CEEMDAN-LSTM-RF-RBF short-term wind power forecast based on incomplete data

A combined CEEMDAN-LSTM-RF-RBF wind power forecast model based on incomplete data is proposed for addressing the problem of incomplete data affecting forecast accuracy due to the difficulty in obtaining wind power data in practical engineering. The wind power historical power data are decomposed using CEEMDAN and classified into high-frequency, medium-frequency, and low-frequency according to the zero crossing rate of each mode component. The forecasts are performed separately, and the final results are obtained by linear summation. The simulation shows that the forecast results of the combined forecast model basically match the trend of the actual values. Its RMSE, MSE, and MAE evaluation indexes are 6.348 kW, 40.2905 kW², and 2.7179 kW, respectively. Compared with the LSTM and RF models, the evaluation indexes of the combined model values are lower, the forecast values are closer to the actual values, and can obtain higher forecast accuracy in the case of incomplete data.