Permeability estimation in nuclear magnetic resonance measurements by artificial intelligence
1.Institute of Porous Fluid Mechanics, Chinese Academy of Sciences, Langfang, Hebei 065007, China; 2.Langfang Branch, Petro China Research Institute of Petroleum Exploration ＆ Development, Langfang, Hebei 065007, China)
Abstract：To solve the problem that it is difficult to simulate the complex relationship between nuclear magnetic resonance (NMR) measurements and permeability in conventional estimation method, a new method was proposed based on artificial intelligence algorithm. Neural network was adopted to establish the relationship between permeability and NMR measurements. Genetic algorithm was used to select the optimal parameters and initial value for the neural network. Based on information gain principle, the neural network input parameters were optimized from corresponding data. The new technique was validated by the experiments of rock sample from Songliao basin. The results show that two major neural network problems of local minima and parameter selection depending on experience can be solved by the new method. The prediction average relative error can be markedly reduced from 207.3% of conventional method to 68.6% of proposed method.
2011, Vol. 32 
