Feng Huang, Rui Wang, Yuelin Yu, Feiyu Hu, Xin Xie, and Lingxiang Huang
WPF, VMD, BA, IPSO, GRU neural network
Wind power forecasting (WPF) has important practical value for grid-connected systems. To address the difficulty of predicting wind power, an improved particle swarm optimisation (IPSO) algorithm for gated recurrent unit (GRU) power forecasting was proposed. Firstly, to deal with problems, such as the instability and large fluctuations of wind power data, the variational mode decomposition (VMD) algorithm was applied to preprocess historical wind power data. Then, a GRU neural network model was established, and the Bat algorithm (BA) and IPSO were used in the PSO process to obtain the hyperparameters of the GRU neural network, to determine the parameters of the forecasting model. Finally, the new VMD-BAIPSO-GRU neural network wind power algorithm was proposed to divide the VMD decomposed data into training and testing sets, and the model was then trained and tested. Compared with other similar algorithms, the VMD-BAIPSO-GRU algorithm has a lowers root mean square error (RMSE) of 0.37 and mean absolute percentage error (MAPE) of 12.09%, indicating higher forecasting accuracy.
Important Links:
Go Back
