454) Jiang, Q., Xi, H. S., Yin, B. Q., Adaptive Optimization of Timeout Policy for Dynamic Power Management Based on Semi-Markov Control Processes, IET Control Theory and Applications, 2010, 4(10): 1945-1958. (SCI收录) Tang, L., Xi, H. S., Zhu, J., Yin, B. Q., Modeling and Optimization of M/G/1-Type Queueing Networks: An Efficient Sensitivity Analysis Approach, Mathematical Problems in Engineering, 2010, 2010: 1-20. (SCI收录) Shan Lu, Baoqun Yin, Dong Guo, Admission Control for P2P-Based Media Delivery Network, Proceedings of the 29th Chinese Control Conference, July 29-31, 2010, Beijing, China, 2010: 1494-1499. ( EI收录20105113504286) ⾦辉宇，康宇，殷保群，局部Lipschitz系统的采样控制，Proceedings of the 29th Chinese Control Conference, July 29-31, 2010, Beijing, China, 2010: 992-997. ( EI收录20105113504436) 江琦，奚宏⽣，殷保群，络新媒体服务系统事件驱动的动态服务组合，Proceedings of the 29th Chinese Control Conference, July 29-31, 2010, Beijing, China, 2010: 1121-1125. ( EI收录20105113504230) Dong Guo, Baoqun Yin, Shan Lu, Jing Huang, Jian Yang, A Novel Dynamic Model for Peer-to-Peer File Sharing Systems, ICCMS, 2010 Second International Conference on Computer Modeling and Simulation, 2010, 1: 418-422. ( EI收录20101812900175) Jing Huang, Baoqun Yin, Dong Guo, Shan Lu, Xumin Wu, An Evolution Model for P2P File-Sharing Networks, ICCMS, 2010 Second International Conference on Computer Modeling and Simulation, 2010, 2: 361-365. ( EI收录20101712882202) 巫旭敏，殷保群，黄静，郭东，流媒体服务系统中⼀种基于数据预取的缓存策略，电⼦与信息学报，2010，32(10): 2440-2445. (EI 收录20104513372577) 马军，郑烇，殷保群，基于CDN和P2P的分布式络存储系统，计算机应⽤与软件，2010，27(2):50-52. Bao, B. K., Xi, H. S., Yin, B. Q., Ling, Q., Two Time-Scale Gradient Approximation Algorithm for Adaptive Markov Reward Processes, International Journal of Innovative Computing, Information and Control, 2010, 6(2): 655-666. (SCI收录, JCR II 区) Jiang, Q., Xi, H. S., Yin, B. Q., Dynamic File Grouping for Load Balancing in Streaming Media Clustered Server Systems, International Journal of Control, Automation, and Systems, 2009, 7(4): 630-637. (SCI收录) 江琦，奚宏⽣，殷保群，动态电源管理超时策略与随机型策略的等效关系，计算机辅助设计与图形学学报，2009, 21(11): 1646-1651. (EI 收录20095012535449) 唐波，李衍杰，殷保群，连续时间部分可观Markov决策过程的策略梯度估计，控制理论与应⽤，2009，26(7)：805-808. (EI 收录20093712302646) 芦珊，黄静，殷保群，基于POMDP的VOD接⼊控制建模与仿真，中国科学技术⼤学学报，2009，39(9)：984-989. 李洪亮，殷保群，郑诠，⼀种基于负载均衡的数据部署算法，计算机仿真，2009，26(4)：177-181. 鲍秉坤，殷保群，奚宏⽣，基于性能势的Markov控制过程双时间尺度仿真算法，系统仿真学报，2009，21(13)：4114-4119. Jin Huiyu; Yin Baoqun; Ling Qiang; Kang Yu; Sampled-data Observer Design for Nonlinear Autonomous Systems, 2009 Chinese Control and Decision Conference, CCDC 2009, 2009: 1516-1520. ( EI收录20094712469527) ⾦辉宇，殷保群，⾮线性采样系统指数稳定的新条件，控制理论与应⽤，2009，26(8)：821-826. (EI 收录20094512429319) Yin, B. Q., Li, Y. J., Zhou, Y. P., Xi, H. S., Performance Optimization of Semi-Markov Decision Processes with Discounted-Cost Criteria. European Journal of Control, 2008, 14(3): 213-222. (SCI收录) Li, Y. J., Yin, B. Q. and Xi, H. S., Partially Observable Markov Decision Processes and Performance Sensitivity Analysis. IEEE Trans. System, Man and cybernetics-Part B., 2008, 38(6): 1645-1651. (SCI收录, JCR II 区) Tang, B., Tan, X. B., Yin, B. Q. , Continuous-time hidden markov models in network simulation, 2008 IEEE International Symposium on Knowledge Acquisition and Modeling Workshop Proceedings, Wuhan, China, DEC 21-22, 2008: 667-670. (EI收录20092812179753) Bao, B. K., Yin, B. Q., Xi, H. S., Infinite-Horizon Policy-Gradient Estimation with Variable Discount Factor for Markov Decision Process. icicic,pp.584,2008 3rd International Conference on Innovative Computing Information and Control, 2008. ( EI收录************) Chenfeng Xu, Jian Yang, Hongsheng Xi, Qi Jiang, Baoqun Yin, Event-related optimization for a class of resource location with admission control, 2008. IJCNN 2008. (IEEE World Congress on Computational Intelligence). IEEE International Joint Conference on Neural Networks, 1-8 June 2008: 1092 – 1097. ( EI收录************)JinHuiyu;KangYu;YinBaoqun; Synchronization of nonlinear systems with stair-step signal, 2008. CCC 2008. 27th Chinese Control Conference,16-18 July 2008: 459 – 463. ( EI收录************)JiangQi;XiHongsheng;YinBaoqun;XuChenfeng;Anevent-drivendynamicload balancing strategy for streaming media clustered server systems, 2008. CCC 2008. 27th Chinese Control Conference, 16-18 July 2008: 678 – 682. ( EI收录************)⾦辉宇，殷保群，唐波，⾮线性采样观测器的误差分析，中国科学技术⼤学学报，2008, 38(10): 1226-1231. 黄静，殷保群，李俊，基于观测的POMDP优化算法及其仿真，信息与控制，2008, 37(3): 346-351. 马军，殷保群，基于POMDP模型的机器⼈⾏动的仿真优化，系统仿真学报，2008, 20(21): 5903-5906. (EI 收录************)江琦，奚宏⽣，殷保群，动态电源管理超时策略⾃适应优化算法，控制与决策，2008, 23(4): 372-377. (EI 收录************)徐陈锋，奚宏⽣，江琦，殷保群，⼀类分层⾮结构化P2P系统的随机切换模型，控制与决策，2008, 23(3): 263-266. (EI 收录************)徐陈锋，奚宏⽣，殷保群，⼀类混合资源定位服务的优化模型，微计算机应⽤，2008，29(9)：6-11. 郭东，郑烇，殷保群，王嵩，基于P2P媒体内容分发络中分布式节点的设计与实现，电信科学，2008，24(8): 45-49. Tang, H., Yin, B. Q., Xi, H. S., Error bounds of optimization algorithms for semi-Markov decision processes. International Journal of Systems Science, 2007, 38(9): 725-736. (SCI收录) 徐陈锋，奚宏⽣，江琦，殷保群，⼀类分层⾮结构化P2P系统的随机优化，系统科学与数学，2007, 27(3): 412-421. 蒋兆春，殷保群，李俊，基于耦合技术计算Markov链性能势的仿真算法，系统仿真学报，2007, 19(15): 3398-3401. (EI收录************)庞训磊，殷保群，奚宏⽣，⼀种使⽤TCP/ IP 协议实现⽆线传感器络互连的新型设计，传感技术学报，2007, 20(6): 1386-1390. Niu, L. M., Tan, X. B., Yin, B. Q. , Estimation of system power consumption on mobile computing devices, 2007. International Conference on Computational Intelligence and Security, Harbin, China, DEC 15-19, 2007: 1058-1061. (EI收录************)Jiang,Q.,Xi, H. S., Yin, B. Q., Dynamic file grouping for load balancing in streaming media clustered server systems. Proceedings of the 2007 International Conference on Information Acquisition, ICIA, Jeju City, South Korea, 2007:498-503. (EI收录************)徐陈锋, 奚宏⽣, 江琦, 殷保群，⼀类分层⾮结构化P2P系统的随机优化，第2xx届中国控制会议论⽂集，2007: 693-696. (EI收录************)Jiang,Q.,Xi,H.S.,Yin,B.Q.,OptimizationofSemi-MarkovSwitchingState-spaceControl Processes for Network Communication Systems. 第2xx届中国控制会议论⽂集，2007: 707-711. (EI收录************) Jiang, Q., Xi, H. S., Yin, B. Q., Adaptive Optimization of Time-out Policy for Dynamic Power Management Based on SMCP. Proc. of the 2007 IEEE Multi-conference on Systems and Control, Singapore, 2007: 319-324. (EI收录************)Jin,H. Y., Yin, B. Q., New Consistency Condition for Exponential Stabilization of Smapled-Data Nonlinear Systems. 第2xx届中国控制会议论⽂集，2007: 84-87. (EI收录************)江琦，奚宏⽣，殷保群，⽆线多媒体通信适应带宽配置在线优化算法，软件学报， 2007, 18(6): 1491-1500. (EI收录************)Ou,Q.,Jin,Y.D.,Zhou,T.,Wang,B.H.,Yin,B.Q.,Power-law strength-degree correlation from resource-allocation dynamics on weighted networks, Physical Review E, 2007, 75(2): 021102 (SCI收录) Yin, B. Q., Dai, G. P., Li, Y. J., Xi, H. S., Sensitivity analysis and estimates of the performance for M/G/1 queueing systems, Performance Evaluation, 2007, 64(4): 347-356. (SCI收录) 江琦，奚宏⽣，殷保群，动态电源管理的随机切换模型与在线优化，⾃动化学报，2007, 33(1): 66-71. (EI收录************)Zhang,D.L.,Yin,B.Q.,Xi,H.S.,Astate aggregation approach to singularly perturbed Markov reward processes. International Journal of Intelligent Technology, 2006, 2(4): 230-239. 欧晴，殷保群，奚宏⽣，基于动态平衡流的络赋权，中国科学技术⼤学学报，2006, 36(11): 1196-1201.殷保群，李衍杰，周亚平，奚宏⽣，可数半Markov控制过程折扣代价性能优化，控制与决策，2006, 21(8): 933-936. (EI收录************)江琦，奚宏⽣，殷保群，动态电源管理的随机切换模型与策略优化，计算机辅助设计与图形学学报，2006, 18(5): 680-686. (EI收录***********)代桂平，殷保群，李衍杰，奚宏⽣，半Markov控制过程基于性能势仿真的并⾏优化算法，中国科学技术⼤学学报，2006, 36(2): 183-186. 