On-Line Optimization of Microgrid Operating Cost Based on Deep Reinforcement Learning. Issue 1 (March 2021)

• Record Type: Journal Article
• Authors: Lin, S H; Yu, H H; Chen, H W
• Is Part Of: IOP conference series. Volume 701:Issue 1(2021)
• Publication Date: 2021-03
• Subjects: Earth sciences -- Periodicals; Environmental sciences -- Congresses; Environmental sciences -- Periodicals; 550.5
• DOI: 10.1088/1755-1315/701/1/012084
• Languages: English
• ISSNs: 1755-1307

Abstract

Abstract: In view of the microgrid random optimization scheduling problem, this paper proposes a microgrid online optimization algorithm based on deep reinforcement learning. By using the approximate state-action value function of the deep neural network, the action of the battery is discretized into the decision variables output by the neural network, and then the remaining decision variables are solved by nonlinear programming and the immediate return is calculated. The optimal strategy is obtained by using the Q-learning algorithm. In order to make the neural network adapt to the randomness of wind, photovoltaic and load power, according to the wind, photovoltaic and load power prediction curves as well as wind, photovoltaic and load prediction errors, Monte Carlo sampling was used to generate multiple sets of training curves to train the neural network. After the training is completed, the weight is fixed and the real-time action value of the battery is output according to the real-time status of the microgrid, so as to realize the online optimal dispatching of the microgrid. Compared with day-ahead optimization results under different fluctuations of wind, photovoltaic and load power, the effectiveness and superiority of this algorithm in online optimization of microgrid are verified.