陈锦铭1,郭雅娟1,伍旺松2,吴倩红2,韩〓蓓2,李国杰2.基于数据预处理与特征表示的多核SVM短期光伏发电预测[J].水电能源科学,2018,36(9):205-208
基于数据预处理与特征表示的多核SVM短期光伏发电预测
Multi kernel SVM Short term PV Power Prediction Based on Data Pre processing and Characteristic Representation
  
DOI:
中文关键词:  数据预处理  特征表示  小波阈值分析  二次聚类  多核支持向量机
英文关键词:data pre processing  characteristic representation  wavelet threshold analysis  twice clustering  multi kernel support vector machine
基金项目:国家电网公司科技项目(52100116001N)
作者单位
陈锦铭1,郭雅娟1,伍旺松2,吴倩红2,韩〓蓓2,李国杰2 1. 国网江苏省电力公司 电力科学研究院 江苏 南京 211103 2. 上海交通大学 电力传输与功率变换控制教育部重点实验室 上海 200240 
摘要点击次数: 49
全文下载次数: 
中文摘要:
      针对气温、降雨等气象数据有时难以获得且仅有的辐照度、光伏发电功率属于多源数据、毛刺多、具有不同特征的问题,提出了基于数据预处理与特征表示的多核支持向量机光伏预测方法,利用小波阈值分析法对光伏发电功率与辐照度进行去噪处理,通过对辐照度特征表示参数提取,进行SOM与k means结合的二次聚类选取相似日,并采用多核支持向量机进行预测。结果表明,小波阈值去噪后能大幅提高预测精度,仅用辐照度与功率数据进行预测也能取得较高的预测精度,且多核预测精度高于单核预测精度。
英文摘要:
      Considering the temperature, rainfall and other meteorological data are unavailable sometimes, only the irradiance and photovoltaic power, which are multi source data with different characteristics, can be obtained. This paper put forward the multi kernel support vector machine to predict PV generation based on data preprocessing and characteristic representation with only irradiance and photovoltaic power. Wavelet threshold analysis was adopted to de noise the photovoltaic power and irradiance signal. Characteristic parameters of irradiance were extracted, and were considered as the input of the SOM and k means clustering to select similar days for prediction. The prediction results show that the wavelet threshold de noising method can improve the prediction accuracy, and it can achieve a satisfied accuracy with only the irradiance and photovoltaic power. The prediction accuracy of multi kernel is higher than single kernel.
查看全文  查看/发表评论  下载PDF阅读器