Photovoltaic station power prediction method, device and equipment based on twin neural network

Abstract

The invention provides a photovoltaic station power prediction method, device and equipment based on a twin neural network, and relates to the technical field of photovoltaic power generation power prediction.The method comprises the steps that weather forecast data corresponding to a to-be-predicted day and weather forecast data corresponding to each typical day in a support set are paired to obtain all data pairs, inputting all the data pairs to a preset twin neural network model to obtain a similarity set; determining a preset number of typical days determined according to the sequence of the similarities from high to low as all similar days, and determining a power prediction result corresponding to the to-be-predicted day according to the power generation power corresponding to all the similar days; and inputting each power prediction result corresponding to the to-be-predicted day under all time scales to a preset bidirectional fusion model to obtain the target power generation power of the to-be-predicted day. According to the method, dynamic changes of power data are comprehensively captured, the most similar typical daily power generation mode is screened out to serve as a reference basis, and the stability and accuracy of prediction are remarkably improved.

Images