摘要
Integrated energy microgrids (IEMs) have developed rapidly in the past years with the advancement of renewable energy and energy storage technologies. As a result, dealing with uncertainty on the source and load sides and optimizing energy storage scheduling in IEMs have become critical research issues. However, some existing methods have limitations in terms of solution accuracy and efficiency. In order to accurately grasp the uncertainty at both source and load sides while effectively reducing the cost and improving the computational efficiency, this study proposes an adaptive and lightweight algorithm to obtain the optimal scheduling strategy for energy storage. First, a modified deep learning method is proposed to predict the PV power and load demand. Then, based on the prediction results, a reinforcement learning algorithm is used to solve the energy storage scheduling model and obtain the optimal scheduling strategy. In addition, to further investigate the effects of greedy and non-greedy actions on the agent's training, this study compares the results under different action exploration policies and different time scales. The outcomes of this study were compared with the ones obtained by mixed-integer linear programming. The results show that the reinforcement learning algorithm reduces 61.17% of the solution time though loses 3.13% of the solution accuracy. The increase in computational efficiency is essential for the real-time energy storage applications.
学科与关键词
- A Data-Driven Energy Management Strategy Based on Deep Reinforcement Learning for Microgrid SystemsBao, Gang; Xu, Rui*Cognitive Computation, 2023, 15(2): 739-750.
- Digital Twin-Based economic assessment of solar energy in smart microgrids using reinforcement learning techniqueYuan, Guanghui; Xie, Fei*Solar Energy, 2023, 250: 398-408.被引 1
- Dynamic Optimization of on-Grid Integrated Energy System Considering Peak-Shaving Demand Via Learning MethodsTang, Hao*; Li, Yijin; Li, Duanchao; Wang, HaiweiIEEJ Transactions on Electrical and Electronic Engineering, 2022, 17(10): 1409-1419.被引 1
