Short-term wind power prediction of wind farms based on LSTM+NARX neural network

Abstract

Accurate short-term wind power prediction is of great significance for large-scale wind power grid security and stability. According to the characteristics of intermittent and randomness of the wind, this paper puts forward a kind of based on Long Short-term Memory network (Long Short Term Memory, LSTM) and Nonlinear regression neural networks (Nonlinear Autoregressive models with Exogenous Inputs, NARX) wind power prediction method. Using the LSTM to short-term prediction of wind speed time series data avoid the problems of ladder loss and gradient explosion. The output of LSTM is used as the input of NARX, the delay of input parameters is determined, and a hybrid model of LSTM+NARX is established to predict the wind power in the future 48h. Compared with the prediction results of BP neural network and NARX neural network, the root mean square error is 2.83%, and the prediction accuracy is higher than that of other methods.