Accurate　evaluation　for　operating　state　of　power　transformer　using　big　data　and　portrait　technology　is　beneficial　to　ensure　the　safe　and　stable　operation　of　power　system.　Aiming　to　the　shortcomings　of　traditional　condition　evaluation　methods,　such　as　too　single　evaluation　dimension　and　strong　subjectivity,　a　construction　method　for　transformer　operating　state　portrait　based　on　multi-dimensional　capability　and　knowledge　graph-multilayer　perceptron　(KG-MLP)　is　proposed.　Firstly,　the　portrait　system　of　transformer　operation　state　is　constructed,　which　is　composed　of　insulation　level,　load　capacity,　anti-short　circuit　ability,　energy　efficiency　level　and　voltage　regulation　capacity.　Then,　by　combining　the　KG　and　MLP,　a　portrait　analysis　model　of　transformer　operating　state　is　built.　Finally,　according　to　the　actual　operation　data　of　1368　110kV　transformers　in　a　certain　area,　an　example　analysis　of　transformer　operation　state　portrait　is　carried　out.　Also,　the　results　are　compared　with　those　of　Random　Forest　(RF)　and　Support　Vector　Machine　(SVM)　methods.　The　results　show　that　the　proposed　method　can　accurately　construct　the　transformer　running　state　portrait　with　an　accuracy　of　96.35%,　which　is　superior　to　RF　algorithm　(accuracy　of　89%)　and　SVM　algorithm　(accuracy　of　77%),　providing　a　new　way　of　evaluation　for　power　transformer　operating　state.　©　2024　Power　System　Technology　Press.　All　rights　reserved.