殷保群，李衍杰，唐昊，代桂平，奚宏⽣，半Markov决策过程折扣模型与平均模型之间的关系，控制理论与应⽤，2006, 23(1): 65-68. (EI收录***********)江琦，奚宏⽣，殷保群，半Markov控制过程在线⾃适应优化算法，第2xx届中国控制会议论⽂集，2006: 1066-1071. (ISTP收录BFQ63) Dai, G. P., Yin, B. Q., Li, Y. J., Xi, H. S., Performance Optimization Algorithms based on potential for Semi-Markov Control Processes. International Journal of Control, 2005, 78(11): 801-812. (SCI收录) Zhang, D. L., Xi, H. S., Yin, B. Q., Simulation-based optimization of singularly perturbed Markov reward processes with states aggregation. Lecture Notes in Computer Science, 2005, 3645: 129-138. (SCI 收录) Tang, H., Xi, H. S., Yin, B. Q., The optimal robust control policy for uncertain semi-Markov control processes. International Journal of System Science, 2005, 36(13): 791-800. (SCI收录) 张虎，殷保群，代桂平，奚宏⽣，G/M/1排队系统的性能灵敏度分析与仿真，系统仿真学报，2005, 17(5): 1084-1086. (EI收录***********)陈波，周亚平，殷保群，奚宏⽣，隐马⽒模型中的标量估计，系统⼯程与电⼦技术，2005, 27(6): 1083-1086. (EI收录***********)代桂平，殷保群，李衍杰，周亚平，奚宏⽣，半Markov控制过程在平均准则下的优化算法，中国科学技术⼤学学报，2005, 35(2): 202-207. 殷保群，李衍杰，奚宏⽣，周亚平，⼀类可数Markov控制过程的平稳策略，控制理论与应⽤，2005, 22(1): 43-46. (EI收录***********)Li,Y.J.,Yin,B.Q.,Xi,H.S.,Thepolicygradientestimationofcontinuous-timeHiddenMarkovDecision Processes. Proc. of IEEE ICIA, Hong Kong, 2005. (EI收录************)Sensitivity analysis and estimates of the performance for M/G/1 queueing systems, To Appear in Performance Evaluation, 2006.Performance optimization algorithms based on potential for semi-Markov control processes. International Journal of Control, Vol.78, No.11, 2005.The optimal robust control policy for uncertain semi-Markov control processes. International Journal of System Science, Vol.36, No.13, 2005.A state aggregation approach to singularly perturbed Markov reward processes. International Journal of Intelligent Technology, Vol.2, No.4, 2005.Simulation optimization algorithms for CTMDP based on randomized stationary policies, Acta Automatics Sinica, Vol. 30, No. 2, 2004.Performance optimization of continuous-time Markov control processes based on performance potentials, International Journal of System Science, Vol.34, No.1, 2003.Optimal Policies for a Continuous Time MCP with Compact Action set, Acta Automatics Sinica, Vol. 29, No. 2, 2003. Relations between Performance Potential and Infinitesimal Realization Factor in Closed Queueing Networks, Appl. Math. J. Chinese Univ. Ser. B, Vol. 17, No. 4, 2002.Sensitivity Analysis of Performance in Queueing Systems with Phase-Type Service Distribution, OR Transactions, Vol.4, No.4, 2000.Sensitivity Formulas of Performance in Two-Server Cyclic Queueing Networks with Phase-Type Distributed Service Times, International Transaction in Operational Research, Vol.6, No.6, 1999.Simulation-based optimization of singularly perturbed Markov reward processes with states aggregation. Lecture Notes in Computer Science, 2005.Markov decision problems with unbounded transition rates under discounted-cost performance criteria. Proceedings of WCICA, Vol.1, Hangzhou, China, 2004.排队系统性能分析与Markov控制过程,合肥：中国科学技术⼤学出版社，2004.可数半Markov控制过程折扣代价性能优化. 控制与决策，Vol.21, No.8, 2006.动态电源管理的随机切换模型与策略优化. 计算机辅助设计与图形学学报，Vol.18, No.5, 2006.半Markov决策过程折扣模型与平均模型之间的关系.控制理论与应⽤，Vol.23, No.1, 2006.⼀类可数Markov控制过程的平稳策略. 控制理论与应⽤，Vol.22, No.1, 2005.G/M/1排队系统的性能灵敏度分析与仿真.系统仿真学报，Vol.17, No.5, 2005.M/G/1排队系统的性能优化与算法，系统仿真学报，Vol.16, No.8, 2004.半Markov过程基于性能势的灵敏度分析和性能优化. 控制理论与应⽤，Vol.21, No.6, 2004.半Markov控制过程在折扣代价准则下的平稳策略. 控制与决策，Vol.19, No.6, 2004.Markov控制过程在紧致⾏动集上的迭代优化算法. 控制与决策，Vol.18, No.3, 2003.闭Jackson络的优化中减少仿真次数的算法研究，系统仿真学报，Vol.15, No.3, 2003.M/G/1排队系统的性能灵敏度估计与仿真，系统仿真学报，Vol.15, No.7, 2003.Markov控制过程基于性能势仿真的并⾏优化，系统仿真学报，Vol.15, No.11, 2003.Markov控制过程基于性能势的平均代价策略. ⾃动化学报，Vol.28, No.6, 2002.⼀类受控闭排队络基于性能势的性⽅程.控制理论与应⽤，Vol.19, No.4, 2002.Markov控制过程基于单个样本轨道的在线优化算法.控制理论与应⽤，Vol.19, No.6, 2002.闭排队络当性能函数与参数相关时的性能灵敏度分析，控制理论与应⽤，Vol.19, No.2, 2002.M/G/1 排队系统的性能灵敏度分析，⾼校应⽤数学学报，Vol.16, No.3, 2001.连续时间Markov决策过程在呼叫接⼊控制中的应⽤，控制与决策，Vol.19, 2001.具有不确定噪声的连续时间⼴义系统确保估计性能的鲁棒Kalman滤波器，控制理论与应⽤，Vol.18, No.5, 2001.状态相关闭排队络中的性能指标灵敏度公式，控制理论与应⽤，Vol.16, No.2, 1999.科研项⽬半Markov控制过程基于性能势的优化理论和并⾏算法,2003.1-2005.12,国家⾃然科学基⾦,60274012隐Markov过程的性能灵敏度分析与优化,2006.1-2008.12,国家⾃然科学基⾦, 60574065部分可观Markov系统的性能优化,2005.1-2006.12,安徽省⾃然科学基⾦, 050420301宽带信息运营⽀撑环境及接⼊系统的研制――⼦课题: 流媒体服务器研究及实现, 2005.1-2006.12, 国家863计划,2005AA103320离散复杂系统的控制与优化研究,2006.9-2008.8,中国科学院⾃动化研究所中国科学技术⼤学智能科学与技术联合实验室⾃主研究课题基⾦络新媒体服务系统的建模及其动⼒学⾏为分析研究，2012.01-2015.12，国家⾃然科学基⾦；⾯向服务任务的快速机器视觉与智能伺服控制，2010.01-2013.12，国家⾃然科学基⾦重点项⽬；新⼀代业务运⾏管控协同⽀撑环境的开发，2008.07-2011.06，国家863计划；多点协作的流媒体服务器集群系统及其性能优化，2006.12-2008.12，国家863计划；获奖情况第xx届何潘清漪优秀论⽂奖联系信息办公室地址：电⼆楼223 实验室地址：电⼆楼227 办公室电话：************。

Detailed Modeling of CIGRÉHVDC Benchmark System Using PSCAD/EMTDC and PSB/SIMULINK M.O.Faruque,Student Member,IEEE,Yuyan Zhang,and Venkata Dinavahi,Member,IEEEAbstract—This paper focuses on a comparative study of the mod-eling and simulation of thefirst CIGRÉHVDC benchmark system using two simulation tools PSCAD/EMTDC and PSB/SIMULINK; an interface between them(PSCAD-SIMULINK)has also been im-plemented and used as a simulator.The CIGRÉHVDC system and its controller has been carefully modeled in all three simulation environments so that the differences are parison of steady-state and transient situations have been carried out,and a high degree of agreement in most of the cases has been observed. Index Terms—HVDC transmission,modeling,simulation.I.I NTRODUCTIONT HE DESIGN,analysis,and operation of complex ac-dc systems require extensive simulation resources that are accurate and reliable.Analog simulators,long used for studying such systems,have reached their physical limits due to the increasing complexity of modern systems.Currently,there are several industrial grade digital time-domain simulation tools available for modeling ac-dc power systems.Among them,some have the added advantages of dealing with power electronics apparatus and controls with more accuracy and efficiency.PSCAD/EMTDC[1]and PSB/SIMULINK[2]are such two simulators that are being increasingly used in the industry as well as in the universities.Both programs allow the user to construct schematic diagram of electrical networks, run the simulation,and produce the results in a user-friendly graphical environment.Furthermore,several real-time digital simulators use models or the graphical front-end that are similar to PSCAD/EMTDC and PSB/SIMULINK.The objective of this paper is to report a detailed compar-ison between PSCAD/EMTDC and PSB/SIMULINK for the modeling and simulation of ac-dc power systems.In a digital simulator,the system model and the algorithm used to solve that model directly affect the accuracy and consistency of the sim-ulation results.Therefore,based on the objective of the study, careful attention should be given to the selection of the model, the numerical solver,and the algorithm.A comparative study among simulation tools will help in identifying the pros and cons that the programs inherit.For the last two decades,digital simulators have been widely used for the simulation of HVDCManuscript received September1,2004;revised December4,2004.This work was supported by the Natural Sciences and Engineering Research Council (NSERC)of Canada and the University of Alberta.Paper no.TPWRD-00406-2004.The authors are with the Electrical and Computer Engineering Depart-ment,University of Alberta,Edmonton,AB T6G2V4,Canada(e-mail: