Study Of Constraint Methodology In Constrained Impedance Inversion

The poststack seismic impedance inversion is an important technology inthe reservoir prediction and discrimination, at present; there are manyimpedance inversion methods, by studying the advantages and disadvantages,applicability, the affecting factors to inversion result and the controls parameterin the inversion process of different inversion algorithms, we can provideguidanceandavoidblindnessinthepracticalimplement.This paper discusses the nonlinear inversion algorithms instructively andmainlystudytwokinds ofnonlinearinversionalgorithms: SimulatedAnnealingConjugate Grads (SA+CG) and Artificial Neural Network Genetic Algorithms(ANN+GA), these two methods can decrease the iteration times and avoiddroppinginto the local extremum, we do the model calculation and confirm thevalidityofthealgorithm.The constrained condition's establishment is also studied in the paper, andadvances extract the reflectivity coefficient data cube by using deconvolutionmethodandaddsittotheinversionsystemasaconstraintcondition,atthesametime, it improves the deconvolution methods by import preconditionalconjugate gradient deconvolution, and get a good result. It possesses fastcalculation velocity and high precision, the addition of constraint reflectivitycubeimprovetheinversionprecision.The paper discusses the principles of matching and calibration of seismicand logging data, wavelet extraction, initial model establishment, andimpedance inversion comprehensively, which are the fundament of jointinversion of cross-well and surface seismic data. This joint inversion uses theimpedance inversion based on modeling method. Firstly, we integrate the welllogging data to inverse the impedance of surface seismic data and cross-wellseismic data separately, and then extract pseudo wells from the impedanceinversionresultsofcross-welldata.Secondly,wecombinethepseudowellsand virtual wells together to establish the 2D or 3D geologic structure model, andthen inverse the lower band impedance and relative impedance separately.Finally, we merge the two impedances above together and generate the ultimatehighresolutionimpedancemodelwhichistheresultofthejointinversion.