Reinforcement learning tuned decentralized synergetic control of power systems

In this paper, decentralized synergetic controllers with varying parameters are developed to dampen oscillations in electric power systems via the excitation systems of the generators. Each generator is treated as a subsystem for which a synergetic controller is designed. Each subsystem is a dynamical system driven by a function that estimates the effect of the rest of the system. A particle swarm optimization (PSO) technique is employed to initialize the controllers’ gains. Then, reinforcement learning (RL) is used to vary the gains obtained after implementing the PSO so as to adapt the system to various operating conditions. Simulation results for a two area power system indicate that this technique gives a better performance than synergetic fixed gains controllers, or conventional power system stabilizers. Simulation results are obtained using the power analysis toolbox (PAT).

► Each generator is modeled as a subsystem and the effect of the rest of system is considered as an external disturbance.
► This disturbance is modeled as a polynomial function of the active power generated by the subsystem.
► Decentralized synergetic controller is designed for each subsystem.
► Particle Swarm Optimization technique is employed to tune the controller parameters.
► Reinforcement learning algorithm is employed to vary some of controller parameters to improve the performance of the controller.