Short-term wind power interval prediction based on fitting error sequences with asymmetric Gaussian distribution

Accurate short-term wind energy prediction is crucial for wind power systems, and over the years, with the rapid development of artificial intelligence and machine learning techniques, researchers have proposed many models to improve the performance of wind power prediction systems. Traditional interval prediction methods usually assume that the residual series obey a simple Gaussian distribution. However, this is not always the case in reality, which can affect the accuracy of interval prediction. In this study, we develop a new interval prediction model that combines the improvement of point prediction accuracy with more efficient interval prediction results based on it. The study improves the accuracy of point prediction by combining SSA data preprocessing with Cao algorithm to mitigate the noise present in complex time series data. Subsequently, we used an asymmetric Gaussian distribution to fit the residual series to obtain more efficient interval prediction results. To evaluate the model, we conducted a case study using wind power data from October to December 2023 from a wind farm in Belgium. The experimental results show that the integrated model proposed in this paper effectively improves the accuracy of point forecasts as well as interval forecasts compared to a single model.