Prior　to　putting　energy　security　plans　into　action,　it　is　crucial　to　conduct　a　scientific　and　methodical　assessment　of　energy　security.　The　energy　security　level　of　30　Chinese　provinces　from　2008　to　2019　is　measured　quantitatively　using　an　energy　security　index　system.　Secondly,　two　machine　learning　methods,　random　forest　and　KNN　model,　are　applied　to　fit　the　energy　security　level　of　each　province.　Finally,　using　the　random　forest　technique,　the　key　risk　factors　affecting　energy　security　are　mined.　Our　study　shows　that　(1)　Energy　security　levels　are　inversely　correlated　with　energy　load　centers　and　vary　widely　within　China,　albeit　the　disparity　is　closing.　(2)　Supply　security　is　the　bottom　line　of　energy　security,　while　energy　carbon　intensity　and　raw　coal　and　crude　oil　consumption　are　also　important　indicators　affecting　the　level　of　energy　security.　(3)　The　machine　learning　approach　can　more　accurately　match　and　forecast　the　level　of　energy　security.　These　findings　provide　theoretical　support　and　practical　guidance　to　optimize　energy　evolution　security　management　and　risk　notification.　©　The　Author(s),　under　exclusive　license　to　Springer　Nature　Switzerland　AG　2024.