A Probabilistic Wind Power Forecasting Approach Based on Gaussian Process Regression

摘要

Wind power forecasting (WPF) is an effective way to eliminate the negative impacts and enhance the market competitiveness of wind power. Probabilistic WPF gives the prediction intervals of possible output power, which may largely improve the confidence level and accuracy of the wind power forecasts. Considering the fluctuation characters of wind power, a combination scheme of different Gaussian kernels is used to establish the Gaussian Process Regression model, which is used to predict the mean value and the confidence interval of output power by taking wind speed and wind direction as input variates. Via the case analysis of a real wind farm in North China, results show that, the proposed Gaussian kernel scheme can well describe the fluctuation process of wind power, and the uncertain intervals produced by the established probabilistic WPF model basically able to cover the measured power. Compared with a typical statistical WPF model based on neural network, the improvement of monthly forecasting precision is up to 1.8 %.