faruque@ece.ualberta.ca;yuyan@ece.ualberta.ca;dinavahi@ece.ualberta.ca). Digital Object Identifier10.1109/TPWRD.2005.852376and its control system.However,to compare the performance of any two simulators,similar circuit topology with control is a prerequisite.To achieve that goal,a benchmark system for HVDC,known as the CIGRÉBenchmark Model,was proposed in1985[3].It provided a common reference system for HVDC system ter in1991,a comparison of four digital models has been carried out by the CIGRÉWorking Group[4],[5],and a benchmark system for HVDC control study was also proposed.A detailed comparison between ATP and NETOMAC for the simulation of HVDC system wasfirst reported in[6],where the fundamental differences between the two software and their effects on simulation results have been discussed.The study found a good agreement between the two simulation results.More recently,custom power con-trollers such as DSTATCOM and DVR have been simulated [7]using PSCAD/EMTDC and SIMULINK to compare their performance.However,for a rigorous comparison between simulation tools and to gain insight into their capabilities and limitations,the modeled system should be able to offer the highest degree of difficulty.The main motivation for using the CIGRÉBenchmark HVDC System in this paper is that not only is it a widely used test system but also it is complex enough, with deliberate difficulties introduced for a comprehensive performance evaluation of the two simulation tools.Section II of this paper gives a brief introduction about the two simulation tools highlighting their solution techniques, and Section III introduces the CIGRÉHVDC benchmark system.Sections IV–VI present the detailed model of HVDC system and its controller in three simulation environments: PSCAD/EMTDC,PSB/SIMULINK,and PSCAD-SIMULINK interface.Results are presented in Section VII,followed by conclusions in Section VIII.II.PSCAD/EMTDC AND PSB/SIMULINK PSCAD/EMTDC is a powerful time-domain transient sim-ulator for simulating power systems and its controls.It uses graphical user interface to sketch virtually any electrical equip-ment and provide a fast andflexible solution.PSCAD/EMTDC represents and solves the differential equations of the entire power system and its control in the time domain(both elec-tromagnetic and electromechanical systems)[8].It employs the well-known nodal analysis technique together with trapezoidal integration rule withfixed integration time-step.It also uses in-terpolation technique with instantaneous switching to represent the structural changes of the system[9],[10].MATLAB/SIMULINK is a high-performance multifunc-tional software that uses functions for numerical computation,0885-8977/$20.00©2006IEEEFig.1.Single-line diagram of the CIGRÉbenchmark HVDC system.system simulation,and application development.Power System Blockset(PSB)is one of its design tools for modeling and simulating electric power systems within the SIMULINK environment[2],[11].It contains a block library with common components and devices found in electrical power networks that are based on electromagnetic and electromechanical equations.PSB/SIMULINK can be used for modeling and simulation of both power and control systems.PSB solves the system equations through state-variable analysis using either fixed or variable integration time-step.The linear dynamics of the system are expressed through continuous or discrete time-domain state-space equations.It also offers theflexibility of choosing from a variety of integration algorithms.III.F IRST CIGRÉHVDC B ENCHMARK S YSTEMThefirst CIGRÉHVDC benchmark system shown in Fig.1 was proposed in[3].The system is a mono-polar500-kV, 1000-MW HVDC link with12-pulse converters on both rec-tifier and inverter sides,connected to weak ac systems(short circuit ratio of2.5at a rated frequency of50Hz)that provide a considerable degree of difficulty for dc controls.Damped filters and capacitive reactive compensation are also provided on both sides.The power circuit of the converter consists of the following subcircuits.A.AC SideThe ac sides of the HVDC system consist of supply net-work,filters,and transformers on both sides of the converter. The ac supply network is represented by a Thévénin equivalent voltage source with an equivalent source impedance.ACfilters are added to absorb the harmonics generated by the converter as well as to supply reactive power to the converter.B.DC SideThe dc side of the converter consists of smoothing reactors for both rectifier and the inverter side.The dc transmission line is represented by an equivalent T network,which can be tuned to fundamental frequency to provide a difficult resonant condition for the modeled system.C.ConverterThe converter stations are represented by12-pulse configura-tion with two six-pulse valves in series.In the actual converter, each valve is constructed with many thyristors in series.Each valve hasa limiting inductor,and each thyristor has par-allel RC snubbers.IV.CIGRÉHVDC S YSTEM M ODEL IN PSCADThe full three-phase model of the CIGRÉHVDC benchmark system is available as an examplefile in PSCAD/EMTDC Ver-sion4.0.1.Data for the CIGRÉHVDC benchmark system[4],[5]is given in Table V.A.Power Circuit Modeling1)Converter Model:The converters(rectifier and inverter) are modeled using six-pulse Graetz bridge block,which in-cludes an internal Phase Locked Oscillator(PLO),firing and valve blocking controls,andfiringangle/extinctionangle measurements.It also includes built-in RC snubber circuits for each thyristor.Thyristor valves are modeled as ideal devices, and therefore,negative turn-off andfiring due tolargeor are not considered.2)Converter Transformer Model:Two transformers on the rectifier side are modeled by three-phase two winding trans-former,one with grounded Wye–Wye connection and the other with grounded Wye–Delta connection.The model uses satura-tion characteristic and tap setting arrangement.The inverter side transformers use a similar model.3)DC Line Model:The dc line is modeled using an equiva-lent-T network with smoothing reactors inserted on both sides.4)Supply Voltage Source:The supply voltages on both rec-tifier and inverter sides have been represented through three-phase ac voltage sources.5)Filters and Reactive Support:Tunedfilters and reactive support are provided at both the rectifier and the inverter ac sides,as shown in Fig.1.B.Control System ModelThe control model mainly consistsof measurements and generation offiring signals for both the rectifier and inverter. The PLO is used to build thefiring signals.The output signal of the PLO is a ramp,synchronized to the phase-A commutatingbus line-to-ground voltage,which is used to generate the firing signal for Valve 1.The ramps for other valves are generated by adding 60to the Valve 1ramp.As a result,an equidistant pulse is realized.The actual firing time is calculated by comparingthe order to the value of the ramp and using interpolation [10]technique.At the same time,if the valve is pulsed but its voltage is still less than the forward voltage drop,this model has a logic to delay firing until the voltage is exactly equal to the forward voltage drop.The firing pulse is maintained across each valve for 120.Theand measurement circuits use zero-crossing informa-tion from commutating bus voltages and valve