中文摘要：

区域风电功率预测对于保障风电消纳及电网安全经济运行具有重要意义。由于新建风电场在并网初期尚未建立预测系统及各风电场预测精度参差不齐,经典的单场功率累加法预测精度并不高。提出一种基于风电功率数据特征聚类的区域风电功率统计升尺度预测方法,首先使用经验正交函数（empirical orthogonal function,EOF）法解析区域内风电出力特征,然后采用层次聚类法划分子区域,并利用风电场的相关系数和预测精度选取代表风电场,最后根据代表风电场的预测功率及权重系数完成区域风电功率的升尺度预测。应用冀北电网2015年的实际数据进行统计升尺度建模和方法验证。结果表明,相比累加法,文中提出的统计升尺度方法可改进区域风电功率预测精度,同时减少区域预测模型对单风电场数据完备性和预测精度的依赖。

英文摘要：

Regional wind power forecasting is of great significance to ensure wind power accommodation and safety and economical operation of power grid. However, classic single-field prediction power accumulation method is poor in accuracy, because new wind farms have not yet established a forecasting system when firstly connected to grid or prediction accuracy of various wind farms is different. This paper proposed a method of short-term regional wind power statistical upscaling forecasting based on feature clustering. Firstly, wind power output characteristics in the region is analyzed with empirical orthogonal function （EOF） decomposition. Then, hierarchical clustering method is used to divide the region and sample wind farms in each subarea are selected based on correlation coefficients and prediction accuracy of the wind farms. Finally, regional wind power prediction is calculated based on predicted power and weight coefficients of the sample wind farms. Actual data of Hebei Province in 2015 are applied to statistical modeling and method validation. Results show that the proposed method can improve prediction accuracy of regional wind power, and reduce dependence of regional prediction model on data integrity and prediction accuracy of single wind farm, compared with traditional accumulation method.