基于单向表示字典学习的电能质量扰动识别方法于华楠，于宏昊(东北电力大学电气工程学院，吉林吉林132012)摘要:电能质量扰动识别是电能质量数据分析问题中极其重要的一个部分。目前已经实现的电能质量扰动识别方法普遍存在识别速度较慢，识别准确率仍有较大提升空间等问题。文章提出一种计算简单但能有效识别分类的方法，即基于单向表示字典学习的电能质量扰动识别方法。对电能质量数据的训练样本进行训练得到与各个类别对应的子字典，提出单向约束以使样本在字典中的表示系数方向可以区分;通过计算测试样本的表示系数方向以及大小来区分所属类别。实验结果表明，所提方法不但识别准确度高于已有的识别方法，而且计算效率也有较大提升。关键词:电能质量;扰动识别;单向表示;字典学习DOI:10．19753/j．issn1001-1390．2023．04．019中图分类号:TM711文献标识码:B文章编号:1001-1390(2023)04-0133-06PowerqualitydisturbancerecognitionmethodbasedonunidirectionalrepresentationdictionarylearningYuHuanan，YuHonghao(SchoolofElectricalEngineering，NortheastElectricPowerUniversity，Jilin132012，Jilin，China)Abstract:Powerqualitydisturbancerecognitionisanessentialpartofpowerqualitydataanalysisproblems．Thecurrentlyimplementedpowerqualitydisturbancerecognitionmethodsgenerallysufferfromslowrecognitionspeedandlowrecognitionaccuracy，andthereisstillroomforimprovementintheaccuracyofidentification．Thispaperproposesamethodthatissim-pletocalculateandcaneffectivelyrecognizeclassification，thatis，apowerqualitydisturbancerecognitionmethodbasedonunidirectionalrepresentationdictionarylearning．Thetrainingsamplesofthepowerqualitydataistrainedtoobtainsub-dic-tionariescorrespondingtoeachtype，aunidirectionalconstraintisproposedsothatthedirectionofthecoefficientsofthesamplesinthedictionarycanbedistinguished．Thetypeisdistinguishedbycalculatingthedirectionandsizeoftherepre-sentationcoefficientofthetestsample．Theexperimentalresultsshowthatthemethodproposedinthispapernotonlyhashigherrecognitionaccuracythanexistingrecognitionmethods，butalsoimprovesthecalculationefficiency．Keywords:powerquality，disturbancerecognition，unidirectionalrepresentation，dictionary...