switching times and then convert this time difference to an angle (using mea-sured PLO frequency).Firingangle (in seconds)is the time when valve turns on minus the zero crossing time for valve .Extinctionangle (in seconds)for valve is the time at which the commutation bus voltage for valve crosses zero (negative to positive)minus the time valve turns off.The control schemes for both recti fier and inverter of the CIGR ÉHVDC system are available in the example file in PSCAD/EMTDC Version 4.0.1.Following are the controllers used in the control schemes:•ExtinctionAngleController;•dc Current Controller;•V oltage Dependent Current Limiter (VDCOL).1)Rectifier Control:The recti fier control system uses Con-stant Current Control (CCC)technique.The reference for cur-rent limit is obtained from the inverter side.This is done to en-sure the protection of the converter in situations when inverter side does not have suf ficient dc voltage support (due to a fault)or does not have suf ficient load requirement (load rejection).The reference current used in recti fier control depends on the dc voltage available at the inverter side.Dc current on the recti fier side is measured using proper transducers and passed through necessary filters before they are compared to produce the error signal.The error signal is then passed through a PI controller,which produces the necessary firing angleorder .The firing circuit uses this information to generate the equidistant pulses for the valves using the technique described earlier.2)Inverter Control:The Extinction Angle Controlor con-trol and current control have been implemented on the inverter side.The CCC with V oltage Dependent Current Order Limiter (VDCOL)have been used here through PI controllers.The ref-erence limit for the current control is obtained through a com-parison of the external reference (selected by the operator or load requirement)and VDCOL (implemented through lookup table)output.The measured current is then subtracted from the reference limit to produce an error signal that is sent to the PI controller to produce the required angle order.The control uses another PI controller to produce gamma angle order for the inverter.The two angle orders are compared,and the minimum of the two is used to calculate the firing instant.V .CIGR ÉHVDC S YSTEM M ODEL IN PSBThe CIGR ÉHVDC system model developed using PSB/SIMULINK Version 6.5release 13is shown in Fig.9.To implement this model,a total of 106states,37inputs,112outputs,and 31switches were used.A.Power Circuit ModelingThe recti fier and the inverter are 12-pulse converters con-structed by two universal bridge blocks connected in series.The converter transformers are modeled by one three-phase two winding transformer with grounded Wye –Wye connection,the other by three-phase two winding transformer with grounded Wye –Delta connection.The converters are interconnected through a T-network.1)Universal Bridge Block:The universal bridge block im-plements a universal three-phase power converter that consists of six power switches connected as a bridge.The type of power switch and converter con figuration can be selected from the di-alog box.Series RC snubber circuits are connected in parallel with each switch device.The vector gating signals are six-pulse trains corresponding to the natural order of commutation.Theand measurements are not realized in this model.2)Three Phase Source:A three-phase ac voltage source in series with a R-L combination is used to model the source,and its parameters are set as in Table V .3)Converter Transformer Model:The three-phase two winding transformers models have been used where winding connection and winding parameters can be set through mask parameters.The tap position is at a fixed position determined by a multiplication factor applied on the primary nominal voltage of the converter transformers (1.01on recti fier side;0.989on inverter side).The saturation has been simulated.The saturation characteristic has been speci fied by a series of current/flux pairs (in p.u.)starting with the pair (0,0).The dc line,ac filters,and reactive support are similar to the ones used in the PSCAD/EMTDC model.B.Control System ModelThe control blocks available in SIMULINK have been used to emulate the control algorithm described in Section IV-B,and enough care has been taken to ensure that exact parameters as in PSCAD/EMTDC simulation are used.Some control param-eters required conversion to their proper values due to differ-ences in units.The recti fier side uses current control with a ref-erence obtained from the inverter VDCOL output (implemented through a lookup table),and the inverter control has both cur-rent controland control operating in parallel,and the lower output of the two is used to generate the firing pulses.Unlike PSCAD/EMTDC,theangle is not provided directly from the converter valve data.It needed to be implemented through mea-surements taken from valve data.The control block diagrams are shown in Fig.2.VI.PSCAD-SIMULINK I NTERFACEPSCAD Version 4.0.1has the capability of interfacing with MATLAB/SIMULINK commands and toolboxes through a special interface.MATLAB programs or block-sets that would be interfaced,with PSCAD needing to be designed and saved as a MATLAB program file or as a SIMULINK block file.Then,a user-de fined block must be provided in PSCAD,with the neces-sary inputs and outputs,to interface the MATLAB/SIMULINK file.In this paper,the power circuit of CIGR ÉHVDC system has been modeled in the PSCAD/EMTDC environment whileFig.2.CIGR ÉHVDC control system in SIMULINK.(a)Recti fier control.(b)Gamma measurement.(c)Inverter control.the control system has been modeled using block-sets from PSB and the SIMULINK Control Library.An interfacing block has been created in PSCAD/EMTDC that linked the SIMULINK files through FORTRAN scripts de fined within the block.A reverse scenario,where the power circuit is mod-eled in PSB/SIMULINK and the control system is modeled in PSCAD/EMTDC,is also feasible.Fig.3shows the block diagram of PSCAD-SIMULINK interface used to simulate CIGR ÉHVDC system.The time step used for the simulation is50;the two programs exchange information between them continuously at every time step.VII.S IMULATION R ESULTSWith the goal of performing a rigorous comparative study among the simulation tools,the CIGR ÉHVDC system has been simulated in three environments:1)using PSCAD/EMTDC only;2)using PSB/SIMULINK only;3)using the PSCAD-SIMULINK Interface.Steady-state and transient results (created through various faults)were recorded and then compared.The comparison reveals a high degree of similarity among the results obtained through the three simulation environments with minor discrep-ancies.A.Steady StateFor steady-state analysis,the system has been simulated for a duration of 2s in all three simulation environments.There were some initial transients that subsided within about 0.5s and the system reached steady state.1)DC Voltages and Currents:Fig.4shows the results where the first column is produced by PSCAD/EMTDC,the second from PSB/SIMULINK,and the third by the PSCAD-SIMULINK interface.Row-wise,the first row shows the recti fier dc voltage produced by the three simulation tools;the second and fourth row are the magni fied view of dc voltages on both the recti fier and inverter side;the third and fifthrowFig.3.PSCAD-SIMULINK interface.show the harmonic spectrum.The following observations can be made from Fig.4.