Surrogate-Assisted Evolutionary Optimization of the Emergency Load Shedding with Parallel Computation

Evolutionary algorithms are widely used for solving the event-driven emergency load shedding (ELS) optimization problem. However, conventional evolutionary ELS optimization methods are inefficient due to the computationally expensive time- domain simulation for evaluating candidate ELS schemes. To improve optimization efficiency, a parallel surrogate-assisted evolutionary algorithm for ELS optimization (PSEALS) is proposed in this paper. In the framework of PSEALS, search operators based on differential evolution are used to generate candidate ELS schemes. The data-driven surrogate model is constructed online to evaluate candidate ELS schemes efficiently. According to evaluation results, elitist ELS schemes are prescreened to direct accurate evolution search. The overall framework of PSEALS is organized with the parallel computation method to accelerate the solving process. The performance of the proposed algorithm is verified through the IEEE 39-bus system.