Deep Reinforcement Learning-Based Adaptive Inertia Control for Grid-Forming Power System

As renewable energy sources (RESs) are increasingly integrated into power grids, the power system exhibits characteristics of low inertia and low damping, which significantly impact system stability. Virtual synchronous generator (VSG) technology allows grid-connected inverters to replicate the external behavior of synchronous generators while providing flexible adjustments to virtual inertia, thereby enhancing power system stability. However, the relationship between frequency and inertia is often simplified as a piecewise or linear function, whereas in practical systems, this relationship is typically nonlinear. To tackle this problem, this study initially examines the influence of inertia and damping parameters on system stability using the VSG mathematical model. Then, an adaptive inertia control approach utilizing deep reinforcement learning is introduced to dynamically adjust the virtual inertia of the VSG, thereby improving system stability. Finally, multiple simulation experiments are performed to demonstrate the effectiveness of the proposed control strategy.