Hierarchical Bayesian Seismic Probabilistic Inversion and Lithofacies Identification using PSO-MCMC Global Sampling Algorithm

Seismic probabilistic inversion predicts the optimal solution and inversion uncertainty of model parameters simultaneously and plays an important role in reservoir characterization. Combining the particle swarm optimization with the Metropolis-Hastings sampling algorithm, we propose one novel seismic probabilistic inversion algorithm in Bayesian framework based on particle swarm optimization Markov Chain Monte Carlo (PSO-MCMC) algorithm, which has the global optimization property of particle swarm optimization and the uncertainty analysis capability of Monte Carlo model. With Bayesian formula, we derive the acceptance probabilities of the candidate model parameters and the equivalent objective function. By introducing the idea of PSO algorithm into the Metropolis-Hastings algorithm, we improve the generation process of candidate states in the Metropolis-Hastings algorithm, and propose probabilistic seismic inversion and the identification algorithm of lithic facies based on PSO-MCMC algorithm. By performing model tests, it is verified that the method has improved noise immunity, accuracy and convergence efficiency compared to the conventional Metropolis-Hastings sampling algorithm. We apply the method to actual seismic data from an eastern prospect and demonstrated its utility in the identification of lithic facies.