Damping inter-area oscillation using reinforcement learning controlled TCSC

Inter-area oscillation is a serious problem that threatens a power system. Appropriate damping control of the inter-area oscillation would ensure the grid stability and maintain the tie-line power transfer capability. In this paper, a novel reinforcement learning (RL) based power oscillation damping (POD) controller is proposed that uses Thyristor Controlled Series Compensators (TCSC) to damp inter-area oscillations. By leveraging the unbiased gradient direction estimation of the natural evolution strategy (NES), the power flows on the tie-lines were successfully regulated and inter-area oscillations were damped through dynamically modulating the inserted reactance of the TCSC. Furthermore, parallel computation techniques were adopted to speed up the training process of the NES. The proposed RL-based POD controller has been tested on both two-area four-machine system and North American MinniWECC system. Extensive studies have demonstrated the excellent performance of the proposed RL-based TCSC POD controller in damping inter-area oscillations.