Nonlinear multivariate statistical regression logging curve prediction method based on quantity protocol

Abstract

The invention discloses a nonlinear multivariate statistical regression logging curve prediction method based on a quantity protocol, and the method increases the dimensionality of attributes througha self-adaptive attribute mapping function, thereby excavating the correlation between a modeling curve and a target curve, indirectly achieving the nonlinear mapping of a model, and improving the expression capability of the model. According to the method, the singular value constraint of the long matrix is utilized to compress the data volume without influencing the statistical regression trendbetween the data, so that the operand of model optimization is greatly reduced, and the optimization algorithm can be selected more freely. The stability of a model structure equation set is improvedby gradually improving the condition number of a coefficient matrix, the iterative variation of the model is inversely calculated by gradually utilizing the loss residual error, the robustness of thealgorithm and the generalization ability of the model are ensured, and in the whole model optimization calculation process, over-fitting and over-fitting do not occur when the model gradually approaches the optimal value. The invention aims to solve the technical problem of low accuracy caused by over-fitting or under-fitting of logging curve prediction in the prior art.

Description

technical field [0001] The invention relates to the technical field of oil and gas exploration, in particular to a non-linear multiple statistical regression logging curve prediction method based on quantity reduction. Background technique [0002] Well logging information is the most important data for oil and gas reservoir research. Well logging curves can provide continuous information on the rock physical response of the entire well section, so well logging has become an important means of oil and gas reservoir research and one of the most important data to characterize reservoirs. In actual work, due to reasons such as borehole collapse, instrument failure, acquisition difficulty and cost, the measured logging curves will be distorted or missing. For example, the shear wave curve is one of the logging curves that often appear to be distorted or missing. It is difficult to re-measure the data. In actual work, people will use various methods to artificially generate well...

AI Technical Summary

Problems solved by technology

[0007] The main purpose of the present invention is to provide a non-linear multivariate statistical regression logging curve prediction method based on quantity reduction, aiming to solve the problem of over-fitting or under-fitting in the logging curve prediction in the prior art, resulting in low accuracy technical issues

Images