CN116205119A - Method and system for evaluating corrosion risk of underground oil pipe - Google Patents

Abstract

The invention discloses a method and a system for evaluating corrosion risk of an underground oil pipe, comprising the following steps of S1, selecting a corrosion risk basic index, establishing an oil pipe risk prediction index system, and calculating a correlation coefficient between the corrosion risk basic index and a corrosion rate; s2, selecting sample data from the corrosion risk basic indexes, and processing the selected sample data to obtain an original sample; s3, taking the RBF kernel function as an SVM classification kernel function, and establishing a corrosion risk prediction SVM model by utilizing the sample obtained in the step S2; and S4, predicting the corrosion rate of the oil pipe by using the obtained corrosion risk prediction SVM model, and evaluating the corrosion risk of the oil pipe according to the predicted corrosion rate. Compared with the prior art, the method and the device have the advantages that the weight does not need to be determined, the nonlinear relation among the factors can be automatically simulated, the influence of human components in the process of weight determination in the traditional evaluation process is avoided, the corrosion rate and other information can be objectively obtained, and the accuracy is higher.

Description

Method and system for evaluating corrosion risk of underground oil pipe

Technical Field

The invention belongs to the technical field of drilling downhole corrosion risk evaluation, relates to a downhole oil pipe corrosion risk evaluation method and system, and in particular relates to a downhole oil pipe carbon dioxide and hydrogen sulfide corrosion risk evaluation method and system based on SVM algorithm.

Background

Downhole tubing corrosion is affected by a number of factors, with medium factors being the most dominant, and its principal corrosion forms are carbon dioxide corrosion and hydrogen sulfide corrosion, and corrosion is most prevalent and most complex when co-existing with carbon dioxide and hydrogen sulfide. The corrosion of the oil pipe not only affects the safe production of the oil and gas well, but also can prevent the trial and well repair operation to a certain extent. By evaluating the risk of oil pipe corrosion, the information of the high-risk corrosion part in the oil pipe is obtained, and precious experience is left for corrosion prevention of the underground oil pipe and safe production of an oil and gas well. The development blocks such as sequential north and jump belong to carbonate reservoirs, the corrosion environment is very complex, the corrosion medium has very strong heterogeneity, the corrosion environments of different fracture zones are very different, the oil pipe corrosion mechanism is also very complex, the development risk is high, and the difficulty is high.

The patent number of China patent No. CN201210548324.X is an oil pipe corrosion degree prediction method and device, an oil pipe corrosion prediction model is provided, corrosion degree prediction is carried out on an in-service oil pipe in a specified environment, and accuracy of underground oil pipe corrosion degree prediction can be improved.

The Chinese patent No. CN201511018634.0 discloses an in-service tubing string corrosion failure prediction system based on data mining, and provides an oil tube failure predictor model which can accurately predict the influence degree of the current environment on the corrosion rate of an oil tube and predict the service life of the oil tube, so that the use safety of the tubing string can be greatly improved, the occurrence of a tubing string leakage accident can be reduced, the cost of enterprise users can be saved, and the environmental pollution can be avoided.

Aiming at development blocks such as northbound, jump and the like, the key problems are that the corrosion environment is very complex, the corrosion medium has very strong heterogeneity, the corrosion environments of different fracture zones are very different, the oil pipe corrosion mechanism is also very complex, the development risk is high, and the difficulty is high. None of the above patents address the problem of predicting the risk of corrosion of tubing in such complex environments.

Therefore, the SVM-based corrosion risk evaluation method based on the statistical learning theory and through learning the accurate information near the interface between sample categories has important significance, and is a problem to be solved in the field.

Disclosure of Invention

The invention aims to provide a method and a system for evaluating the corrosion risk of an underground oil pipe, which can automatically simulate the nonlinear relation among factors without determining weights, avoid the influence of human components in the process of determining the weights in the traditional evaluation process, further objectively obtain the corrosion rate and other information, and have higher accuracy.

