This paper presents intelligent access control for a charging station and a framework for dynamically and adaptively managing charging requests from randomly arriving electric vehicles(EVs),to increase the revenue of ...This paper presents intelligent access control for a charging station and a framework for dynamically and adaptively managing charging requests from randomly arriving electric vehicles(EVs),to increase the revenue of the station.First,charging service requests from random EV arrivals are described as an event-driven sequential decision process,and the decision-making relies on an eventextended state that is composed of the real-time electricity price,real-time charging station state,and EV arrival event.Second,a state aggregation method is introduced to reduce the state space by first aggregating the charging station state in the form of the remaining charging time and then further aggregating it via sort coding.Besides,mathematical calculations of the code value are provided,and their uniqueness and continuous integer characteristics are proved.Then,a corresponding Q-learning method is proposed to derive an optimal or suboptimal access control policy.The results of a case study demonstrate that the proposed learning optimisation method based on the event-extended state aggregation performs better than flat Q-learning.The space complexity and time complexity are significantly reduced,which substantially improves the learning efficiency and optimisation performance.展开更多
基金the National Natural Science Foundation of China under Grant Nos.61871412,61972439。
文摘This paper presents intelligent access control for a charging station and a framework for dynamically and adaptively managing charging requests from randomly arriving electric vehicles(EVs),to increase the revenue of the station.First,charging service requests from random EV arrivals are described as an event-driven sequential decision process,and the decision-making relies on an eventextended state that is composed of the real-time electricity price,real-time charging station state,and EV arrival event.Second,a state aggregation method is introduced to reduce the state space by first aggregating the charging station state in the form of the remaining charging time and then further aggregating it via sort coding.Besides,mathematical calculations of the code value are provided,and their uniqueness and continuous integer characteristics are proved.Then,a corresponding Q-learning method is proposed to derive an optimal or suboptimal access control policy.The results of a case study demonstrate that the proposed learning optimisation method based on the event-extended state aggregation performs better than flat Q-learning.The space complexity and time complexity are significantly reduced,which substantially improves the learning efficiency and optimisation performance.