Abstract: In order to obtain a more accurate light crude water-in-oil (W/O) emulsion viscosity model, current models with preferable computational accuracy are improved. Light crude is used to prepare the W/O emulsion for viscosity test under different temperature and shear rate. Then, the commonly used emulsion viscosity prediction models are compared to define the optimal one based on the experimental data, suggesting that the Brinkman model features the highest accuracy. Consequently, the Brinkman model is improved by considering the influences of hydration, flocculation and water content on the calculation of emulsion viscosity, and also directly incorporating the influence of temperature, regardless of shear rate. The improved model is verified by laboratory loop test. The result shows that the improved model has higher accuracy, with average relative error less than 10%.