Research on Generation Wind Onshore Forecasting Method Based on Enhanced Long Short-Term Memory Hybrid Network Model

Generation wind onshore and unit electricity price are important factors for wind power generation enterprises’ revenue and power grid stability. In this study, we proposed a method to predict wind power generation and unit price by using historical wind power generation, unit electricity price and current weather conditions, and designed a parallel convolutional neural network (CNN) architecture to map low-dimensional data into high-dimensional space by extracting features and embedding codes from historical data and current data, and used the long short-term memory (LSTM) network model to learn time series data after flattening the data. The experimental results show that the fit of the improved parallel CNN-LSTM hybrid model to the model reaches $\mathbf{9 7 . 1 1 \%}$, and the training times are reduced by $50 \%$ compared with the LSTM model, we used lower training times to obtain a model with higher fitting accuracy.