•For both inverter and recti fier,the dc voltages show small oscillations around the reference value (1p.u);however,they are almost identical,except for minor differences at start-up.During initialization,PSCAD/EMTDC and PSCAD-SIMULINK interface do not show any negative dc voltage,whereas PSB/SIMULINK shows a negativetransientp.u .However,all three simulation tools have produced identical waveforms in terms of phase and magnitude in steady state,and there is hardly any discrep-ancy among them.The zoomed view of their steady-state waveforms re flects that fact.•The mean of output dc voltage produced by PSCAD/EMTDC and PSCAD-SIMULINK interface falls short of reference outputs by (1–2)%(0.99p.u.for PSCAD/EMTDC and 0.98p.u.for PSCAD-SIMULINK interface),whereas it is 1.0p.u.for PSB/SIMULINK.•The Fourier spectrum of the corresponding waveforms have very few differences.The dc component is close to 1.0for all environments,but in the plot,it is not shown in full magnitude,for the sake of highlighting the harmonicsFig.4.Steady-state results for dc voltage.TABLE IC OMPARISON OF THD(%),M EAN,AND MAD FOR DC V OLTAGES AND C URRENTS P RODUCED BY D IFFERENT S IMULATION TOOLSpresent in the signal.The THDs found in the three cases for different waveforms are also very close.•Table I shows further information about thefine differ-ences in terms of mean value,THD,and Mean Absolute Deviation(MAD).Close results have also been observed for dc currents on both the rectifier and inverter sides. 2)AC Voltages and Currents:All ac side waveforms on both the rectifier and inverter sides have also been compared. The results were found to be similar,and only the ac current waveforms and their harmonic spectrum are shown in Fig.5. Both rectifier and inverter ac currents are similar in terms of phase angle and their magnitude;their spectrum is identical, which reaffirms the accuracy of all three simulation techniques. The11th and13th harmonics are the dominant harmonics on both rectifier and inverter sides;their THDs have been found to be close for all simulation environments.Table II compares other control outputs(,inverter,and rectifier). PSCAD-SIMULINK shows the maximum rectifier(17.28), while PSB/SIMULINK shows the minimum(14.44).This result agrees with the mean value of the dc voltage produced on the rectifier side by the three simulation environments.Sim-ilarly,for,PSCAD-SIMULINK shows the highest(15.72), and PSB/SIMULINK shows the lowest(14.95).However, these differences are small,and the produced results are con-sistent.B.TransientsDc and ac faults have been simulated in the three simula-tion environments.The instant and duration of faults have beenFig.5.Steady-state results for ac currents on the recti fier and inverter side.TABLE IIC OMPARISON OF O UTPUTS P RODUCED BY C ONTROL SYSTEMSmaintained the same for all types of faults,i.e.,the fault is ap-plied at 1.0s and cleared after 0.15s.The fault resistance hasbeen chosen as0.1and1for fault-on and fault-off situa-tions,respectively.Clearing of the fault has been allowed,even when there is a fault current flowing.1)Dc Fault:This fault has been located at the midpoint of the dc line.Fig.6illustrates different output parameters of the system.The transient response in all three simulation environ-ments has been found almost similar.During the fault,the dc voltage has gone down to zero (the small oscillation is due to the energy stored in the capacitor),and a momentary transient dc current has been observed.However,control response forces recti fier andinverter to reach maximum andinverter to reach minimum,thereby reducing the current flow.VDCOL forces the current to stay minimum until the dc voltage situation is improved.Once the fault is cleared,the dc voltage is recov-ered,and the control system brings the system back to normal operation.The transient response for all three simulation envi-ronments has been compared in terms of Rise Time (RT),Set-tling Time (ST),and Overshoot (OS)in Table III.2)AC Faults:Two cases of ac faults have been simulated:One is a single line-to-ground fault (see Fig.7),and the other is a three-phase to ground fault (see Fig.8).Both are applied on the inverter side of the system.During fault,the dc voltage has gone down to zero (neglecting oscillation due to capacitor energy storage),the dc current faces a momentary overshoot,and then goes to minimum limit with some oscillation present (due to the oscillation of dc voltage).Recti fier ,inverter ,andinverter reach to maximum value,thereby blocking the system for the fault duration.Once the fault is cleared,the system comes back to its normal operation.During an ac fault,commutation failures happen,resulting in a momentary drop-down of dc voltage.This causes the VDCOL to limit the dc current to a minimum,and ac voltages also get disturbed (not shown in the figure).The voltage and current waveforms for the three environments are similar;however,the following minor discrepancies were observed.•In all three cases,the rise-time for inverter dc voltage for ac faults is shorter than that for the dc fault.Even though PSCAD/EMTDC and PSCAD-SIMULINK showFig.6.V oltages and currents under a short duration dc fault.a faster rise than PSB/SIMULINK case,PSB/SIMULINKreaches steady state before the other two cases.•The highest transient values of inverter dc currents during the phase-to-ground fault are 2.58p.u.for PSCAD/EMTDC, 2.4p.u.for PSB/SIMULINK,and2.39p.u.for PSCAD-SIMULINK.•The peak value of inverter ac current for the single phase-to-ground fault is13.2kA for PSCAD/EMTDC, 14kA for PSB/SIMULINK,and13kA for PSCAD-SIMULINK.•In case of the three-phase to ground fault,when the faultis clearedat,in all the three environments,thesystem is brought back to normal operation within0.05 s;however,PSCAD/EMTDC and PSCAD-SIMULINK could not stabilize the system,i.e.,after a small over-shoot,the system collapses again,though it regains the control very fast,and the system stability is restored.PSB/SIMULINK,however,does not show this behavior.C.Execution Time and MemoryAll three environments were run on a Pentium IV1.5–GHz processor running Windows2000operating system.The ex-ecution time was recorded from the Time Summary shown on the output window in PSCAD/EMTDC and by using the cputime function at the start and end of the simulation in PSB/rmation on memory usage was collected from the System Performance Monitor on Windows2000. Although resource requirements for both programs may not be the same,attempts have been made to allow no other programs except the systemfiles to run while the simulations were performed.Table IV shows the execution time and memory usage for the three environments.PSCAD/EMTDC was found to be the fastest environment,while PSCAD-SIMULINK Interface was the slowest.For a simulation duration of2s, PSCAD/EMTDC took30.5s with a memory usage of42 MB,whereas PSB/SIMULINK took72.2s with a memory usage of107MB.In comparison to these two environments, PSCAD-SIMULINK took a much longer execution time;using a50sampling period in both PSCAD and SIMULINK environments,the execution time was found to be12503.58s with a memory usage of63MB(24MB for PSCAD and39MB for SIMULINK).The reason for such a long simulation time is the necessity of data exchange between the two programs at every50;the memory usage for the interface was less than the other two environments due to the task partition(electrical system in PSCAD and control system in SIMULINK).A higher control sampling period reduced the execution time by a very small margin(2.5%for100);however,it also reduced the accuracy of the simulation.TABLE IIIC OMPARISON OF R ISE T IME (RT)(S ),S ETTLING T IME (ST)(S ),AND O VERSHOOT (OS)(P .U .)D URING R ECOVERY F ROM A DC FAULTFig.7.V oltages and currents under a short duration ac fault (phase-A to ground)on the inverter side.D.General RemarksBoth PSCAD/EMTDC and PSB/SIMULINK provides user-friendly graphics for modeling power and control systems through simple functional blocks.However,the following minor differences particular to this case study,are worth men-tioning.