Dissolved　gas　analysis　in　oil　is　one　of　the　most　common　methods　for　monitoring　the　operating　state　of　power　transformers.　The　existing　standard　set　the　early-warning　threshold　for　volume　fraction　of　dissolved　gas　in　oil　as　a　fixed　and　too　large　value.　In　response　to　this　problem,　a　dynamic　early-warning　method　based　on　data　statistics　and　distribution　model　is　proposed　in　this　paper.　Firstly,　nearly　100,000　volume　fraction　data　of　dissolved　gas　in　oil　for　220　kV　transformer　in　a　certain　power　grid　is　statistically　and　classified,　and　the　volume　fraction　distribution　model　of　dissolved　gas　in　oil　is　established.　Then,　based　on　the　actual　volume　fraction　early-warning　data　of　dissolved　gas　in　oil,　the　dynamic　early-warning　value　is　calculated　by　the　inverse　accumulation　operation　of　distribution　model.　Thus,　a　dynamic　early-warning　method　for　the　volume　fraction　of　dissolved　gas　in　oil　has　been　realized.　Finally,　the　dynamic　early-warning　method　is　compared　with　the　existing　standard　and　the　calculation　results　of　the　10%　partition　principle　and　the　dynamic　characteristics　of　early-warning　threshold　is　discussed.　The　results　show　that　this　method　can　reduce　missed　alarms　compared　with　the　existing　standard,　also　can　reduce　false　alarms　compared　with　the　10%　partition　principle.　In　other　words,　a　good　balance　between　missed　alarm　and　false　alarm　has　been　achieved,　which　has　a　certain　value　of　engineering　application.　©　2019,　Xi＇an　High　Voltage　Apparatus　Research　Institute　Co.,　Ltd.　All　right　reserved.