Rotor angle stability enhancement using DDPG reinforcement agent with Gorilla troops optimized input scaling factors

This paper introduces a Reinforcement Learning (RL)-based Power System Stabilizer (PSS) with a Deep Deterministic Policy Gradient (DDPG) algorithm for rotor angle stability. The proposed stabilizer uses scaled values of the generator's accelerating power, a derivative of accelerating power, integration of accelerating power, and generator real power as inputs. The stabilizer uses the DDPG algorithm to train The RL agent. Moreover, to further enhance the PSS performance, the scaling factors of the input observations are optimized using the Gorilla Troops Optimization (GTO) algorithm, which is known for its robustness, fast convergence. Furthermore, the RL reward considered is a discrete function that rewards the generators' accelerating power samples when they are below a defined value. The proposed PSS is tested on three popular case studies: a Single Machine connected to an Infinite Bus (SMIB), Kundur's four-machine system, and the IEEE 39 bus ten machine system. The proposed stabilizer performance is compared with three common IEEE common PSSs: the Multiband dw speed-based PSS (MB-PSS), the lead-lag dw speed-based PSS (dw-PSS), and the lead-lag dPa accelerating power-based PSS (dPa-PSS). MATLAB simulations prove that the proposed PSS performs better than the other PSSs. It shows higher transient stability capability than the compared PSS even with long duration faults.