Smart Scheduling Strategy for Islanded Microgrid Based on Reinforcement Learning Algorithm

• 作 者: Lingxiao Gan Tao Yu Jing Li
• 机 构: Electric Power College,South China University of Technology,Guangzhou 510640,China
• 出 处: 上海市研究生“新能源与智能电网”学术论坛
• 发表日期: 2012年6期
• 关键词: Distributed generation islanded microgrid hierarchical AGC reinforcement learnin

　　This paper investigates a hierarchical Automatic Generation Control (AGC) strategy for an islanded microgrid, including wind power, solar photovoltaic, micro turbines, small hydropower and energy storage devices.The upper AGC is for central scheduling.The bottom AGC is to optimize the allocation factors, expecting to meet the requirement of energy-saving generation dispatching (ESGD).Three different bottom controllers are presented.Two of them are designed based on reinforcement learning (RL) algorithm.In order to evaluate their control performance, another proportion-based (PROP) controller which has been put into practical application is also presented.Detailed dynamic models of distributed generations and loads are built to simulate the microgrid.System responses to wind turbine hipping and to large load disturbances are tested.The results indicate that the proposed strategy based on RL algorithm can not only achieve reliability and stability of microgrid in islanded mode, but also reduce fossil energy consumption.This approach is a possible candidate for future microgrid control approaches.