A Prediction of Clastic Reservoir Facies Using Hidden Markov Model Combined With Viterbi Algorithm

Abstract

A facies of clastic rock affects reservoir properties such as permeability and porosity. A facies analysis in well is crucial to do a conditional facies modeling. Since a recovery of core cannot be performed through total depth due to the cost, the facies analysis is preferred to link well log data to stochastics methods. Hidden Markov model (HMM) method, which predicts the facies from sedimentary -transition information and well log data, has been used for reservoir characterization. The conventional method is based on maximum a posteriori (MAP)

selects a facies based on only maximum probability. This method performs poor with very thin layers where probabilities are similar or input data are noisy. In this study, Viterbi algorithm which can decrease uncertainty of decision and remove the meaningless thin layer, is included in HMM. The proposed method is applied to the three clastic reservoir including two synthetic reservoirs and a shoreface reservoir. For C field in Louisiana, USA, it performed 12% better than conventional method and the transition-consistency ratio was increased by 70% for 18 wells.