Adaptive Control of Synchronous Generator with Transient Model Using Deep Reinforcement Learning

This study proposes a novel approach to improve the stability of a Single Machine Infinite Bus (SMIB) system by developing a learning-based Power System Stabilizer (PSS) utilizing the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. The proposed controller adjusts the gain of the PSS in response to disturbances, ensuring robust control against various disturbances. In this research, TD3 is used because of its capability to reduce sensitivity to actions and enhance overall stability. The performance of the learning-based PSS is evaluated under both small and transient disturbances, and its effectiveness is compared to a Conventional Power System Stabilizer (CPSS). Under small disturbances, both learning-based PSS and CPSS can maintain system stability. However, compared with the metrics used which are Integral of Absolute Error (IAE) Integral of Squared Error (ISE), and settling time, learning-based PSS performs better. Furthermore, under transient disturbance, only learning-based PSS can maintain system stability. The results demonstrate that the learning-based PSS can significantly enhance system stability against various disturbances, making it a promising solution for improving the robustness of power systems.