Quantitative Reservoir Characterization Based On Probabilistic Inversion

This dissertation addresses the problem of reservoir characterization,uncertainty evaluation,and reservoir modelling.The major research contains reservoir properties estimation,seismic facies classification,quantitative uncertainty evaluation,and geostatistical modelling.Several new strategies were proposed for the problems in the processing of reservoir characterization and modelling.Synthetic models and field data application validated the methodology.The objective functions of reservoir properties estimation by elastic data are strongly nonlinear,because of the complexity of the mathematical expressions in rock physical models.The nonlinear objective functions cause negative influences on both the accuracy and stability of the inversion results.Hence,Monte Carlo simulation was introduced in the processing of reservoir properties estimation.Multiple samples of reservoir properties were first derived stochastically.Meanwhile,an artificial intelligent algorithm was explored for sample selection.The posterior probabilities of reservoir properties were then achieved.In regard to the water saturation(Sw)estimation,the probabilistic relationship among reservoir properties,such as porosity,clay content,and Sw,was constructed,since the Sw estimation by low-quality seismic derived elastic information is inaccurate.A reliable result of Sw estimation was then achived.Field data application validated the methodology.Facies distribution in target zone serves as an important reference for reservoir characterization.Generally,the results of seismic facies classification share a great deal of uncertainty.In this dissertation,a probabilistic multistep approach was exploited for the estimation of seismic facies probabilities.The statistical relationships between input and output parameters of the multiple steps in seismic facies classification were first constructed.Then,seismic facies probabilities were derived from the statistical information.Moreover,parameter spaces were restricted according to the data distribution characteristics in the cross-plot.Parameter vectors that went beyond the restricted scopes were excluded,reducing the computational time as well as uncertainty.The concept of entropy were explored for quantitative uncertainty evaluation.To achieve this,facies probabilities conditioned on different properties in each step of seismic facies classification were first derived.Then,the associated uncertainty and maximum a posterior(MAP)of facies probabilities were evaluated by entropy and reconstruction rate,which assesses the degree of similarity between MAP and facies sequence within the range [0,1].This enables us to investigate the influence of the uncertainty propagation on seismic facies classification.The uncertainty of the inversion results was finally characterized by the calculated entropy and its indicator-transform.Uncertainty in reservoir modelling cannot be neglected when well logs are sparsely distributed in the target reservoir.Reservoir characterization between wells is relatively unreliable.For this reason,facies probability derived from seismic facies classification was introduced as constraint information for reservoir facies modelling.With the help of sequential indicator simulation,well logs and seismic facies probabilities were integrated by Tau-model for constructing the local conditional probability.The results of reservoir facies modelling conditioned with seismic facies probabilities shown more accuracies and stabilities than the original results.