In order to achieve the above object, the present invention provides the following technical solutions:

an evaluation method of corrosion risk of an underground oil pipe comprises the following steps:

s1, selecting a corrosion risk basic index, establishing an oil pipe risk prediction index system, and calculating a correlation coefficient between the corrosion risk basic index and the corrosion rate;

s2, selecting sample data from the corrosion risk basic indexes, and processing the selected sample data to obtain an original sample;

s3, taking the RBF kernel function as an SVM classification kernel function, and establishing a corrosion risk prediction SVM model by utilizing the sample obtained in the step S2;

and S4, predicting the corrosion rate of the oil pipe by using the obtained corrosion risk prediction SVM model, and evaluating the corrosion risk of the oil pipe according to the predicted corrosion rate.

Preferably, the corrosion risk base indicator in step S1 includes water content, temperature, pressure, CO2/H2S partial pressure, cl-concentration, corrosion evaluation test period, stress cracking evaluation test period, and tubing material of the environment in which the tubing is located.

Preferably, the calculating method of the correlation coefficient in step S1 is as follows: and analyzing the correlation coefficient between the corrosion risk basic index and the corrosion rate by using the Pearson correlation, wherein the correlation coefficient changes from negative correlation to positive correlation, and the correlation coefficient is irrelevant when the coefficient is 0.

Further preferably, the formula of the pearson correlation is:

wherein X is a corrosion risk basic index, and Y is a corrosion rate.

Preferably, the selected sample data in step S2 is positively correlated with the corrosion rate.

Preferably, the processing in step S2 is: and carrying out normalization processing on the selected sample data, wherein after the normalization processing, the value range x of the sample data is limited between [ -1,1 ].

Further preferably, the conversion function of the normalization process is:

where x represents the original data, x represents the processed data, min represents the minimum value in the original data sample, and max represents the maximum value in the original data sample.

Preferably, the method for establishing the corrosion risk prediction SVM model in step S3 includes the following steps:

(1) Carrying out pre-dimension reduction treatment on data in an original sample, and dividing the data into a training sample and a prediction sample;

(2) The RBF kernel function is used as an SVM classification kernel function, an oil pipe corrosion risk prediction SVM model is constructed according to the training sample, and the risk category of the prediction sample is predicted according to the obtained oil pipe corrosion risk prediction SVM model, so that a penalty factor C and RBF kernel function parameters g are obtained;

(3) Optimizing the obtained penalty factor C and RBF kernel function parameter g to obtain an optimal parameter value of the penalty factor C and RBF kernel function parameter g;

(4) Optimizing the oil pipe corrosion risk prediction SVM model according to the obtained optimal parameter values of the punishment factor C and the RBF kernel function parameter g as parameter values of the oil pipe corrosion risk prediction SVM model;

(5) And (3) comparing the optimal parameter values (C, g) determined in the last step one by one to construct an oil pipe corrosion risk prediction SVM model, comparing the accuracy with historical data to verify that the SVM model with the accuracy more than or equal to 85% has higher reliability, can be used for predicting the oil pipe corrosion risk, and returns to the step (4) if the accuracy is lower than 85%.

Further preferably, the optimizing process in step (4) is: optimizing penalty factor C and RBF kernel function parameter g by a K-time cross validation method;

the method comprises the following steps: dividing a training sample set into subsets with the same size as K, taking one subset as a test set, taking the rest K-1 subsets as training sets, and checking the accuracy of a training classifier on the training set by using the test subset; the method is sequentially circulated until each subset is tested once, and training and testing are respectively carried out K times; meanwhile, setting the change range of the penalty factor C and the RBF kernel function parameter g and the step size of each change, and respectively completing K times of circulation for each group of changed (C, g); and finally, selecting (C, g) corresponding to the classifier with the highest cross-validation accuracy as the optimal parameter value.

Preferably, the evaluation of the corrosion risk of the oil pipe in step S4 is classified as:

the invention also provides an SVM-based oil pipe corrosion risk evaluation system, which comprises:

the oil pipe corrosion risk basic index system is used for classifying oil pipe corrosion risk basic indexes, so that sample data can be conveniently processed;

the sample data selecting module is used for selecting sample data from the oil pipe corrosion risk basic indexes and processing the sample data;

the risk prediction module is used for establishing an oil pipe corrosion risk prediction SVM model, predicting the oil pipe corrosion risk and obtaining an optimal parameter value of the punishment factor C and the RBF kernel function parameter;

the model optimization module is used for taking the obtained optimal parameter value as the parameter value of the oil pipe corrosion risk prediction SVM model and optimizing the oil pipe corrosion risk prediction SVM model.

