基于偏最大信息系数与组合XGBoost的短期风功率预测

作为新能源领域的课题热点之一，短期风功率预测的研究在提高预测精度的同时也应重视模型的工程化应用。据此，提出一种基于偏最大信息系数的组合XGBoost预测模型。首先，设计一种基于偏最大信息系数的特征选择算法，通过引入偏互信息，在挖掘出对风功率影响较大的气象特征的同时，也能消除耦合信息带来的不利影响。在此基础上，为兼顾模型的精度和计算效率，降低单个模型的预测风险，构建以XGBoost为底层算法的组合预测模型，进一步实现风功率预测。采用2个具有较大差异的风电场作为算例进行验证分析，结果表明，基于偏最大信息系数特征选择算法的组合XGBoost预测模型不但能提升短期风功率的预测精度，与相近的组合预测模型相比，也具备更高的计算效率，有利于工程化应用。

Combined XGBoost short-term wind power forecasting model based on partial maximum information coefficient

As one of the hot topics in the field of new energy forecasting,it is necessary for the research of short-term wind power forecasting to pay attention to the engineering application of the model while improving forecasting accuracy. Hence,a combined XGBoost forecasting model based on partial maximum information coefficient is proposed. To begin with,a feature selection algorithm based on partial maximum information coefficient is designed. By introducing partial mutual information,while mining meteorological features that have a greater impact on wind power,it can also eliminate the adverse effects of coupled information. On this basis,in order to take the accuracy and computational efficiency of the model into account and reduce the forecasting risk of a single model,a combined forecasting model with XGBoost as the underlying algorithm is constructed to further realize wind power forecasting. Two wind farms with large differences are used as examples for verification analysis. The results show that the combined XGBoost forecasting model based on partial maximum information coefficient feature selection algorithm can not only improve the forecasting accuracy of short-term wind power,but also has higher calculation efficiency compared with similar combined forecasting models,which is beneficial to engineering application.