•PSCAD/EMTDC is a specialized software designed mainly for the analysis of ac/dc systems.Therefore,it has added advantages such as built-in PLO-based firingcontrol and the measurementofangles embedded inside the six-pulse Garetz bridge .On the other hand,PSB/SIMULINK requires these blocks and the measure-ment system to be developed by the user.•PSB/SIMULINK offers more flexibility in terms of choice of the solution techniques:fixed or variable time-step-based solutions.However,for this study,a fixed step-size trapezoidal rule has been used to be consistent with PSCAD/EMTDC.•The error debugging system in PSCAD/EMTDC is quite complex.In some cases,instead of locating the exact source of error,it returned general error messages.VIII.C ONCLUSIONSA detailed comparison of the performance of three simu-lation environments (PSCAD/EMTDC,PSB/SIMULINK,and PSCAD-SIMULINK Interface)has been demonstrated by mod-eling the CIGR ÉHVDC Benchmark System.All three environ-ments produced almost identical and consistent results during steady-state and transients situations,validating the accuracy of the modeling and solution algorithms.In terms of computationalFig.8.V oltages and currents under a short duration ac fault(three-phase to ground)on the inverter side.Fig.9.CIGRÉHVDC benchmark system model in PSB/SIMULINK.TABLE IVE XECUTION T IME (ET)(S )AND M EMORY U SAGE (MU)(MB)FORAS IMULATION D URATION OF 2S W ITH A T IME S TEP OF 50sTABLE VCIGR ÉHVDC B ENCHMARK S YSTEM DATAspeed and memory usage,PSCAD/EMTDC was found to be the most ef ficient environment.A PPENDIXTable V shows the CIGR ÉHVDC benchmark system data.R EFERENCES[1] D.A.Woodford,A.M.Gole,and R.W.Menzies,“Digital simulationof DC links and AC machines,”IEEE Trans.Power App.Syst.,vol.102,no.6,pp.1616–1623,Jun.1983.[2]L.-A.Dessaint,H.Le-Huy,G.Sybille,and P.Brunelle,“A power systemsimulation tool based on SIMULINK,”IEEE Trans.Ind.Electron.,vol.46,no.6,pp.1252–1254,Dec.1999.[3]J.D.Ainsworth,“Proposed benchmark model for study of HVDC con-trols by simulator or digital computer,”in Proc.CIGRE SC-14Colloq.HVDC With Weak AC Systems ,Maidstone,U.K.,Sep.1985.[4]M.Szechtman,T.Wess,and C.V .Thio,“First benchmark model forHVDC control studies,”Electra ,no.135,pp.54–67,Apr.1991.[5],“A benchmark model for HVDC system studies,”in Proc.Int.Conf.AC/DC Power Transmission ,Sep.17–20,1991,pp.374–378.[6]P.Lehn,J.Rittiger,and B.Kulicke,“Comparison of the ATP versionof the EMTP and the NETOMAC program for simulation of HVDC systems,”IEEE Trans.Power Del.,vol.10,no.4,pp.2048–2053,Oct.1995.[7]W.Freitas and A.Morelato,“Comparative study between power systemblockset and PSCAD/EMTDC for transient analysis of custom power devices based on voltage source converter,”in Proc.Int.Conf.Power Systems Transients ,New Orleans,LA,2003,pp.91–96.[8] F.Jurado,N.Acero,J.Carpio,and M.Castro,“Using various computertools in electrical transient studies,”in Proc.Int.Conf.30th ASEE/IEEE Frontiers in Education ,Kansas City,MO,Oct.18–21,2000,pp.F4E17–F4E22.[9]PSCAD/EMTDC User ’s Manual ,Manitoba HVDC Research Centre,2001.[10]G.D.Irwin,D.A.Woodford,and A.Gole,“Precision simulation ofPWM controller,”in Proc.Int.Conf.Power System Transients ,Rio de Janeiro,Brazil,2001,pp.301–306.[11]Power System Blockset User ’s Guide ,TEQSIM International,Inc.,2001.M.O.Faruque (S ’03)received the B.Sc.Engg degree in 1992from Chittagong University of Engineering and Technology (CUET),Chittagong,Bangladesh,and M.Eng.Sc.degree in 1999from the University of Malaya,Kuala Lumpur,Malaysia,in the area of power engineering.He is working toward the Ph.D.degree from the Department of Electrical and Computer Engineering,University of Alberta,Edmonton,AB,Canada.For the last ten years,he has been working in both academia and industry,and his research interests include FACTS,HVDC,and real-time digital simulation of power electronics and power systems.Yuyan Zhang received the Bachelor ’s degree in electrical engineering from South-East University,Nanjing,China,in 1990and the M.E.degree from the University of Alberta,Edmonton,AB,Canada in 2003.Since then,she has worked as an Electrical Engineer for a variety of indus-tries,including the Beijing Chemical Plant,Hongkong Well fine Ltd.,Motorola China Ltd.,and Siemens Ltd.,China,in the power generation group.Her re-search interests are in the area of power distribution systems,HVDC,and digital simulation.Venkata Dinavahi (M ’00)received the B.Eng.degree in electrical engineering from Nagpur University,Nagpur,India,in 1993,the M.Tech.degree from the Indian Institute of Technology,Kanpur,India,in 1996,and the Ph.D.degree in electrical and computer engineering from the University of Toronto,Toronto,ON,Canada,in 2000.Presently,he is an Assistant Professor at the University of Alberta,Edmonton,AB,Canada.His research interests include electromagnetic transient analysis,power electronics,and real-time simulation and control.Dr.Dinavah is a member of CIGR Éand a Professional Engineer in the Province of Alberta,Canada.。

pd控制舵机Chapter 1: IntroductionIn this chapter, the background and significance of PD control for servo motors will be discussed. Additionally, the objectives and structure of this paper will be outlined.1.1 BackgroundServo motors are widely used in various fields such as robotics, industrial automation, and aerospace applications. They are responsible for converting electrical energy into precise mechanical movements. However, ensuring accurate control of servo motors can be challenging due to factors such as external disturbances, non-linearities, and system uncertainties.1.2 SignificancePD control, which stands for Proportional-Derivative control, is a commonly used method to improve the performance of servo motors. By combining proportional and derivative terms, PD control can provide better stability, reduce steady-state errors, and enhance the response speed of the servo motor. Therefore, understanding and implementing PD control for servo motors is of great significance in improving the overall system performance.1.3 ObjectivesThe main objective of this paper is to investigate and implement PD control for servo motors. Specifically, the following aspects will be covered:- Theoretical background of PD control- System modeling of servo motors- Design and tuning of PD controller parameters- Experimental verification and performance evaluation1.4 StructureThe