The beneficial effects of the invention are as follows:

according to the oil pipe corrosion risk basic index system, the SVM classification model is used for optimizing parameters, the oil pipe corrosion risk evaluation method and system based on the SVM have high prediction accuracy for nonlinear relations and small sample models, and good prediction effects are achieved in oil pipe corrosion risk evaluation. Compared with the prior art, the method and the device have the advantages that the weight does not need to be determined, the nonlinear relation among the factors can be automatically simulated, the influence of human components in the process of determining the weight in the traditional evaluation process is avoided, the corrosion rate and other information can be objectively obtained, and the accuracy is higher.

Drawings

FIG. 1 is a graph showing actual prediction of corrosion rate according to an embodiment of the present invention.

Detailed Description

The method and system for evaluating the risk of corrosion of a downhole oil pipe according to the present invention will be described more fully hereinafter with reference to the accompanying drawings, in which it is shown, however, that the embodiments shown are only some, and not all, of the embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Before the embodiments of the invention are explained in further detail, it is to be understood that the invention is not limited in its scope to the particular embodiments described below; it is also to be understood that the terminology used in the examples of the invention is for the purpose of describing particular embodiments only, and is not intended to limit the scope of the invention.

An evaluation method of corrosion risk of an underground oil pipe comprises the following steps:

s1, selecting a corrosion risk basic index, establishing an oil pipe risk prediction index system, and calculating a correlation coefficient between the corrosion risk basic index and the corrosion rate;

s2, selecting sample data from the corrosion risk basic indexes, and processing the selected sample data to obtain an original sample;

s3, taking the RBF kernel function as an SVM classification kernel function, and establishing a corrosion risk prediction SVM model by utilizing the sample obtained in the step S2;

and S4, predicting the corrosion rate of the oil pipe by using the obtained corrosion risk prediction SVM model, and evaluating the corrosion risk of the oil pipe according to the predicted corrosion rate.

Preferably, the corrosion risk base indicator in step S1 includes water content, temperature, pressure, CO2/H2S partial pressure, cl-concentration, corrosion evaluation test period, stress cracking evaluation test period, and tubing material of the environment in which the tubing is located.

Preferably, the calculating method of the correlation coefficient in step S1 is as follows: and analyzing the correlation coefficient between the corrosion risk basic index and the corrosion rate by using the Pearson correlation, wherein the correlation coefficient changes from negative correlation to positive correlation, and the correlation coefficient is irrelevant when the coefficient is 0.

Further preferably, the formula of the pearson correlation is:

wherein X is a corrosion risk basic index, and Y is a corrosion rate.

Preferably, the selected sample data in step S2 is positively correlated with the corrosion rate.

Preferably, the processing in step S2 is: and carrying out normalization processing on the selected sample data, wherein after the normalization processing, the value range x of the sample data is limited between [ -1,1 ].

Further preferably, the conversion function of the normalization process is:

where x represents the original data, x represents the processed data, min represents the minimum value in the original data sample, and max represents the maximum value in the original data sample.

Preferably, the method for establishing the corrosion risk prediction SVM model in step S3 includes the following steps:

(1) Carrying out pre-dimension reduction treatment on data in an original sample, and dividing the data into a training sample and a prediction sample;

(2) The RBF kernel function is used as an SVM classification kernel function, an oil pipe corrosion risk prediction SVM model is constructed according to the training sample, and the risk category of the prediction sample is predicted according to the obtained oil pipe corrosion risk prediction SVM model, so that a penalty factor C and RBF kernel function parameters g are obtained;

(3) Optimizing the obtained penalty factor C and RBF kernel function parameter g to obtain an optimal parameter value of the penalty factor C and RBF kernel function parameter g;

(4) Optimizing the oil pipe corrosion risk prediction SVM model according to the obtained optimal parameter values of the punishment factor C and the RBF kernel function parameter g as parameter values of the oil pipe corrosion risk prediction SVM model;

(5) And (3) comparing the optimal parameter values (C, g) determined in the last step one by one to construct an oil pipe corrosion risk prediction SVM model, comparing the accuracy with historical data to verify that the SVM model with the accuracy more than or equal to 85% has higher reliability, can be used for predicting the oil pipe corrosion risk, and returns to the step (4) if the accuracy is lower than 85%.

