Probability Estimation of Distributed Photovoltaic Power Prediction Based on KMC-optimized Copula

Abstract

Because of the natural randomness of solar energy and the unique property of distributed photovoltaic (DPV), the uncertainty of its output affects the safe and stable operation of the distribution network. DPV power probability prediction can provide more comprehensive prediction information, and it is one of the most important measures to solve the problem of large-scale DPV integration. Therefore, a DPV probability prediction algorithm based on k-means clustering(KMC)-optimized Copula is proposed in this paper. Firstly, the weather types are classified by the clustering method. Then, according to the characteristics of each weather attribute, the optimal representative weight of the Copula function is constructed via Genetic Algorithm (GA), and the KMC-optimized Copula algorithm is proposed to quantify the dynamic power space correlation between concentrated photovoltaic (CPV) and DPV. Finally, the DPV prediction value is dynamically expressed by the CPV power in the coming prediction duration, and the conditional prediction probability estimation of the DPV power is obtained. The case study shows that the proposed KMC-optimized Copula Model is of great importance for distributed power predictionand compared with the normal Copula model, the model performances of point prediction and probability prediction are both improved under different weather types.