Research On Seismic Inversion Of Key Parameters During Reservoir Development

Reservoir parameter prediction is an important part of seismic reservoir characterization during development.At present,the prediction is mainly based on the statistical relationship or using artificial intelligence to build seismic attributes model,while the physical meaning is unclear,and it depends on the accuracy of prior data.Besides,rock physics modeling can more accurately simulate the underground rock physics.Combined with the research status and actual needs,this paper proposes a method for predicting seismic reservoir parameters based on rock-physics inversion according to Bayesian theory framework.In this paper,the reservoir prediction is based on the rock physical model.According to the equivalent medium theory and the wave propagation theory in the pore medium,the Xu-White model establishes a rock physical model that can describe the rock elasticity and physical properties of the underground medium.The forward modeling and AVO reflection characteristics analysis of the rock physics model are conducted to discuss the parameter sensitivity and inversion multiple solutions.The feasibility of parameter inversion is verified.According to the model test results,the rock physical model inversion system based on the least square method has noise amplification effect.The rock physics inversion method based on trust region constraint reduces the inversion error.This spatial structure constraint method also has the error transmission limitation in theory.Therefore,the Bayesian theory is introduced into the inversion process.The method is by adding prior distribution constraints to the rock physics inversion objective function.A rock physics inversion method combining low-frequency logging constraints and spatial difference constraints is proposed.Experiments show that the inversion method based on the rock physics model under the Bayesian framework can overcome the influence of data error propagation.Besides,it reduces the inversion multi-solution and suppress outliers,and enhance inversion stability.Therefore,such methods are worth popularizing in reservoir prediction during development.