The　common　energy　efficiency　model　of　chiller　in　subway　station　has　nonlinear　interference,　which　leads　to　low　prediction　accuracy　and　reliability.　To　solve　the　above　problems,　this　paper　proposes　a　model　of　chiller　efficiency　prediction　based　on　error　correction.　This　model　focuses　on　the　error,　trains　the　historical　data　of　prediction　error　by　SVR,　and　then　corrects　the　prediction　result　of　traditional　model　to　get　the　final　prediction　value　of　energy　efficiency.　Through　the　verification　of　operation　data　from　air　conditioning　units　on　a　Beijing　subway　platform　,　the　results　show　that　this　model　is　superior　to　the　traditional　model　,　the　average　relative　error　of　predictive　value　decreased　from　4.04*10(-2)　to　6.69*10(-)(11).