Small-sample short-term photovoltaic output prediction model based on GRA-SSA-GNNM method

Abstract

The precision of photovoltaic (PV) output forecasting results is crucial to the reliability of the intelligent distribution network and multi-energy supplementary system. This work aims to address problems of insufficient research related to the short-term prediction of small-sample PV power generation and the low prediction accuracy in the previous research. A hybrid prediction model based on grey relation analysis (GRA) combined by sparrow search algorithm (SSA), and grey neural network model (GNNM) is proposed. In this paper, GRA is utilized to reduce the dimension of meteorological features of the samples. Then, a high-precision day-ahead short-term PV production forecast based on the SSA-GNNM model is established. The GNNM is used to perform regression analysis on the input features after reducing the dimension of meteorological features of the samples, and the parameters of the GNNM are optimised via SSA. The prediction results agrees well with the data from PV power plant in Xinjiang, indicating that the GRA-SSA-GNNM model developed in this work effectively achieves a high precision estimation in short-term PV power generation output prediction and has a promising application in this field.