A long short-term memory based wind power prediction method

Abstract

Wind power prediction is the basis of power grid energy dispatching. However, wind instability increases the difficulty of wind power prediction. The paper proposes a wind power prediction method based on long and short-term memory network to improve the accuracy of wind power prediction. First, wind power sequence is decomposed by empirical mode decomposition (EMD) method, and the noise in the original sequence was removed by effective component reconstruction. Then, long shortterm memory (LSTM) with the ability of information memory predicts model of wind power sequence. The improved particle swarm optimization algorithm (IPSO) optimized the parameters of LSTM to solve the problem that the parameters of LSTM, such as the number of neurons, the learning rate and the number of iterations, are difficult to determine and thus affect the prediction accuracy of the model. Finally, the proposed EMD-IPSO-LSTM method makes rolling prediction of wind power series of actual wind farm, and the prediction results are compared with other prediction models. The results show that the prediction model has higher accuracy.