Wind Power Forecasting Based on CNN and LSTM Models

Abstract

Nowadays more and more sustainable and environmentally friendly energy sources are demanded and wind power is playing an increasingly significant role in the energy market. For example, according to the Chinese Academy of Sciences, China plans to generate more than ten times the current scale of wind power in the next decade. Accurate prediction of wind generation is essential for the scheduling, operation, and energy trading of power systems. In this study, a method combining long and short-term memory network (LSTM) and convolutional neural network (CNN) is proposed for wind power generation prediction. In this study, a wind power generation data set from Longyuan Power Group is used to adapt to the input requirements of the model. The data are entered into the LSTM model to capture long-term trends in the data. Next, this paper uses the CNN to process the local spatial features to improve the prediction accuracy. Combining the outputs of these two models, this paper obtains an efficient wind power generation prediction method. In order to evaluate the performance of the proposed method, root mean square error is used to analyze the prediction data. The results show that the combined method of LSTM and CNN is better than individual methods in prediction accuracy and stability. In conclusion, this study proposes a wind power forecasting method combining LSTM and CNN. This method helps to improve the dispatching and operation efficiency of the power system and provides strong support for the promotion of wind power in the energy market.