Robust Adaptive Load Frequency Controller Based on Reinforcement Learning in an Inter-Connected Power System

The aim of this paper is using reinforcement learning for designing of robust and adaptive PID and SMES controllers to load frequency control in a two area thermal power system. Thus, in first setting of PID and SMES controller parameters formulated as an optimization problem and solved using teaching-learning optimization algorithm. Then the simultaneous performance of designed controllers improved using proposed reinforcement learning based controller. Simple and understandable structure and easy to use are distinguished advantages of q-learning based controllers. In order to evaluate the performance of the proposed controller, computer simulations have been done by using MATLAB software. Simulation results verified that the proposed q-learning based controller exhibits much better performance from the conventional optimization based controllers from viewpoint of time domain performance indices like over shoot, under shoot, ITAE, ITSE, and IAE.