APPLICATION OF ARTIFICIAL NEURAL NETWORKS IN RESERVOIR PARAMETER PREDICTION IN XIFENG OILFIELD
Fengjie Li, Wang Duo-yun, Yuan Ke-zeng, Zheng Xi-min · Tianranqi diqiu kexue · 2004
To compare with the traditional log data comprehensive explanation and information disposing technique, BP artificial neural network method possesses great ability of adapting itself, learning itself in parameter prediction of reservoirs which have great nonhomogeneity and big progression error of physical parameters. By use of the wonderful nonlinear mapping ability, BP artificial neural network method can build accurately nonlinear model between reservoir parameters and log response. On the basis of expounding essential principle of BP artificial neural network, the prediction of reservoir physical parameters (porosity and permeability) is carried out the method in Yanchang formation and Yan'an formation of Xifeng oil field, and the application result is staisfied. The permeability parameters predictive accuracy is high if the progression error (10~(2)) of permeability is little. When the change range of permeability is big (10~(3)), the predictive accuracy of high permeability reservoir is high, and the one of low permeability reservoir has big relative error.