Study On Chaotic Genetic Optimization Method For Multi-wave Pre-stack AVA Inversion

According to the relationship of changes in the amplitude characteristic pick lithological parameters from the pre-stack seismic data, it is very important to lithology identification and oil or gas detection. Because of this nonlinear relationship, people usually solve the nonlinear problem by gradient information in mathematical method, or the best model by random search in the known model space. With the development and application of the nonlinear science, more and more nonlinear optimization algorithm is introduced to the geophysical field. This paper studied in the application of multi-wave AVA inversion around genetic optimization method.The angular trace gather record is used to AVA inversion that amplitude vary with angle. So we forward simulate P-P and P-SV wave's AVA curve by Zoeppritz exact formula and its approximate formula based on the theoretical model, then forward simulate multi-wave AVA angular trace gather record.The simple genetic algorithm initializes the individual with binary code method, while AVA inversion can be seen as an optimization problem, real-coded scheme have high precision to the function optimization was used in this paper. chaotic system was introduced in the algorithm for the initial population is more uniform distributed in the solution space. As an example of Logistic model, we verified sensitive dependence on initial conditions, randomness, ergodicity characteristics of chaotic system. The use of an chaotic mapping mechanism to initialize population distributed is more evenly based on the uniformity.Genetic algorithm's population generate new individuals by crossover and mutation at a setting probability. We use dynamic adaptive crossover and mutation probability to generate new individuals in order to avoid premature convergence phenomenon in the course of evolution. AVA inversion is on the basis of Zoeppritz formula, including a mount of parameters. So we introduced multi-agent genetic algorithm that adapt to optimize high-dimensional function. they combine with chaotic system and design adaptive chaotic genetic algorithm and chaotic multi-agent genetic algorithm, then test their computing performance with five standard test functions. The computing performance of standard algorithm is compared and analyzed with the two algorithms.Both of them are used for two-layer model in the single P-P wave and P-SV wave inversion or the joint inversion. It is shown that joint inversion can improve the accuracy of AVA inversion to a certain extent. we conduct joint inversion to multi-layer model, its result show that the number of layers increase, while the parameters of simultaneous inversion also increased. The chaotic multi-agent genetic algorithm has better advantage when AVA inversion with multi-layer model. The result of AVA inversion explain that the given initial search range have an certain effect on searching accuracy through the different search range. We add some noise to theoretical model, and inversion results indicate that chaotic multi-agent genetic algorithm has ability of anti-noise, then conduct joint inversion to a sample model, finally, this algorithm is applied to real seismic data processing.