Further preferably, the optimizing process in step (4) is: optimizing penalty factor C and RBF kernel function parameter g by a K-time cross validation method;

the method comprises the following steps: dividing a training sample set into subsets with the same size as K, taking one subset as a test set, taking the rest K-1 subsets as training sets, and checking the accuracy of a training classifier on the training set by using the test subset; the method is sequentially circulated until each subset is tested once, and training and testing are respectively carried out K times; meanwhile, setting the change range of the penalty factor C and the RBF kernel function parameter g and the step size of each change, and respectively completing K times of circulation for each group of changed (C, g); and finally, selecting (C, g) corresponding to the classifier with the highest cross-validation accuracy as the optimal parameter value.

Preferably, the evaluation of the corrosion risk of the oil pipe in step S4 is classified as:

the invention also provides an SVM-based oil pipe corrosion risk evaluation system, which comprises:

the oil pipe corrosion risk basic index system is used for classifying oil pipe corrosion risk basic indexes, so that sample data can be conveniently processed;

the sample data selecting module is used for selecting sample data from the oil pipe corrosion risk basic indexes and processing the sample data;

the risk prediction module is used for establishing an oil pipe corrosion risk prediction SVM model, predicting the oil pipe corrosion risk and obtaining an optimal parameter value of the punishment factor C and the RBF kernel function parameter;

the model optimization module is used for taking the obtained optimal parameter value as the parameter value of the oil pipe corrosion risk prediction SVM model and optimizing the oil pipe corrosion risk prediction SVM model.

In order to better understand the technical scheme and technical effect of the invention, the technical scheme and technical effect of the invention are specifically described below for the development blocks of north-south, jump-in and the like. The corrosion of the downhole oil pipe refers to the corrosion of the downhole oil pipe by carbon dioxide and hydrogen sulfide.

Example 1

An evaluation method of corrosion risk of an underground oil pipe comprises the following steps:

step 1: establishing an oil pipe corrosion risk prediction index system; the pearson phase correlation analysis is used for analyzing the relation between each factor and the corrosion rate, the correlation coefficient changes from negative correlation to positive correlation, and no relation is shown when the correlation coefficient is 0.

The calculation method comprises the following steps:

wherein X is a corrosion risk basic index, and Y is a corrosion rate.

The correlation coefficients of the 8 corrosion risk base indicators and the corrosion rates are shown in table 1 below.

TABLE 1

Step 2: sample data is selected and processed;

step 2.1: there may be 3 relationships between the actually selected indices and the risk: the larger the index value is, the larger the risk is;

the smaller the index value, the greater the risk; the oil pipe corrosion rate is in a safe state in a certain interval, and the more the index value deviates from the interval, the greater the corresponding risk. In order to unify the variation directions of each index and risk, namely, the larger the index value is, the larger the risk value is, forward processing is carried out on various indexes;

step 2.2: in order to eliminate the influence of the dimensional differences of different index variables on the reliability of the verification result, the data are normalized, and the value range x of each index is limited between [ -1,1 ]. The normalization method adopts a maximum and minimum method, and the conversion function is as follows:

where x represents the original data, x represents the processed data, min represents the minimum value in the original data sample, and max represents the maximum value in the original data sample.

Step 3: evaluating the corrosion risk degree of the oil pipe;

step 3.1: dividing a sample into two types, namely a training sample and a prediction sample, constructing an SVM model by using the training sample, and predicting risk categories of the prediction sample by using the constructed model, namely constructing an SVM model by using sample data (215 samples) of an indoor laboratory, and simultaneously, predicting risk categories of a northbound 1 fracture zone, a 5 fracture zone, a southward region and a southward region by using the constructed SVM model;

step 4: establishing and checking an oil pipe corrosion risk prediction SVM model;

step 4.1: before the model is constructed, carrying out pre-dimensionality reduction treatment on sample data;

step 4.2: the kernel functions are key factors for realizing mapping of the problems from the input space to the high-dimensional space in the SVM algorithm, and different support vector machine algorithms are adopted by different kernel functions, and the form and the parameters of the kernel functions determine the type and the complexity of the classifier. The functions satisfying the Mercer theorem can be used as kernel functions of the model, and it should be noted that selecting different inner-product kernel functions in the support vector machine model results in different algorithms. Because only one parameter g in the RBF kernel function is adjustable, the operation difficulty of the model is greatly reduced; meanwhile, research has shown that RBF kernel functions are superior to other functions in most cases, and have strong universality. Therefore, RBF is selected as the SVM classification kernel function.

