Volume 63

Improved Probability Prediction of Distributed Photovoltaic Power Based on Optimized KMC-Vine Copula Xiaotong Yang, Zuan Fu, Jie shi

Abstract

Distributed photovoltaic (DPV) power generation has increasingly emerged as one of the most promising renewable energy sources due to its extensive distribution and flexible installation. However, its output is highly influenced by solar irradiation conditions, exhibiting diurnal and seasonal variations, with weather conditions being the most significant influencing factor. To ensure the secure and stable operation of the power system and maintain an appropriate balance between supply and demand, predicting DPV power output presents a viable solution. This paper proposes an improved KMC-Vine Copula model for probabilistic prediction of DPV power by integrating spatio-temporal influencing factors and the prediction results of centralized PV power. Using copula correlation theory, the spatio-temporal dependencies between centralized and distributed PV systems are figured out. Conventional single copula functions are inadequate in capturing complex correlations. Thus, an optimized copula model is introduced to address the above limitation. However, due to its inability to represent high-dimensional dependencies, a flexible Vine Copula model is constructed as the final prediction framework. Through the case study, compared with the optimized copula model, the proposed model improved by 26.87% (interval width), 41.6% (reliability), and 39.23% (sharpness), respectively, under three evaluation indexes.

Keywords Terms-distributed photovoltaic, k-means clustering, optimized copula, quantile regression, vine copula