Based on the vortex correlation system and evapotranspiration observation data in the Liulin watershed of Hebei Province, the correction effect of the Bowen ratio method on energy closure was evaluated. Particle swarm optimization algorithm and machine learning model were used to optimize the key parameters in the Penman-Monteith-Leuning(PML) model, and the long term evaporation of the study area was simulated and its spatiotemporal variation was analyzed. The results showed that after energy closure correction, the determination coefficient (R2) of the fitting results between the observed values of the vortex related system and the lysimeters in the study area increased by 0.04. The use of machine learning models to optimize the dynamic soil evaporation coefficient improved the simulation performance of the PML model. The simulation results showed an R2 of 0.79 and a Nash efficiency coefficient of 0.76, with a relative error of -5.06% compared to the evapotranspiration simulated based on the watershed water balance method. The average annual evapotranspiration in the Liulin Watershed from 2001 to 2022 was 477.8 mm, with evapotranspiration concentrated from May to August. It had a spatial distribution characteristic of being higher in the northwest and lower in the southeast. Forest land had the highest average annual evapotranspiration, followed by grassland and farmland, and building land had the smallest.