16388

1709 دكتري

دكتري

مهندسي كامپيوتر

اصفهان : دانشگاه صنعتي اصفهان

1399

چهارده، [142]ص. : مصور، جدول، نمودار

مسعودرضا هاشمي

حامد نريماني

واژه نامه

محمد رضا خيام باشي، محمد امين لطيفي

1400/01/17

كتابنامه

مهندسي كامپيوتر

مهندسي برق و كامپيوتر

1400/01/17

2681113

1 Modeling the EV Drivers Charging Behavior and Estimating its Parameters via Reinforcement Learning Approach Zohreh Fotouhi z fotouhi@ec iut ac ir Date of Submission September 21 2020 Doctor of Philosophy Thesis Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 IranSupervisor Dr Massoud Reza Hashemi hashemim@cc iut ac irAdvisor Dr Hamed Narimani h narimani@cc iut ac irDepartment Graduate Program Coordinator Dr Behzad Nazari Isfahan University of Technology Isfahan 84156 83111 IranAbstract Modeling and identifying EV drivers behavior is an essential prerequisite for charging stations CSs management which improves their ef ciency and service quality In this thesis we introduce a non homogeneousbehavioral Markov model BMM for describing a typical EV driver s charging behavior where the modelparameters depend on the behavioral characteristic of the driver the vehicle speci cations and the EV batterylevel of charge Validating the model based on real data shows its ability to describe the drivers charging patternwell The sensitivity analysis of the model indicates that the EV drivers charging behavior affects the statisticalcharacteristics of the CSs behavioral related parameters This model is then applied to simulate the congestionstatus in public CSs and predict their future capacity to guarantee an appropriate service quality level Theresults show that studying and controlling the EV drivers behavior leads to a signi cant saving in CS capacityand results in consumer satisfaction thus affecting the station owners pro tability Applying BMM to designa system to identify the EV drivers charging behavior needs an accurate and feasible parameter estimator Inthis regard we propose an RL based algorithm to estimate the behavioral parameters The evaluation resultsdemonstrate the convergence of the proposed algorithms and validate the estimated behavioral parameters Key Words Electric Vehicle EV Characteristic Modeling Electric Vehicle Charging Congestion Control Reinforcement Learning Parameter EstimationIntroductionThe uncertainties and dynamism imposed by the individual and collective behavior of EVdrivers need to be considered in predicting the charging stations CSs requirements when theEV population rises in the future Kim 2019 Speci cally the random characteristics of EVdrivers behavior greatly in uence many aspects of the design and development challengesof the CS including capacity provisioning congestion management and economic aspectsof investment planning In this regard developing an applicable system for exploring the EVdrivers charging behavior taking into account the highly dynamic CS environment wouldbe valuable Various studies has been conducted in the area of design and development of CSs attempt ing to analyze the statistical characteristics of CS behavioral related parameters including EV

مسعودرضا هاشمي

حامد نريماني

محمد رضا خيام باشي، محمد امين لطيفي