Short-term Wind Speed Prediction Model based on GA-LSTM Network

With the large-scale wind power grid connection, accurate wind farm wind speed prediction is of great significance to promote the consumption of wind power energy and the safe and stable operation of power grid. Short-term memory(LSTM) has been widely used in wind farm short-term wind speed prediction because of its ability to capture dynamic time series effectively. However, the optimal parameters of the wind speed prediction model based on LSTM are closely related.This paper proposes a LSTM network optimized by genetic algorithm (GA). Firstly, a dual-layer LSTM network is built using LSTM units. Then, the number of hidden neurons in the LSTM network and the Dropout ratio are optimized using GA. Next, the time-series feature information is extracted from historical data of wind speed, wind direction, temperature, and humidity using the optimized LSTM network. Finally, the extracted features are used for wind speed prediction. An experimental simulation was conducted on the public data of a wind farm, and the experimental results show that the proposed method has a high accuracy in wind speed prediction.