remaining chapters of this paper are organized as follows:- Chapter 2: Theoretical Background- Chapter 3: System Modeling and Controller Design- Chapter 4: Experimental Verification and Performance Evaluation- ConclusionChapter 2: Theoretical BackgroundThis chapter will provide a comprehensive background on the principles of PD control for servo motors. The fundamental concepts, including proportional and derivative actions, will be explained in detail. Additionally, the advantages and limitations of PD control will be discussed.2.1 Proportional ControlProportional control is the simplest form of control that relies on the relationship between the error and the control input. The control action is proportional to the error, resulting in a linear relationship. The proportional gain parameter determines the response speed and robustness of the control system.2.2 Derivative ControlDerivative control is based on the rate of change of the error. By considering the instantaneous rate of change, derivative control can provide anticipatory control action. The derivative gain parameter affects the stability and damping of the system.2.3 PD ControlPD control combines the proportional and derivative actions to improve system performance. It offers both quick response time and accurate steady-state control. The tuning of the two gains is crucial for obtaining optimal performance.Chapter 3: System Modeling and Controller DesignThis chapter focuses on the mathematical modeling of servo motors and the design of a PD controller. The steps involved in designing the PD controller, including parameter tuning, will be presented. Additionally, simulation techniques may be utilized to evaluate the performance of the designed controller.3.1 System ModelingThe mathematical representation of servo motors, such as transfer function models or state-space models, will be established. The physical parameters and characteristics of the motor will be considered in the modeling process.3.2 Controller DesignThe design of the PD controller will involve determining appropriate values for the proportional and derivative gains. Various methods, such as trial and error, Ziegler-Nichols method, or optimization algorithms, can be applied to achieve optimal controller performance.Chapter 4: Experimental Verification and Performance Evaluation In this chapter, the designed PD controller will be implemented on a physical servo motor system. The experimental setup and measurement instruments will be described. The performance ofthe controller will be evaluated in terms of tracking accuracy, stability, and disturbance rejection.4.1 Experimental SetupThe hardware and software components required for the experimental setup will be outlined. This may include the servo motor, motor driver, position encoder, microcontroller or PLC, and the control algorithm implementation.4.2 Performance EvaluationThe performance of the PD controller will be evaluated through a series of experiments. The tracking accuracy, response time, and disturbance rejection capability will be quantitatively and qualitatively analyzed. Comparisons with other control methods may also be included.ConclusionIn this final chapter, the key findings and contributions of this paper will be summarized. The effectiveness and limitations of PD control for servo motors will be discussed. Suggestions for further research and improvements will also be provided.秋天是一年中最美的季节之一。

构建模型的英语Building a Model: Key Steps and ConsiderationsIntroduction:Building a model is a crucial process in various fields such as machine learning, data analysis, and statistical modeling. It involves constructing a representation of a system or phenomenon to understand, predict, or analyze its behavior. In this article, we will discuss the key steps and considerations involved in building a model.1. Defining the Problem:The first step in model building is to clearly define the problem or objective. This involves understanding what needs to be achieved, identifying the available data, and setting realistic expectations. A well-defined problem statement helps guide the entire modeling process.2. Data Collection and Preparation:Once the problem is defined, the next step is to gatherthe relevant data. This may involve sourcing it from various databases, utilizing existing datasets, or conducting experiments to generate new data. Data preparation is equally crucial, which includes cleaning, transforming, andformatting the data in a suitable manner for modeling.3. Exploratory Data Analysis (EDA):EDA involves analyzing the collected data to understandits characteristics, identify patterns, and detect outliers.It helps to gain insights into the data, validate assumptions, and select appropriate modeling techniques.4. Feature Selection and Engineering:Feature selection refers to the process of identifyingthe most relevant variables or features from the data that contribute significantly to the outcome. Feature engineeringinvolves creating new features or transforming existing ones to improve the model's performance. This step requires domain knowledge and creativity.5. Selecting a Modeling Technique:Choosing an appropriate modeling technique depends on the nature of the problem, available data, and desired outcome.It could range from traditional statistical methods such as linear regression and decision trees to more advanced techniques like neural networks and deep learning.6. Model Training and Evaluation:Once the modeling technique is selected, the model needs to be trained using the prepared data. This involvessplitting the data into training and validation sets,defining performance metrics, and fine-tuning the model parameters. After training, the model's performance is evaluated using various metrics such as accuracy, precision, recall, or mean squared error.7. Model Optimization:Model optimization aims to improve the model's performance by fine-tuning its parameters or exploring different algorithms. Techniques like cross-validation, hyperparameter tuning, and regularization can be employed to prevent overfitting and achieve better generalization.8. Model Deployment and Monitoring:After optimizing the model, it is deployed for real-world usage, integrating it into the existing architecture or systems. It is important to continuously monitor the model's performance, assess its accuracy, and retrain or update it periodically to adapt to changing data patterns.9. Model Interpretation and Communication:Model interpretation helps in understanding the factors influencing the model's predictions and gaining insights. It is crucial for decision-making and explaining the model'soutcomes to stakeholders. Effectively communicating the model's results and limitations is essential for gainingtrust and facilitating its practical implementation.Conclusion:Building a model involves a series of well-defined steps, from problem definition to model interpretation. Each step requires careful consideration and expertise in order to develop an accurate and reliable model. By following these steps and considering the specific requirements of the problem at hand, one can build effective models to gain insights, make predictions, or solve complex problems in various industries.。

