Abstract: The highly complexity of petroleum hydrocracking process results in the application of artificial neural network (ANN) in this field. In this paper a BP-ANN was used to model a VGO hydrocracking unit with a capacity of 2.80 Mt/a. The effect of feed rate, inlet temperatures of reactors, and amount of quench H2 used on product distribution and outlet temperatures of reactors were well predicted by the model. The results show that the model has a higher accuracy, especially in the prediction of temperatures (less than 0.1 ℃ ) and a good ability of reproducibility and generalization ability and that the model is able to guide practical operation.