基于相似日和特征提取的短期风电功率预测  

Short-Term Wind Power Prediction Based on Similar Day and Feature Extraction

作　　者：张颖超[1,2] 成金杰 邓华[1,2] 宗阳 章璇 ZHANG Yingchao;CHENG Jinjie;DENG Hua;ZONG Yang;ZHANG Xuan(School of Automation,Nanjing University of Information Science&Technology,Nanjing 210044,China;Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters,Nanjing University of Information Science&Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学自动化学院,江苏南京210044 [2]南京信息工程大学气象灾害预报预警与评估协同创新中心,江苏南京210044

出　　处：《郑州大学学报（工学版）》2020年第5期44-49,共6页Journal of Zhengzhou University（Engineering Science）

基　　金：国家自然科学基金资助项目(41675156);江苏省六大人才高峰项目(WLW-021)。

摘　　要：为提高短期风电功率预测精度,增强预测模型对特定天气状况的代表性和适应性,提出一种基于离散Frechet距离与核熵成分分析(KECA)相结合的数据处理方法。通过引入离散Frechet距离,建立匹配相似日的数学模型,提取与预测日相似的样本,使用KECA从多种气象要素中提取合适的非线性主元作为支持向量机(SVM)模型的输入。实验结果表明:所提出的方法明显提高了预测精度并具有一定的适用性。In order to improve the short-term wind power prediction accuracy and enhance the representativeness and adaptability of the prediction model to specific weather conditions,a data processing method was proposed based on discrete Fréchet distance and kernel entropy component analysis(KECA).Using Fréchet distance,this paper established a mathematical model matching similar days,extracted samples similar to the prediction date,and then used KECA to extract suitable nonlinear principal elements from various meteorological elements as input of support vector machine(SVM)model.The simulation verification showed that the proposed method could significantly improve the prediction’s accuracy and has certain applicability.

关 键 词：离散Frechet距离 相似日 核熵成分分析 支持向量机 功率预测 

分 类 号：TM614[电气工程—电力系统及自动化]