阐述了神经网络用于过程建模目前存在的问题，提出了改进的方法，并将此应用到了大型炼油厂脱蜡过程的建模中。通过比较改进前后的ＢＰ算法仿真结果可以看出，改进后的算法在网络的训练收敛速度上有了较大提高。 The existing problems of neural network for process modeling are expatiated. An improved method is put forward and applied to the modeling of dewaxing process in large refinery. By comparing the BP algorithm simulation results before and after the improvement, we can see that the improved algorithm has greatly improved the training convergence speed of the network.