基于集群划分和LightGBM-KDE的区域风电中期概率预测

The Medium-Term Probabilistic Prediction of Regional Wind Power Based on Cluster Division and LightGBM-KDE

随着新能源保供电重要性日渐增强,调度人员对中期风电功率预测的需求程度也进一步加深。针对气象资源时空分布时序差异性波动大以及随着预测前瞻时间的延迟,区域风电功率预测不确定性合理评估难的问题,提出了基于集群划分和LightGBM-KDE的区域风电中期概率预测。首先对空间气象特征参数进行有效识别,并对风电集群出力波动性进行研究,利用减法聚类对风电场集群进行合理划分;然后基于集群划分结果,利用LightGBM算法建立中期风电功率预测模型;最后对功率预测误差采用非参数核密度估计计算概率密度分布,最终建立基于集群划分和LightGBM-KDE的240 h区域风电中期概率预测模型。算例结果表明,所提方法在区域风电中期概率预测中具有更高的精度。

With increasing importance of ensuring power supply through new energy resources,the demand for mediumterm wind power prediction by power dispatchers has deepened.Focusing on large fluctuations in temporal and spatial distribution of meteorological resources and the difficulty of reasonably evaluating the uncertainty of regional wind power forecast with the delay of forecast prospect time,this paper proposes a medium-term probabilistic forecast method for regional wind power based on cluster division and LighTBM-KDE.Firstly,the spatial meteorological characteristic parameters are identified effectively,and the output fluctuation of wind power clusters is studied,and the wind power clusters are divided reasonably by subtraction clustering.Secondly,based on the cluster division results,the LightGBM algorithm is used to build a medium-term wind power prediction model.Furthermore,the non-parametric kernel density estimation method is used to calculate the probability density distribution of power prediction errors,and finally a medium term probability prediction model of 240-hour regional wind power based on cluster division and LighGBMKDE is established.The numerical results show that the proposed method has higher precision in the medium-term probabilistic prediction of regional wind power.