A Sensitivity Based Strategy for Wide-Area Voltage Control by Q-Learning

A smart scheme for dynamic voltage control in a both wide-area and real-time scheme is presented in this article. The proposed strategy is based on a combination of Q-learning concept and multi-agent system to determine appropriate coordination of reactive power resources for maintaining voltage security of power system using synchrophasor data. The proposed algorithm takes the advantages of sensitivity analysis to create Q-table in an offline manner and then updating this table after applying remedial actions in grid in an online method. Here, these remedial actions are reactive power resources (such as capacitor banks, reactors, and on-load tap changers). To evaluate the effectiveness and robustness of the proposed scheme, various simulations have been performed in the Nordic test system.