Abstract: The neural network has been widely used in reservoir parameter calculation with logging data and has got good effects. But under the control of logging data from multiple wells, it is crucial to construct a unified mathematical model, which helps to promote the contrast of calculating outputs among all the wells and in turn to promote the accuracy of interwell prediction. Based on the edition, standardization, normalization and depth correction of logging data, this research constructed a unified BP neural network model for each sand group. Compared with real samples, the predicting effect was good.