No abstract

An advanced control scheme to deal with transient angle instability and low-frequency angle oscillations in power systems is proposed in this paper. The control combines an existing controller (derived from the concept of Lyapunov energy functions) and deep Q-network (a deep reinforcement learning algorithm) to control a static VAr compensator. This control is modified in this paper in a way to directly control thyristor controller reactor part of the compensator, offering an easier implementation. The aim is to improve performances of existing control through the use of deep Q-network while retaining its basic system stabilization characteristics. Advantages of the proposed control scheme are illustrated by implementation of a model of four-machine test power system (this system is considered as the benchmark when studying the phenomena dealt with in this paper) in MATLAB/Simulink environment and using TensorFlow toolkit for deep Q-network usage.