As　an　important　hub　equipment　of　power　system,　the　safe　and　stable　operation　of　transformer　is　the　top　priority　to　ensure　the　continuous　supply　of　high-quality　electric　energy　and　the　normal　operation　of　social　life.　The　state　estimation　of　the　transformer　is　the　key　to　the　operation　state　maintenance　method.　The　existing　transformer　state　estimation　methods　mainly　use　gas　content　and　other　data,　but　can　not　use　the　massive　transformer　electrical　quantity　monitoring　data　accumulated　in　the　monitoring　system.　Therefore,　a　k-means　clustering　method　for　transformer　state　anomaly　detection　based　on　voltage,　current　and　power　data　of　transformer　is　proposed.　Firstly,　based　on　the　monitoring　data　of　transformer　with　normal　maintenance　history,　a　state　detection　model　based　on　K-means　clustering　is　constructed.　Then,　according　to　the　clustering　results　of　historical　normal　data,　the　appropriate　threshold　is　selected,　and　the　distance　between　the　new　data　and　each　cluster　center　is　analyzed　to　judge　the　operation　status　of　the　transformer.　Finally,　the　correctness　of　the　model　is　verified　by　an　example.　The　results　show　that　the　proposed　method　can　make　full　use　of　the　electrical　data　of　the　transformer　and　realize　the　real-time　detection　of　the　transformer　state,　which　is　convenient　for　engineering　application.　©　2021　ACM.