Coordinated deep policy learning for frequency-Constrained energy management for second-to-second energy balancing of microgrids

Abstract

Hybrid microgrids (HMGs) are susceptible to frequency distractions due to their low-inertia and randomness of renewable resources. The traditional energy management systems (EMSs) and linear controllers have difficulty dealing with the uncertainty prevalent within the system due to nonlinearity and time-based conditions. This necessitates considering short-term imbalances in HMGs while the intermittency of renewable resources can highly affect the second-to-second time frame of the power system. To address this issue, this work proposes two stage frameworks for an HMG with second-to-second power imbalances: i) an efficient energy management system is developed to reduce costs and to improve reliability of microgrids. The proximal policy optimization (PPO) with actor and critic neural networks is utilized to solve EMS problem, ii) a secondary controller based on the non-linear backstepping controller (NBC) is developed to mitigate the dynamic fluctuations of frequency deviation. In this application, the IEEE 39-bus is considered as the benchmark system to study second-to-second power imbalances in the HMGs. The risk of bottlenecks for the test-system with various risk indices is calculated. Transient simulations of the HMG reveal the improvement of operation of the power system from security and stability point of view. The comparison analysis with the prevalent scheme demonstrates the suggested NBC scheme can provide a higher level of stability than prevalent state-of-the-art controllers.