Short-term wind power forecasting method based on spatial-temporal graph neural network

Abstract

Abstract In order to solve the problem that the weather factor is neglected in the current short-term wind power forecasting process, which leads to a big difference between the power forecasting result and the actual power, a short-term wind power forecasting method based on the spatial-temporal graph neural network is proposed. According to the operating power data, the short-term wind power forecasting sequence is calculated, and the wind farm is regarded as a graph. The dependence of space and time series is captured by a graph neural network, and the spatial-temporal graph neural network model is constructed. Combined with the wavelet decomposition process, short-term wind power forecasting is realized. The experimental results show that the average absolute error and average relative error of this method are less than 10%, and the wind power prediction results at different times are all on the actual wind power curve, which shows that this method can accurately predict short-term wind power.