Prediction Model of Total Organic Carbon Content on Hydrocarbon Source Rocks in Coal Measures Established by BP Neural Network Based on Logging Parameters

Abstract

The total organic carbon content (TOC) is an important parameter for source rocks evaluation in coal measures. Total organic carbon content determined from logging parameters using back propagation neural network technique, which provide a new method for hydrocarbon source rock evaluation. Use the Turpan Basin Xishanyao formation as the research object. The five logs which consist of volume gamma logging (GR), acoustic logging (AC), density logging (DEN), resistivity logging (RT) and compensation neutron logging (CNL) were selected optimally based on the correlation analysis of the total organic carbon content measured data and well logging parameters as the input vector of BP neural network, and the total organic carbon content was selected as the output vector of BP neural network. Then the BP neural network model was established and applied to predict total organic carbon content for Xishanyao formation of B1 well in the Turpan Basin, with a competitive analysis of the prediction errors. The error between prediction values and measured values is small, and the majority of the relative errors are less than 8%. The results show that the BP neural network model based on logging with optimal parameters has a very strong generalization ability, and can approximate the nonlinear relationship between total organic carbon content and logging parameters of coal measure source rocks with high accuracy. Keywords—coal measures; logging parameters; hydrocarbon source rocks; BP neural network; total organic carbon content