基于支持向量机的光伏电站异常数据序列提取  

Extraction of abnormal data sequence of photovoltaic power station based on support vector machine

作　　者：郭泽阳 程拥军 温家铭 卫东 顾鑫磊 GUO Zeyang;CHENG Yongjun;WEN Jiaming;WEI Dong;GU Xinlei(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China;State Grid Quzhou Power Supply Co.,Ltd.,Quzhou 324000,China)

机构地区：[1]中国计量大学机电工程学院,浙江杭州310018 [2]国网衢州供电公司,浙江衢州324000

出　　处：《中国测试》2023年第7期156-161,共6页China Measurement & Test

基　　金：浙江省自然科学基金(LGG22E070003)。

摘　　要：根据华为生产管理系统提供的光伏阵列直流端数据,研究电气数据序列的异常特征,提出一种基于支持向量机的异常数据序列提取方法,实现组串异常判定。通过分析异常数据序列波形特征,总结其波形变化规律;对数据序列进行偏差率及马氏距离计算,设定健康阈值,实现正常、异常标记,建立训练样本集以训练支持向量机模型,利用网格搜索与交叉验证法确定模型最优参数。将该模型对电站进行异常数据序列提取,常规/非常规类异常状态判定。结果表明,所构建的模型对电流、电压数据序列进行分类时的误判率分别为3.19%、2.03%,漏判率分别为2.35%、2.16%,该模型具有较高的可靠性。According to the DC terminal data of photovoltaic array provided by Huawei production management system,the abnormal characteristics of electrical data sequence are studied,and an abnormal data sequence extraction method based on SVM is proposed to realize the judgment of string abnormality.By analyzing the waveform characteristics of abnormal data sequence,the waveform variation law is summarized.The deviation rate and Mahalanobis distance of the data sequence are calculated,the threshold is set,the normal and abnormal markers are realized,the training sample set is established to train the SVM model,and the optimal parameters of the model are determined by grid search and cross validation method.The model is used to extract the abnormal data sequence of the power station and determine the conventional/unconventional abnormal state.The results show that the misjudgment rates of the constructed model for classifying current and voltage data sequences are 3.19%and 2.03%respectively,and the missed judgment rates are 2.35%and 2.16%respectively.The model has high reliability.

关 键 词：光伏电站 支持向量机 网格搜索 交叉验证 

分 类 号：TM615[电气工程—电力系统及自动化] TB9[一般工业技术—计量学]