Accurate determination of crop evapotranspiration (ET) in the greenhouse is critical for making scientific irrigation decisions and improving water productivity. In this study, the dynamics of grapevine latent heat (LE), ET and its components were monitored from 2017 to 2019, 2021 by using sap flow + micro-lysimeters and environmental monitoring system. The Penman-Monteith dual crop coefficient (PM), Shuttleworth-Wallace (SW) model and modified SW model (SWm) which was proposed considering the effect of the surface resistance difference between wet and dry soil under drip irrigation, were applied to simulate evaporation (E), transpiration (T) and ET. The results showed that SW and SWm models overestimated T (overestimation rate was 3-11%), while the PM model significantly underestimated T (underestimation rate was 9-18%). The simulation accuracy of three models for E was in the order of SWm>SW>PM. Compared with the SW model, SWm model could significantly reduce the simulation error of E and improved the simulation accuracy (R2 is increased from 0.75-0.81 to 0.92-0.94), while PM model had a poor simulation effect on E (R2 is 0.74-0.89, RMSE is 0.16-0.23mm d-1). ET was underestimated by PM model (underestimation rate was 4-38%) but overestimated by SW and SWm models (overestimation rate was 3-33%). All three models had high simulation accuracy for grapevine ET (R2 >0.97, RMSE is 0.25-0.50 mm d-1), and simulation accuracy of ET was in the order of SWm>SW>PM. Therefore, the SWm model is recommended to simulate ET and its components of greenhouse grapevines in the cold region of northeast China.