Neural Network In The Research And Application Of Seismic Inversion

Seismic inversion is the basic issue of seismic exploration, its purpose is to extract informations contained by the underground strata seismic velocity structure and the physical parameter spatial distribution of geological structure, to provide the important basis for the resource exploration and development. Since most of the geophysical inversions belong to the target function optimization multi-optima, linear inversion method can lead to the inversion results in local optimum. This paper proposes the neural network training algorithm (PSO_BP) based on the particle swarm optimization, that is applied to nonlinear seismic inversion calculation. In this way, not only has the BP neural network search performance been improved, but also the seismic inversion of operation efficiency and inversion precision improve. Moreover, for seismic wave impedance inversion, it provides a new efficient method.This paper containing the research achievements and work are as follows:(1)Around the existing problems of seismic inversion technology, this paper discusses application status and technology principle in the neural network and the seismic inversion.In research BP algorithm application, learning mode and the improved method, this paper bases on the problems, such as in BP neural network in seismic inversion prediction, the network structure design being not nimble, weights to gain in single, easy to fall into the local rules of the most superior limitations, it proposes ways to the particle swarm optimization algorithm to optimize neural network connection weights and threshold. It also designs the PSO_BP algorithm of the two algorithms mutual infiltration and reciprocal.(2)This paper in-depth researches and summarizes the principle of PSO_BP algorithm, basic structure and realization model. Through the standard test function experiment, it further discusses the operation parameters value range, feasibility and stability of the particle swarm optimization algorithm, especially engages in a more detailed research in the convergence of PSO algorithm, and summarizes some guiding rules.(3)In seismic inversion simulation experiments, this paper compares the PSO_BP algorithm with the BP algorithm. It makes use of the optimization to the oilfield actual seismic data, to hold the logging data and seismic wave impedance inversion. The test confirms the PSO_BP algorithm has fast convergence rate, less error rate, strong global optimization ability, in seismic inversion prediction applications it can get better effect.At the end, this paper summarizes the PSO_BP algorithm, and prospects the further research of the seismic wave impedance inversion algorithm.