heterogeneous-agent reinforcement learning-回复Heterogeneous Agent Reinforcement Learning: A Step-by-Step IntroductionIntroduction:Reinforcement learning (RL) is a subfield of machine learning that focuses on developing algorithms capable of making decisions and learning from experience. Traditionally, RL assumes a homogeneous environment where a single agent interacts with its surroundings. However, many real-world scenarios involve multiple agents with different capabilities and objectives. Heterogeneous agent reinforcement learning (HARL) addresses this challenge by considering diverse agents in an environment. In this article, we will explore the steps involved in building and applying HARL algorithms.1. Understanding Heterogeneous Agents:In HARL, each agent has its own unique observations, actions, and rewards. Agents can have varying learning capabilities, goals, and levels of expertise. For example, in a collaborative robotic setting, one robot may be focused on manipulation tasks while anotherspecializes in navigation. Understanding the differences between agents is crucial in designing effective HARL approaches.2. Multi-Agent System Modeling:To begin with HARL, we need to model the multi-agent system (MAS). MAS involves defining the agents, their interactions, and their influence on the environment. One common approach is to make the assumptions of either cooperative or competitive behavior between the agents. Cooperative behavior aims to maximize a collective reward, while competitive behavior focuses on individual rewards. The modeling process sets the stage for training and evaluating the HARL algorithms.3. MARL Algorithms:HARL leverages existing multi-agent reinforcement learning (MARL) algorithms, making adaptations to handle heterogeneity. Some popular MARL algorithms include Q-learning, Policy Gradient, and Deep Deterministic Policy Gradient. These algorithms can be modified to account for variations in agent capabilities and goals. One approach is to incorporate multiple value functions or policies, each tailored for a specific agent type. Another approach involves designing reward shaping mechanisms that consider individualagent contributions. The choice of algorithm depends on the specific requirements of the environment and the agents involved.4. Reward Design:Designing suitable reward functions is critical for effective HARL. Agents with different capabilities and goals may require distinct reward structures. For example, an agent focused on exploration may receive higher rewards for visiting uncharted areas, while a more experienced agent might prioritize exploiting known territories. Balancing the rewards to encourage cooperation among heterogeneous agents is a challenge. Iterative refinement of reward structures is often required to incentivize desired behaviors.5. Training:HARL training involves iteratively updating the agents' policies or value functions to maximize their objectives. The training process typically utilizes experiences gained during interactions with the environment. Each agent's learning rate, exploration-exploitation trade-off, and update mechanism can be customized to align with their capabilities and objectives. In complex environments, training may take an extended period to converge.6. Evaluation and Performance Metrics:Assessing the performance of HARL algorithms requires suitable evaluation and performance metrics. These metrics can vary depending on the specific MAS setting. Examples include the average cumulative reward, coordination level among agents, and fairness of outcomes. Effective evaluation helps identify areas for improvement and comparison among different algorithms.7. Real-World Applications:HARL finds applications in various domains such as autonomous driving, swarm robotics, and smart grids. By considering heterogeneity among agents, HARL algorithms can facilitate efficient coordination, resource allocation, and decision-making in complex environments. Real-world applications often require adapting HARL techniques to specific use cases, considering factors like scalability, real-time decision-making, and safety.Conclusion:Heterogeneous agent reinforcement learning (HARL) extends traditional reinforcement learning to tackle scenarios involving multiple agents with different capacities and objectives. Bymodeling the multi-agent system, adapting existing MARL algorithms, designing appropriate rewards, and tailoring training and evaluation strategies, HARL enables effective coordination and decision-making among diverse agents in complex environments. The field of HARL holds significant potential for solving real-world problems requiring collaboration among heterogeneous agents.。

the metamorphoses of performancebudgetingthe metamorphoses of performance budgeting翻译为：绩效预算的变化and physical control in the public sector, of which empirical assessment becomes increasingly important.Performance budgeting is a form of budgeting that evaluates the performance of government programs in order to determine how the budget should be allocated. It is a tool used to assess and manage the efficiency of government programs and operations. Performance budgeting looks at the use of funds and how they are used to achieve goals and objectives. Performance budgeting measures the actual activities of a program or organization, rather than just its outputs or outcomes. Performance budgeting is a process designed to ensure that the spending of public funds is properly monitored and managed in order to ensure that it is achieving its intended purpose.Physical control in the public sector often involves the control of physical assets such as buildings and property. It is the process of monitoring and controlling the physical assets of a government organization or public service in order to ensure they are properly maintained and used in acost-effective manner. Physical control includes activities such as ensuring the maintenance of facilities, monitoring safety and security, and ensuring that assets remain up to date and in good condition.The metamorphosis of performance budgeting and physical control in the public sector has seen significant changes over the years, the most notable being the shift from the traditional focus on outputs to a more comprehensive evaluation of outcomes. This shift was brought about by an increased focus on results-based management, which emphasizes the importance of assessing the effectiveness of public programs and initiatives in delivering services. With an increasing emphasis on the need for accountability in the public sector, performance budgeting and physical control now also need to include elements of empirical assessment. By collecting data and conducting analysis, governments can track the performance of their programs and make necessary adjustments as needed in order to ensure that they are being used as efficiently as possible. Furthermore, empirical assessment allows governments to gain insight into how funds are being used, what the outcomes of their programs are, and whether or not they are achieving their set goals. Through the implementation of empirical assessment, performance budgeting and physical control in the public sector can be improved, and help ensure that public resources are being used in a waythat is beneficial to the general public.。