考虑山东近海不同风能天气特征的风电功率区间预测模型  

Wind Power Interval Forecasting Model Considering Different Wind Energy Weather Characteristics in Shandong Offshore Areas

作　　者：余沣 董存 王铮 蒋建东[1] 王勃 YU Feng;DONG Cun;WANG Zheng;JIANG Jiandong;WANG Bo(School of Electrical Engineering,Zhengzhou University,Zhengzhou 450001,Henan Province,China;National Electric Power Dispatching and Control Center,Xicheng District,Beijing 100031,China;State Key Laboratory of Operation and Control of Renewable Energy&Storage Systems(China Electric Power Research Institute),Haidian District,Beijing 100192,China)

机构地区：[1]郑州大学电气工程学院,河南省郑州市450001 [2]新能源与储能运行控制国家重点实验室(中国电力科学研究院有限公司),北京市海淀区100192 [3]国家电力调度控制中心,北京市西城区100031

出　　处：《电网技术》2020年第4期1238-1246,共9页Power System Technology

基　　金：国家重点研发计划项目(2018YFB0904200);国家电网有限公司科技项目(SGLNDKOOKJJS1800266)。

摘　　要：风电功率区间预测是应对大规模风电机组并网运行的有效手段之一。针对山东风电并网运行建立了一种考虑山东半岛不同风能特征的风电功率区间预测模型。对比了不同风能条件下半岛内风电场出力特征和风电功率历史预测误差分布特点,发现风电场出力分布范围随风速增加呈"先增后减"趋势,在出力分布范围较大的风速区间内,预测误差也相对较大。以风速、风向和预测功率为特征变量,在利用层次聚类法对样本数据进行聚类分析基础上,采用非参量核密度估计方法,建立了各类样本在不同风向条件下风速-风电功率预测误差的联合概率密度分布模型。将该模型与NARX(nonlinear auto regressive models with exogenous inputs)网络确定性风电功率预测结果相结合,得到一定置信水平的风电功率区间预测结果,最后通过实际算例验证了模型的有效性。Wind power interval forecasting is one of effective means to cope with the challenges of large-scale grid integration of wind turbines. This paper builds an interval forecasting model of wind power considering the different characteristics of wind resource in Shandong Peninsula. By comparing the output characteristics of wind farms and the distribution characteristics of historical forecasting errors of wind power in the Peninsula on the condition of different wind energy, it turns out that for the winds with equal speeds but different directions, the distribution range of wind farms’ output presents a trend of " increase first and then decrease" with wind speed increase. In the distribution range of larger wind speed, the forecasting errors of wind power are relatively big. Taking wind speed, wind direction and forecasting power as characteristic variables, hierarchical clustering method is used to carry out analysis on the sampled data. Nonparametric kernel density estimation is used to build a joint probability density distribution model of the forecasting errors of wind speed and wind power for a variety of samples in the conditions with different wind directions. Combined with the power forecasting results of NARX(nonlinear auto regressive models with exogenous inputs) network, the model proposed in this paper can obtain an interval forecasting result with certain confidence level. The performance of this model is verified through actual case studies.

关 键 词：区间预测 层次聚类 非参量核密度估计 联合概率密度分布 NARX网络 

分 类 号：TM614[电气工程—电力系统及自动化]