Economic Operation and Management of Microgrid System Using Deep Reinforcement Learning

Abstract

The paper presents an adaptive Backstepping sliding mode control (BSMC) management method based on deep reinforcement learning to manage and stabilize DC/DC boost converter with constant power loads (CPLs) in Microgrids market. To perform the BSMC, the system's zero dynamic stability with diverse output functions has been presented via applying the input/output precise feedback linearization. The suggested layout has been modeled in Brunovsky's canonical model to solve the nonlinear issue resulting from the CPLs and the non-minimum phase problem. In this offered controller, the gains of the switching have been considered to being the adjustable controller coefficients that are chosen adaptively via the DRL method by online learning. This topology makes sure the rigid stability of the power electronic system by simultaneous adaptively tuning the gains. Eventually, the results prove the suggested control method owns stronger robustness efficiency and better dynamic regulation in comparison to the nonlinear control strategies.