In order to predict the sticking temperature of high water-cut crude oil in unheated gathering, an indoor simulation tank device was used to measure sticking temperatures under various conditions. A BP neural network model was then proposed, with hyperparameters optimized by grid search and cross-validation, and a Python GUI interface was developed for real-time prediction. The results show that the model achieves an RMSE of 0.62 and an R2 of 0.88. The sticking temperature is positively correlated with wax appearance temperature and negatively correlated with water cut and density, with wax appearance temperature being the dominant factor. It is concluded that the BP neural network model effectively captures the complex nonlinear relationship, shows good generalization and engineering practicality, and provides a new technical means for setting the lower limit of safe operating temperature in high water-cut oilfields.