Geophysical model optimization method integrating machine learning principle and gradual deformation method

Abstract

The invention relates to a geophysical model optimization method integrating a machine learning principle and a gradual deformation method. According to the optimization method, machine learning and agradual deformation method are combined, a known geophysical model is deeply optimized, the interpretability and accuracy of the model are greatly improved, meanwhile, the model has good superiorityin the aspect of information integration, and certain efficiency can be improved for data interpretation and application of a physical parameter model in oil exploration.

Description

technical field [0001] The invention relates to a geophysical model optimization method that combines machine learning principles and gradual deformation methods, and belongs to model optimization in the field of geophysics. Background technique [0002] With the gradual development of oil and gas resources, accurate geophysical models are becoming more and more important for reservoir evaluation and production plan formulation. Reservoir geophysical models include structural models, sedimentary facies models and physical parameter models. The first two types of models Usually, it is a deterministic model constructed based on initial exploration information and experience, while a physical parameter model can use stochastic modeling to obtain multiple model results with equal probability. [0003] At present, for the constraint, selection and optimization process of the stochastic model, the logging of individual points is usually used to constrain the model, and the accurac...

AI Technical Summary

Problems solved by technology

[0003] At present, for the constraint, selection and optimization process of the stochastic model, the logging of individual points is usually used to constrain the model, and the accuracy of the model is judged by calculating the average value of multiple models. These optimization methods usually cannot be based on the results of the current model. Given the optimal value, choosing any random model cannot fully express the geophysical parameters of the current region

The gradual deformation method is a way to gradually optimize a group of random models by using the disturbance mechanism. Each random model is weighted and extracted in proportion and superimposed on the optimized model. Theoretically, the optimal modeling result can be obtained, but among them coefficients are difficult to obtain

Images