Step 4.3: after the kernel function is determined, two important parameters in the SVM model need to be determined: penalty factor C and RBF kernel parameter g. The adopted parameter optimization method is a K-time cross validation method, and the basic thought is as follows: dividing a training sample set into subsets with the same size as K, taking one subset as a test set, combining the rest K-1 subsets as a training set, and checking the accuracy of a training classifier on the training set by comparing the classification rate with the test subset; the method comprises the steps of sequentially cycling until each subset is tested once, training a model or a hypothesis function according to training, and respectively carrying out training and testing K times; simultaneously, setting the variation range of the parameters C and g and the step size of each variation, and respectively completing K times of circulation for each group of variation (C, g); and finally, selecting (C, g) corresponding to the classifier with the highest cross-validation accuracy as the optimal parameter value.

The optimizing result shows that when the cross-validation accuracy is 100%, the SVM model parameter C value is 138, and the g value is 0.088.

Step 5: and (3) model inspection and prediction result analysis, after a kernel function and model parameters are selected, comparing the optimal parameter values (C, g) determined in the last step one by one to construct an SVM risk prediction model, and comparing the accuracy of the SVM risk prediction model with historical data to verify that the SVM model with high accuracy (the accuracy is more than or equal to 85 percent) has higher reliability, can be used for predicting the corrosion risk of the oil pipe, and returns to the step (4.3) if the accuracy is not high (less than 85 percent).

According to the selected sample data, after the kernel function and the optimization model parameters are selected, a libsvm tool box is used in MATLAB to construct an SVM risk prediction model, and the accuracy of the SVM risk prediction model is verified. The model training results show that 28 samples in 30 test samples are correctly classified, and the prediction accuracy reaches 93.333%. Therefore, the trained SVM model has higher reliability and can be applied to the evaluation of the corrosion risk of the oil pipe.

In summary, the oil pipe corrosion risk basic index system is established, the SVM classification model is used for optimizing parameters, the oil pipe corrosion risk evaluation method and system based on the SVM have higher prediction accuracy on the nonlinear relation and the small sample model, and the method and system have good prediction effect in oil pipe corrosion risk evaluation as shown in fig. 1.

Compared with the prior art, the method and the system for evaluating the corrosion risk of the underground oil pipe do not need to determine the weight, can automatically simulate the nonlinear relation among all factors, avoid the influence of human components in the process of determining the weight in the traditional evaluation process, further objectively obtain the information such as the corrosion rate and the like, and have higher accuracy.

The foregoing has outlined and described the basic principles, features, and advantages of the present invention in order that the description that follows is merely an example of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, but rather that the foregoing embodiments and description illustrate only the principles of the invention, and that the invention is susceptible to various equivalent changes and modifications without departing from the spirit and scope of the invention, all of which are intended to be within the scope of the invention as hereinafter claimed. The scope of the invention is defined by the appended claims and their equivalents.

Claims (10)

1. The method for evaluating the corrosion risk of the underground oil pipe is characterized by comprising the following steps of:

s1, selecting a corrosion risk basic index, establishing an oil pipe risk prediction index system, and calculating a correlation coefficient between the corrosion risk basic index and the corrosion rate;

s2, selecting sample data from the corrosion risk basic indexes, and processing the selected sample data to obtain an original sample;

s3, taking the RBF kernel function as an SVM classification kernel function, and establishing a corrosion risk prediction SVM model by utilizing the sample obtained in the step S2;

and S4, predicting the corrosion rate of the oil pipe by using the obtained corrosion risk prediction SVM model, and evaluating the corrosion risk of the oil pipe according to the predicted corrosion rate.

