Inversion Research of Horizontal Well Array Induction Logging Based on Neural Network

In order to improve the computational efficiency and accuracy of the nonlinear inversion of horizontal well array induction logging, the neural network is introduced into the inversion of induction logging data. Firstly, the database of horizontal well array induction logging samples is established. Then, BP, RBF and convolutional neural network prediction models are built respectively, and training and testing are carried out based on sample database. During training, log response is taken as the input of neural network, and formation parameters are taken as the output of neural network. The prediction results of different neural network models are compared and analyzed, and the prediction results are visualized by graphs. The results show that the relative error between the inversion results of the formation parameters and the true value of the formation model is lower than 5%, and the evaluation indexes and prediction accuracy of the convolutional neural network prediction model are better than those of the BP and RBF neural network prediction models.