Dual-source models demonstrate strong performance in simulating evapotranspiration (ET) for sparse vegetation, where stomatal resistance is a critical parameter. While multiple stomatal resistance models have been used in field conditions, their suitability and parameterization under greenhouse conditions remain limited. This study aimed to identify an optimal stomatal resistance model and its parameters for the cultivation of grapevine in a greenhouse, a typical sparsely spaced crop. Three modified dual-source models (SW-ST, SW-JS, and SW-CO) were developed by coupling the Shuttleworth-Wallace (SW) model with the Stannard (ST), Jarvis-Stewart (JS), and coupled surface resistance (CO) stomatal resistance models, respectively. Using measured latent heat flux (lambda ET) from a greenhouse energy flux monitoring system, the stomatal resistance parameters were calibrated and the performance of the coupled models was evaluated. Results showed that the SW-CO model outperformed the other two models at hourly timescale, with a coefficient of determination (R2)of 0.86-0.87, root mean square error (RMSE) of 4.24-5.08 W m-2, mean absolute error (MAE) of 3.19-3.81 W m-2, and index of agreement (IA) of 0.98. At daily timescale, both SW-CO and SW-JS had a high accuracy - R2 >= 0.85, RMSE <= 5.84 W m-2, MAE <= 4.97 W m-2, and IA >= 0.98. However, while SW-CO and SW-ST overestimated lambda ET respectively by 4.7-9% and 7-10%, SW-JS underestimated it by 6.4-8.8%. The SW-JS model performed best during early growth stage (R2 = 0.96-0.98, RMSE = 11.7-18.4 W m-2), whereas SW-CO excelled in both mid-growth (R2 = 0.91-0.95, RMSE = 16.2-28.1 W m-2) and late-growth (R2 = 0.92-0.96, RMSE = 17.4-22.1 W m-2) stages. Sensitivity analysis showed that SW-JS had the highest stability (sensitivity coefficient <= 0.15%), with radiation as the most driving factor. On the basis of accuracy measure, SW-CO model was recommended for simulating ET of solar greenhouse grapevine.