Research And Application Of Neural Network In Logging Information Interpretation

The core of logging interpretation is to determine the relationship between logging information and geological information,that is,to use the local data obtained from logging to determine the lithology,reservoir fluid properties,and reservoir parameters(porosity,permeability)of the entire underground reservoir.And saturation,etc.)Traditionally,relying on mathematical formulas established by geologists through past experience,local data have been used to make predictions.However,in the face of complex geology with unknown underground conditions,these methods have great limitations.The appearance of artificial neural network technology,especially new neural network models such as deep learning,opens up new ideas for solving the problems of expression and mapping of complex nonlinear relationships.It is important theoretical and practical value to study the new technology of artificial intelligence and use it to solve the decision-making problems in the oil and gas field.Based on an in-depth analysis of the development of logging interpretation-related technologies and the application status of computer technology in well logging interpretation,the paper focuses on the application of artificial neural networks in well logging interpretation.The purpose is to find better use.Logging data A method for predicting geological attribute parameters.The paper first studies and analyzes the characteristics and deficiencies of several typical shallow neural network models used for log interpretation,and designs a combined neural network model that consists of three shallow neural networks that are organically combined.The characteristics of the model are: Strong expressiveness and adaptability,good scalability.The deep neural network model was studied,and based on the actual conditions of well logging interpretation,a solution for porosity prediction using deep confidence neural network(DBN)was presented.Aiming at the difficult problem of low accuracy in the interpretation of neural networks in low-permeability and tight special oil and gas reservoirs,the idea of pre-processing pre-processing modules in artificial neural network models was proposed by referring to the ideas of log data pre-processing introduced in the manual log interpretation process.Strategies to increase binding information input.This improved neural network model can effectively improve the prediction accuracy in complex strata.Using the actual data,the paper designs and improves the artificial neural network model respectively.Through the analysis of the experimental results,it can be concluded that the use of neural network to realize well logging interpretation is a kind of recognition of formation lithology and prediction of reservoir parameters.The effective method is an important technology update for traditional well logging interpretation and has a good application prospect in the interpretation of well logging information.