The Study Of Geostatistics Inversion Based On Simulated Annealing Method

As the exploration of oil and gas become more and more extensive, exploration targethas been transferred from structural reservoirs to complicated reservoirs such as lithologyand thin bed reservoirs. Traditional seismic inversion has many solutions and its resolutionis very low because of seismic is band-limited. Seismic inversion alone can not be capableof combination of various sources of information.Stochastic inversion is based on random simulation theory in geostatistics and seismicinversion theory. It can fully utilize the high vertical resolution from logging data andlateral continuity from seismic data. Geostatistical methods and techniques are used toacquire more piror information on reservoir space. Posteriori information of accuratereservoir parameters can be obtained by seismic inversion. Stochastic inversion canachieve many stochastic simulations at one time, which provide the basis for theuncertainty analysis and risk assessment of reservoirs parameters. Statistical information isthe input parameters of Stochastic inversion, including probability density function andvariogram (covariance) of random variables. The probability density function (PDF) isused to describe the probability distribution of variable in space. The variogram reflectsgeological body’s distribution and continuity in horizontal and vertical. First of all paperintroduce the nonlinear optimization algorithm, which is represented by simulatedannealing algorithm, then establish object function in frequency domain and use the veryfast simulated annealing (VFSA) algorithm solving the equation. Through the model test,the spectral inversion can identify the thin layer, but sensitive to the noise, in the practicalapplication is not ideal. This paper derives kriging equations in detail. The interpolationresult of well data only is unsatisfactory, this is proved by model tested. By introduction ofstochastic simulation technology, the smoothing effect created by Kriging is effectivelyavoided. It brings many realizations. Each realization has error to a certain degree, butthese results can be selected by seismic inversion. VFSA and Heat Bath SimulatedAnnealing (HBSA) nonlinear algorithm is used in the process. This can jump out of localoptima, and converge to the global minimum. On the basis of above research, Stochasticinversion is implemented by model and real data. The result is stable, high resolution by comparative analysis with Constrained Sparse Spike Inversion (CSSI) result and spectruminversion result. Finally, the seismic inversion method based on multi-point geostatisticshas carried on the exploring research, it can obtain the lithology, impedance resultsconstrained by seismic data. at the same time.