2. The method of claim 1, wherein the corrosion risk base indicator in step S1 comprises water content, temperature, pressure, CO2/H2S partial pressure, cl-concentration, corrosion evaluation test period, stress cracking evaluation test period, and tubing material of the environment in which the tubing is located.

3. The evaluation method according to claim 1, wherein the calculation method of the correlation coefficient in step S1 is as follows: and analyzing the correlation coefficient between the corrosion risk basic index and the corrosion rate by using the Pearson correlation, wherein the correlation coefficient changes from negative correlation to positive correlation, and the correlation coefficient is irrelevant when the coefficient is 0.

4. A prediction method according to claim 3, wherein the pearson correlation is formulated as:

wherein X is a corrosion risk basic index, and Y is a corrosion rate.

5. The evaluation method according to claim 1, wherein the processing in step S2 is: and carrying out normalization processing on the selected sample data, wherein after the normalization processing, the value range x of the sample data is limited between [ -1,1 ].

6. The evaluation method according to claim 5, wherein the conversion function of the normalization process is:

where x represents the original data, x represents the processed data, min represents the minimum value in the original data sample, and max represents the maximum value in the original data sample.

7. The method according to claim 1, wherein the method for creating the corrosion risk prediction SVM model in step S3 includes the steps of:

(1) Carrying out pre-dimension reduction treatment on data in an original sample, and dividing the data into a training sample and a prediction sample;

(2) The RBF kernel function is used as an SVM classification kernel function, an oil pipe corrosion risk prediction SVM model is constructed according to the training sample, and the risk category of the prediction sample is predicted according to the obtained oil pipe corrosion risk prediction SVM model, so that a penalty factor C and RBF kernel function parameters g are obtained;

(3) Optimizing the obtained penalty factor C and RBF kernel function parameter g to obtain an optimal parameter value of the penalty factor C and RBF kernel function parameter g;

(4) Optimizing the oil pipe corrosion risk prediction SVM model according to the obtained optimal parameter values of the punishment factor C and the RBF kernel function parameter g as parameter values of the oil pipe corrosion risk prediction SVM model;

(5) And (3) comparing the optimal parameter values determined in the last step one by one to construct an oil pipe corrosion risk prediction SVM model, comparing the accuracy of the oil pipe corrosion risk prediction SVM model with historical data, verifying that the oil pipe corrosion risk prediction SVM model with the accuracy of more than or equal to 85 percent has higher reliability, can be used for predicting the oil pipe corrosion risk, and returns to the step (4) if the accuracy is lower than 85 percent.

8. The evaluation method according to claim 7, wherein the optimization process is: optimizing penalty factor C and RBF kernel function parameter g by a K-time cross validation method;

the method comprises the following steps: dividing a training sample set into subsets with the same size as K, taking one subset as a test set, taking the rest K-1 subsets as training sets, and checking the accuracy of a training classifier on the training set by using the test subset; the method is sequentially circulated until each subset is tested once, and training and testing are respectively carried out K times; meanwhile, setting the change range of the penalty factor C and the RBF kernel function parameter g and the step size of each change, and respectively completing K times of circulation for each group of the changed C and g; and finally, selecting C and g corresponding to the classifier with the highest cross-validation accuracy as optimal parameter values.

9. The method of evaluation according to claim 1, wherein the evaluation of the risk of corrosion of the oil pipe in step S4 is classified as:

10. an SVM oil pipe corrosion evaluation system according to the evaluation method of any one of claims 1 to 9, comprising:

the oil pipe corrosion risk basic index system is used for classifying oil pipe corrosion risk basic indexes, so that sample data can be conveniently processed;

the sample data selecting module is used for selecting sample data from the oil pipe corrosion risk basic indexes and processing the sample data;

the risk prediction module is used for establishing an oil pipe corrosion risk prediction SVM model, and evaluating the oil pipe corrosion risk to obtain optimal parameter values of the punishment factor C and the RBF kernel function parameter;

the model optimization module is used for taking the obtained optimal parameter value as the parameter value of the oil pipe corrosion risk prediction SVM model and optimizing the oil pipe corrosion risk prediction SVM model.

Images