考虑爬坡特征量的海上风电短期分区功率预测  

SHORT-TERM ZONING POWER PREDICTION OF OFFSHORE WIND POWER CONSIDERING CLIMBING FEATURE QUANTITIES

作　　者：时帅 张皓 黄冬梅 李媛媛 米阳 杨晓东 Shi Shuai;Zhang Hao;Huang Dongmei;Li Yuanyuan;Mi Yang;Yang Xiaodong(School of Electric Power Engineering,Shanghai University of Electric Power,Shanghai 200090,China;School of Electronics and Information Engineering,Shanghai University of Electric Power,Shanghai 201306,China;Clean Energy Branch of Huaneng(Zhejiang)Energy Development Co.LTD.,Hangzhou 310014,China)

机构地区：[1]上海电力大学电气工程学院,上海200090 [2]上海电力大学电子与信息工程学院,上海201306 [3]华能(浙江)能源开发有限公司清洁能源分公司,杭州310014

出　　处：《太阳能学报》2024年第12期258-268,共11页Acta Energiae Solaris Sinica

基　　金：国家重点研发计划(2021YFC3101602);华能集团总部科技项目(HNKJ20-H66)。

摘　　要：考虑复杂海洋工况,提出一种考虑爬坡特征量的分区预测混合模型。首先,采用改进的bump事件检测技术对功率波动时段进行识别与划分;其次,综合考虑不同典型气象日的功率波动特征,对气象数据进行分类;最后,计及风电功率波动性的基础上提出一种混合预测模型,在功率非连续波动段采用一种LightGBM决策树与长短期记忆神经网络(LSTM)的点预测组合模型,在功率连续波动时段采用随机森林(RF)与LSTM的区间预测组合模型,并获得较好的预测效果。最后,选取中国东部某海上风电场的数据进行改进模拟与算例分析,结果表明,相较于传统的风电功率点预测与区间预测方法,考虑风电爬坡与气象日分类的分区混合预测模型的预测精度有明显提升。Considering complex marine conditions,this paper proposes a zonal hybrid prediction model considering climbing features quantities.Firstly,the improved bump detection technology is used to identify and divide the power fluctuation period.Secondly,this paper comprehensively considers the power fluctuation characteristics of different typical meteorological days to classify the meteorological data.Thirdly,considering the power volatility of wind power,this paper proposes a hybrid prediction model,which adopts a LightGBM-LSTM point prediction model in the power discontinuous fluctuation section and an RF-LSTM interval prediction model in the power continuous fluctuation period,and obtains a good prediction effect.Finally,the data of an offshore wind farm in eastern China is selected for improved simulation and case analysis.The results show that compared with the traditional wind power point prediction and interval prediction methods,the prediction accuracy of the partition hybrid prediction model considering wind power climbing and meteorological daily classification proposed in this paper is significantly improved.

关 键 词：爬坡特征量 风电爬坡事件 bump事件检测 气象日分类 组合模型 分区预测 

分 类 号：TM614[电气工程—电力系统及自动化]