Bi-objective Pareto optimization for clustering-based hierarchical power control in a large-scale PV power plant

Abstract

To assist the power grid for the real-time power balance, more and more PV power plants have participated in automatic generation control (AGC). In this process, a PV power plant should distribute the dynamic total AGC signal to all the controllable inverters. Firstly, it is difficult to satisfy multiple operation requirements for the whole PV power plant. Secondly, it easily results in a high optimization complexity due to the large amount of controllable inverters, especially for a large-scale PV power plant. To handle these two problems, a clustering-based hierarchical power control is constructed with two conflict objectives, including the AGC responding performance and the reactive power reserve. With the discrepant responding performance of different inverters, all the inverters are clustered into multiple groups by the K-medoids algorithm according to their regulation parameters. Hence, the hierarchical power control contains a top-layer control and a bottom-layer control. An efficient non-dominated sorting genetic algorithm is adopted to acquire the Pareto optimal solutions for the top-layer control, then the bottom-layer control is implemented in proportion to their regulation capacities. Finally, the case studies are carried out to verify the proposed technique for a large-scale PV power plant under different AGC signals and dynamic irradiations.