Physics-informed reward framework for virtual inertia control in VSG

The high penetration of renewable energy reduces system inertia, leading to instability and decreased power system reliability. In this context Virtual Synchronous Generator emerges as a method that can provide virtual inertia to improve system stability. The trade-off between frequency response and power output of VSG requires a strategy to adapt its virtual inertia. This paper proposes a reinforcement learning-based method with fast reward shaping capability to enhance the system’s learning and stability. The results of the proposed method are implemented through MATLAB/Simulink and compared with other methods to demonstrate the effectiveness.