The multiagent cooperative AGC strategy for multiregional energy interconnection

Abstract

<p indent=0mm>Traditional reinforcement learning cannot effectively solve the strong random disturbance problem caused by large-scale new energy access; therefore, the response speed of AGC (automatic generation control) becomes slower and the performance becomes poorer. This study considers the double Q learning that can solve the problem of overestimation of the value of state-action pairs in the reinforcement learning algorithm based on the Q framework as the fulcrum. It integrates a grey wolf optimization algorithm that can quickly search for the optimal solution in the unknown search space. Additionally, it proposes a multiagent cooperative AGC strategy for multiregional energy interconnection, namely, the GWDQ strategy, to quickly obtain the multiregional cooperative optimal solution in the AGC process. The two-area integrated energy system model, such as the hybrid-electric gas turbine system, CCHP and other forms of energy, and the northeast power grid model with a multiregional energy interconnection, are simulated. The results showed that the proposed strategy has a strong learning ability. Besides, the convergence speed and control performance are superior to that of the traditional reinforcement learning algorithm. The proposed strategy can quickly obtain the multiregional collaborative optimal solution in the AGC process.