Stochastic inversion of AVO based on an improved MCMC method
Li Wang, Guangzhi Zhang, Xingyao Yin · 2018
Markov Chains Monte Carlo is an important type of posterior probability sampling method in the probability inversion. However, the search process of algorithm is time consuming and often comes to the local optimal solution, which limited its application in the inverse problem of non-unique solutions. In view of the above limitation, this paper improves the traditional Metropolis-Hastings algorithm and proposes the GAMH algorithm based on the genetic crossover operation. The improved method is able to exchange information among multiple chains according to the search results, so as to search as many solutions as possible to reduce the possibility of global random walk into local solution and then improve the accuracy of inversion. This paper points out that in the process of managing model data, using GAMH-MCMC method in the AVO inversion is capable to obtain more accurate results than using the traditional method. Furthermore, GAMH-MCMC method can improve computational efficiency to some extent. Presentation Date: Wednesday, October 17, 2018 Start Time: 1:50:00 PM Location: 206A (Anaheim Convention Center) Presentation Type: Oral