基于贝叶斯优化XGBoost的短期风功率预测

Short-Term Wind Power Prediction Based on Bayesian Optimization XGBoost

针对风电场输出功率不稳定导致风功率预测精度低的现象,提出一种基于贝叶斯优化极端梯度提升(BOA_XGBoost)的短期风功率预测模型。将极端梯度提升(XGBoost)、支持向量机(SVM)、核非线性回归(KNR)三个基础预测模型以及贝叶斯优化后的极端梯度提升(BOA_XGBoost)、支持向量机(BOA_SVM)、核非线性回归(BOA_KNR)进行对比,通过内蒙古自治区某风电场的实测数据对六种预测模型进行验证,结果表明BOA_XGBoost模型预测效果最佳,有效改善了风功率预测效果,提高了风电并网运行的安全性。

A short-term wind power prediction model based on BOA_XGBoost is proposed to solve the low accuracy of wind power prediction caused by unstable output power of wind farms.The three basic prediction models of XGBoost,SVM and KNR,and the models of extreme gradient lifting(BOA_XGBoost),support vector machine(BOA_SVM)and kernel nonlinear regression(BOA_KNR)after Bayesian optimization are compared.The six prediction models are verified by the measured data of a wind farm in Inner Mongolia autonomous region.The results show that BOA_XGBoost model has the best prediction effect,effectively improving the wind power prediction effect and improving the safety of wind power grid connected operation.