Interpretable wind power probabilistic prediction based on NGBoost

Abstract

To realize the probabilistic prediction of wind power and analyze the influencing factors of the prediction results， this paper proposes a probabilistic prediction method of wind power based on natural gradient boosting （NGBoost） and takes account of interpretable wind power probabilistic forecast method. Firstly， the definition of a probabilistic wind power prediction model is given based on the analysis of wind power characteristics. The NGBoost algorithm is used to train the prediction model to achieve probabilistic prediction considering the heteroskedasticity characteristics of wind power. Secondly， the Shapley value in cooperative game theory is used to interpret the prediction model and analyze the influence of meteorological factors on the prediction results. Finally， the prediction performance of the model is verified using actual wind farm data and compared with other methods. The results show that the proposed method achieves good prediction effect and can explain the prediction results and analyze the influence of meteorological factors on the prediction results. The method is practical and effective.

Keywords

• wind power probabilistic prediction
• wind power uncertainty
• ngboost
• shapley value