Short-Term Power Prediction of a Wind Farm Based on Empirical Mode Decomposition and Mayfly Algorithm-Back Propagation Neural Network

With the improvement of energy consumption structure, the installed capacity of wind power increases gradually. However, the inherent intermittency and instability of wind energy bring severe challenges to the dispatching operation. Wind power forecasting is one of the main solutions. In this work, a new combined wind power prediction model is proposed. First, a quartile method is used for data cleaning, namely, identifying and eliminating the abnormal data. Then, the wind power data sequence is decomposed by empirical mode decomposition to eliminate non-stationary characteristics. Finally, the wind generator data are trained by the MA-BP network to establish the wind power prediction model. Also, the simulation tests verify the prediction effect of the proposed method. Specifically speaking, the average MAPE is decreased to 12.4979% by the proposed method. Also, the average RMSE and MAE are 107.1728 and 71.604 kW, respectively.