Research On Tubing And Casing Leak Diagnosis Method Based On Multi-source Information Fusion

The downhole tubing is in a harsh environment for a long time.It is very likely to leak which is caused by corrosion and perforation,and pose a serious threat to the safety of oil and gas wells.Accurately judging whether the downhole tubing is leaking,and where is the source and location of the leakage is of great significance for the safety production of oil and gas wells.However,the single detection method is mostly used to detect the leakage now.And the collected information is imperfect,which reduces the accuracy of diagnosis.In view of the above problems,this paper adopts multi-source information fusion method for comprehensive treatment.The main research content is as follows.(1)Downhole tubing leak simulation system independently developed by the laboratory is used to carry out the tubing and casing leakage sound wave experiment,the annulus liquid level detection experiment,the annulus pressure relief-pressure recovery experiment,the component analysis experiment and Tracer detection test.And the collected experimental data provides support for the following research.(2)An improved neural network fusion method based on ant colony algorithm is used to determine whether the oil casing is leaking.Adopt wavelet threshold denoising method to denoise the acoustic wave signal and annulus pressure relief-pressure recovery signal,and extract the time domain and frequency domain eigenvalues.To solve the shortcomings of BP neural network,an ant colony algorithm is proposed.The improved neural network fusion model uses the extracted eigenvalues as input vectors and the state of the tubing as output.Experiments show that the accuracy of the fusion model recognition is 97.5%.(3)After determining the leakage of the tubing,LS-SVM is used to analyze the components of the annular air body to determine the source of the leakage.The RBF function with stability is selected,and the cross-validation method is used to find the optimal parameters to construct LS-SVM model.Experiments show that the model can determine whether the leakage is from the production or non-production layer,with an accuracy rate of 98%.(4)Aiming at the leakage from the production zone,this paper uses D-S evidence theory with the leakage positioning method of tubing and casing based on autocorrelation analysis and tracer location method to determine the location of the leakage.Firstly,establish the target recognition framework.And then calculate the basic probability assignment function of the two evidence bodies separately,which merge the basic probability assignment function of the two evidence bodies according to the Dempster fusion rule.Judging from the decision rules based on basic probability assignment,experiments show that the model can choose one with higher positioning accuracy from the results obtained by the two positioning methods.