ETRI-Knowledge Sharing Plaform



논문 검색
구분 SCI
연도 ~ 키워드


학술대회 A study on the reward generation method to be used in reinforcement learning to reduce the peak load
Cited 0 time in scopus Download 3 time Share share facebook twitter linkedin kakaostory
International Conference on Information and Communication Technology Convergence (ICTC) 2022, pp.2369-2372
21PR4100, 친환경에너지 공급자원 제어시스템 개발, 신철호
This paper proposes a method of generating reward according to the ESS(Energy Storage System) charge/discharge action that is most important in developing a reinforcement learning algorithm to reduce the peak load on a building. The peak load of power used by a building can occur at various times depending on the characteristics of the individual building, such as office or residential use. In order to reduce the peak load of a building through ESS optimal control, a reinforcement learning model should be trained to optimally control the ESS using the power consumption data monitored by the building. In this paper, the reinforcement learning policy was designed so that the sum of the reward values by successive ESS actions for each control time unit (at) for 24 hours in 1 day would be the maximum as the peak load was reduced, and the performance was verified by simulation. As a result of the simulation, it was confirmed that the reinforcement learning algorithm for ESS control using the reward generation method proposed in this paper can well track and reduce the peak load point that occurs at various times according to the characteristics of the power consumption pattern of individual buildings.
KSP 제안 키워드
Consumption pattern, Control time, Learning model, Peak Load, Power Consumption, Reinforcement Learning(RL), Reinforcement learning algorithm, Residential use, Well track, consumption data, energy storage system