Online self-learning optimal control for residential energy system based on adaptive dynamic programming
Xin Tong, Chao Huang, Xu-guang Hu, Dazhong Ma · 2021 China Automation Congress (CAC) · 2021
The optimization and management of residential energy system is more complex because it contains time-varying load and random renewable energy. This paper designs an online self-learning strategy to control energy storage system (ESS), so as to realize the optimal economic dispatching of residential energy system. Firstly, the optimal system operation model is established and analyzed. In order to prolong the service life of energy storage, a penalty function is added to the utility function. Then, a data-based action dependent adaptive dynamic programming (ADP) is designed to reduce the operation cost of system, and the value iteration algorithm is applied to approximate the optimal control strategy and value function. The online method can flexibly deal with the fluctuation of data in the system. In addition, reasonable control of the charge and discharge behavior of residential energy storage can increase the utilization of photovoltaic (PV) and decrease the operation cost of residential energy system. Finally, the simulation results proved the effectiveness of the designed optimal control method.