Abstract: Lithology identification is the basis of formation recognition and reservoir parameter calculation, and the traditional lithology identification method can not meet the needs of actual production because of the complexity and heterogeneity of sedimentary environment. Aiming at the problem of traditional identification method such as the fault tolerance ability is poor, the degree of automation is low and the interpretation accuracy is low. By using the neural network autonomous learning prediction analysis method, the comparison study of several popular lithologic identification methods, a more suitable field practical method was applied to the drilling system. The study found that in the case of the same prediction method and well logging curve, the more the number of standard stratigraphic samples is, the higher the correct rate. By comparing probabilistic neural networks method in the application in the production of better effect, the recognition accuracy rate was high, training and recognition time was the shortest, a high level of recognition can be still maintained when less logging data are got.