An AI Combined with Mathematical Analysis Energy Management System for Active Distribution Network

In recent years, a notable breakthrough has been made in artificial intelligence (AI) as well as smart grid. Even though it is exciting to see more and more AI-based industrial applications, the big concern of its reliability and interoperability has not been well addressed, especially for critical systems such as the power grid. Therefore, an AI combined with a mathematical analysis method is proposed for active distribution system (ADS) management. First, the classic deep reinforcement learning (DRL) method is applied to realize the data-driven optimal control. Due to the practice limitation, the purely data-driven solution is easily dropped into local optimal points. Thus, the auxiliary mathematical analysis is added to solve this problem. The proposed control method is tested on a typical ADS. The methodology of this work can be generalized into other application systems to overcome the black-box characteristics of deep learning methods.