Photovoltaic typical output scenario clustering method based on PSO-WKM algorithm

Abstract Extracting typical photovoltaic output scenarios is key to the efficient planning and configuration of hydro-photovoltaic complementary systems. Accurately capturing the characteristics of photovoltaic output curves and generating representative output scenarios presents a significant challenge. The Particle Swarm Optimization Weight K-Means (PSO-WKM) clustering algorithm is proposed to optimize the feature dimension weights of photovoltaic output curves. The algorithm converts the original curves to feature vectors, assigns initial weight factors to their components, and optimizes these factors using PSO to determine the importance of each feature, weakening the impact of noise dimensions on clustering precision. The final result is a set of photovoltaic output scenarios that consider weights of feature dimensions. The algorithm was used to cluster the 2022 photovoltaic daily output curves from the Beipan River Basin, resulting in a typical set of output scenarios. The results indicate that the PSO-WKM can accurately classify and extract typical photovoltaic output scenarios, outperforming other models in all clustering evaluation indicators, thus confirming its effectiveness and providing vital data for hydro-photovoltaic complementary system output decisions.