The oil-water two-phase relative permeability curve has important influence on water cut and the oil production decline, which is the basic data of reservoir development.Correcting and predicting the endpoint value of relative permeability curve based on reservoir physical property is one of the keys to improve the accuracy of reservoir numerical simulation.In order to accurately predict the end value of the relative permeability curve,the relationship between the reservoir physical property and the end value of the relative permeability curve was established which is important to improve the reservoir numerical simulation accuracy.Porosity and permeability of the 149 relative permeability curves are taken as the input variables,while the endpoint values of relative permeability curves (irreducible water saturation, residual oil saturation and residual oil saturation of water phase permeability) as the output variables.On the above basis,a BP neural network prediction model is established to predict the end point of relative permeability curve upon reservoir physical property.The accuracy of the model is tested with 14 other relative permeability curve data.The results show that the absolute error of the model is within 0.03 and the relative error is within 9%,which meets the calculation requirements of the relative permeability curve.This method is used to predict the relative permeability curves of BZ reservoir under different physical properties,and the results can reflect the seepage characteristics of actual oil fieldand provide reference for the prediction of relative permeability and improvement of numerical simulation accuracy.