Oil reservoir automatic history fitting system and method based on transfer learning

Abstract

The invention relates to an automatic reservoir history fitting system based on transfer learning. The automatic reservoir history fitting system comprises a data reading module, a population reinitialization module, an optimization module, an analog calculation module, a comparative judgment module and an output module. The data reading module reads an optimization result of an existing oil reservoir, outputs the optimization result to the population reinitialization module, calculates to obtain an initial population of a new oil reservoir, outputs the initial population to the optimization module, outputs an optimized result to the simulation calculation module, obtains oil reservoir production simulation data, and outputs the oil reservoir production simulation data to the comparison and judgment module; when the error between the simulation data and the observation data meets the requirement, an optimization result is output to an output module, and system operation is completed; if the error does not meet the requirement, optimization is performed again. According to the method, the experience of historical fitting model adjustment in an old instance oil reservoir model can be used for reference, the initial population closer to the optimization result is constructed according to the fitting experience of the existing model, and the method can be integrated with any evolutionary optimization algorithm and is more suitable for being applied to practical engineering problems.

Description

technical field [0001] The invention belongs to the field of petroleum engineering, and in particular relates to an automatic history fitting system and method for oil reservoirs based on migration learning. Background technique [0002] In reservoir numerical simulation, in order to make the reservoir model match the real situation of the reservoir as accurately as possible, the history matching method is usually used to correct the model by using the reservoir data; During the development process, the main dynamic indicators such as pressure, gas-oil ratio, production, water cut, etc., if the calculated results are quite different from the real situation, the static parameters of the reservoir will be continuously modified until the measured dynamic indicators and the calculated results reach the allowable error range. Applying this model for dynamic prediction at this time is considered to have fairly accurate results. The traditional history fitting method requires cumb...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the defects of the prior art, the technical problem to be solved by the present invention is to provide an automatic reservoir history fitting system and method based on migration learning, adopting a reinitialization strategy based on migration learning with directionality and random changes, and migrating old optimized The experience of adjusting the history fitting model in the instance, solving the dynamic history fitting optimization problem of the new optimization instance

Images