In　this　paper,　a　non-parametric　method　based　on　Data　Envelopment　Analysis　(DEA)　is　proposed　to　measure　the　energy　efficiency.　Majority　of　the　previous　studies　did　not　consider　the　pollution　when　measuring　energy　efficiency,　and　hence　the　results　are　partially　distorted　on　the　perspective　of　environment.　However,　this　paper　takes　into　account　the　undesirable　factors　such　as　waste　water,　waste　gas　and　solid　wastes　under　the　DEA　technology　that　exhibits　Constant　Returns　to　Scale　(CRS)　and　Variable　Returns　to　Scale　(VRS)　respectively.　The　result　indicates　that　the　DEA　model　with　the　undesirable　output　as　byproduct　is　more　appropriate　in　energy　efficiency　measuring.　Moreover,　comparing　to　DEA　model　based　on　CRS,　the　model　supported　by　VRS　is　more　preferable,　since　in　real　cases,　production　process　of　VRS　is　more　likely　to　be　observed.　Finally,　an　empirical　study　on　measuring　the　energy　efficiency　of　30　provinces　in　China　is　presented.　©2009　IEEE.