A　method　of　identifying　the　operating　parameters　of　harmonic　customers　based　on　historical　monitoring　data　is　introduced　in　this　paper.　Firstly,　PCA　is　used　to　reduce　the　original　harmonic　data　dimensions　and　determine　the　appropriate　number　of　clusters.　Then　k-means　is　used　to　partition　harmonic　mode　on　low　dimensions.　Lastly　group　feature　parameters　are　calculated　from　the　clustered　typical　harmonic　operating　conditions.　Experimental　results　showed　that　the　operating　parameters　of　harmonic　customers　can　be　identified　from　a　large　number　of　high-dimensional　historical　statistic　data　by　the　proposed　method,　which　will　also　contribute　to　the　harmonic　source　location　and　optimal　operation　strategy.　©　2018　IEEE.