A Multi-dimensional Copula Wind Speed Correction Method for Ultra-Short-Term Wind Power Prediction

Abstract

The focus of this paper is how to make comprehensive use of numerical weather prediction (NWP) and real power data to correct the forecasted wind speed, to improve the accuracy of ultra-short-term wind power prediction. Firstly, the multi-dimensional Copula model is built by 3 time series: current wind speed, previous wind speed error and current wind speed error. Determinate the first two values, and the current wind speed error can be obtained by gradually solving the conditional probability. Then, use the current wind speed error to correct and replace forecasted wind speed in NWP. Finally, take corrected NWP as the input of Long Short-Term Memory (LSTM), and take real power as the output. After selecting appropriate parameters to train the LSTM model, this hybrid model can be used to predict the ultra-short-term wind power. Compared with Autoregressive Integrated Moving Average (ARIMA), which only considers the time-series characteristics of wind speed error, this hybrid model can bring 2.3% reduction of mean absolute percentage error in prediction.