基于时间序列的多采样尺度风电功率多步预测研究  

Multi-Step Prediction Study on the Multi-Sampling Scale Wind Power Based on the Time Series Method

作　　者：杨茂[1] 孙涌[1] 王东[2] 穆钢[1] 张明明[3] 

机构地区：[1]东北电力大学电气工程学院,吉林吉林132012 [2]国网北京市电力公司,北京100031 [3]天津国投津能发电有限公司,天津300480

出　　处：《电测与仪表》2014年第23期55-59,109,共6页Electrical Measurement & Instrumentation

基　　金：国家重点基础研究发展计划项目(973计划)(2013CB228201);国家自然科学基金(51307017);吉林省科技发展计划项目(20140520129JH);吉林省教育厅"十二五"科学技术研究项目(吉教科合字[2014]第474号);吉林市科技发展计划资助项目(2013625004)

摘　　要：风力发电联网运行是实现风能大规模开发利用的主要途径,准确的风电功率实时预测是实现风能大规模合理调度的有效手段。文章建立了基于时间序列的多采样尺度风电功率多步预测模型。对于周期分量利用建模域均值法进行预测,剩余分量通过平稳性检验后利用多采样尺度时间序列法进行预测。利用两个不同风场的实测风电功率数据,进行不同时间段的实时预测,与持续法和时间序列法的滚动预测模型相比,该预测模型能有效地提高预测精度。Wind power generation with network running is the main means of realizing the exploitation and utilization of wind energy on a large scale.Accurate real-time prediction of wind power is an effective means of achieving large-scale rational management of wind energy.A multi-step prediction model of multi-sampling scale wind power was established based on the time series method.The mean method of domain modeling was adopted to predict the periodic component, while the multi-sampling scale time series method was adopted to predict the remaining components after the stability test.By means of the measured data of the wind power from two different wind farms, real-time predic-tion was made at different times, which was compared with the rolling prediction models based on the persistence method and the time series method.The results showed that this prediction model could effectively improve prediction accuracy.

关 键 词：风电功率 周期分量 建模域均值法 时间序列 多采样尺度 

分 类 号：TM71[电气工程—电力系统及自动化]