Bayesian facies identification based on Markov-chain prior model
Wang Fangfang1,2, Li Jingye1,2, Chen Xiaohong1,2
1. State Key Laboratory of Petroleum Resources and Prospecting, China University of Petroleum (Beijing), Beijing 102249, China;
2. CNPC Key Lab of Geophysical Exploration, China University of Petroleum (Beijing), Beijing 102249, China
Abstract：According to Bayesian classification theory, we propose in this paper a workflow for facies identification based on Markov-chain prior model. First we choose the key well logs and define different facies based on the log data and core data. Then we extend log data as training data through rock physics modeling, synthetize seismic attributes, and estimate facies-dependent conditional probability density function. After that, we upscale seismic attributes by Backus average model. Finally we obtain posterior probability density functions by Bayesian classification on the basis of a Markov-chain prior model, and obtain the solution of the maximum posterior probability. We have tested this method on model data and marine seismic data with entropy and Bayesian confusion matrix. Test results prove that this method provides very good facies prediction.
