Optimisation of photovoltaic storage and charging system operation based on photovoltaic power generation prediction analysis

Abstract

Abstract The power generation of photovoltaic power generation systems has more influencing factors. Accurate PV power prediction data is needed in order to maintain the stable operation of the power system. In this paper, a photovoltaic power generation prediction scheme combining the integrated weather-similar day clustering method is designed with reference to the weather characteristics in the northeast region. The training samples are classified into three generalized weather small sample datasets. The radial basis (RBF) neural network prediction optimized with the Fruit Fly Algorithm (FOA) is chosen as the main body of the prediction system model. The prediction model is established. The power prediction is compared with the power prediction using the BP neural network model. The simulated prediction results are analyzed. The results indicate that the FOA-RBF neural network model exhibits superior performance in approximating nonlinear functions compared to the BP neural network. Ultimately, the forecasted data is utilized for operational enhancement, aiming for peak economic efficiency as the goal of optimization. This process considers limitations like energy equilibrium to enhance the financial viability of the optical storage charging facility, which improves the economic efficiency of the photovoltaic storage charging station, enhances the reliability of the system, and verifies the usability of the small-sample PV power prediction strategy based on FOA-RBF neural network designed in this paper for the northeastern region.