Ultra-Short-Term Wind Speed Forecasting Based on CNN-LSTM-ARIMA
Improving wind speed prediction accuracy is important for timely adjustments in power system scheduling plans and enhancing competitiveness in the wind energy market.This paper presents an ultra-short-term wind speed forecasting method based on an ensemble of convolutional neural network(CNN),long short-term memory network(LSTM),and autoregressive integrated moving average(ARIMA).Firstly,CNN convolutional layers capture patterns and local features in time series data.Subsequently,it utilizes LSTM models to learn and train on the extracted features.Based on the CNN-LSTM composite architecture model,it predicts future wind speeds and compares them with actual data to obtain residuals.Finally,it employs ARIMA to analyze historical residuals to correct future prediction errors,achieving ultra-short-term wind speed forecasting.Using measured wind speed data from a wind farm in Turkey as an example,it predicts wind speeds for the next 10 minutes.The results indicate that compared to traditional neural network models like CNN-LSTM and LSTM-LSTM,the CNN-LSTM-ARIMA model has reduced the mean absolute error in wind speed forecasting by 16.40%and 26.92%,respectively,significantly enhancing the prediction accuracy.
wind speed predictionconvolutional neural networkslong short-term memory networkautoregressive integrated moving average model
NETL
NSTL
万方数据
