Well drilling leakage pressure prediction method based on SVR algorithm

Abstract

The invention belongs to the field of drilling leakage prediction and the field of artificial intelligence machine learning, and particularly relates to a drilling leakage pressure prediction method based on an SVR algorithm. The method comprises the steps of collecting data from a drilling well history database for preprocessing, creating a sample set and a test set, training an SVR model, introducing a relaxation factor, introducing a Lagrange function for unconstrained and dual derivation, selecting optimal parameters by using an SMO algorithm, and constructing a regression hyperplane to obtain a prediction function. And inputting the preprocessed data set into the trained SVR model to obtain a predicted value of the drilling leakage pressure. The method solves the problem that the current leakage pressure data has great significance for leakage stoppage and leakage prevention but cannot be directly obtained, has the advantages of reliable result, lower cost, convenience in operation and high generalizability, and has certain guiding significance and reference value in the field of leakage stoppage and leakage prevention.

Description

technical field [0001] The invention belongs to the field of drilling leakage prediction and artificial intelligence machine learning, and in particular relates to a method for predicting drilling leakage pressure based on an SVR algorithm. Background technique [0002] Loss of circulation is a common problem in major oil and gas production areas at home and abroad, and the complex downhole accidents caused by it have caused great harm to drilling and completion projects and caused serious economic losses. At present, the technology of leakage prevention and plugging at home and abroad is mainly based on drilling, bridge plugging, and cement solidification. One of the necessary prerequisites. At present, the judgment of lost circulation and the decision of lost circulation prevention and plugging are highly dependent on personal experience, and the technology has poor reproducibility, which cannot effectively meet the needs of on-site lost circulation control. As one of th...

AI Technical Summary

Problems solved by technology

[0004] One method of predicting lost circulation based on artificial intelligence and machine learning is to use the SVR algorithm. The traditional regression method has high requirements for data noise, poor generalization performance, and it is difficult to fit complex probability distribution models.

Images