Search

CN-121993744-A - Method and system for monitoring leakage of buried gas pipeline under external disturbance

CN121993744ACN 121993744 ACN121993744 ACN 121993744ACN-121993744-A

Abstract

The invention discloses a monitoring method and a monitoring system for leakage of a buried gas pipeline under the action of external disturbance, wherein the monitoring method comprises the following steps of collecting multi-source data of the gas pipeline and preprocessing the multi-source data to obtain input characteristics, constructing a data set according to the input characteristics, dividing the data set into a training set and a testing set, establishing three CatBoost regression prediction models, namely a first model, a second model and a third model, and training the Catboost regression prediction model by utilizing training set data, wherein the first model is used for judging whether the pipeline is leaked or not, the second model is used for identifying external disturbance, the third model is used for identifying the positions of leakage points/external disturbance of the pipeline, obtaining optimal super-parameter combinations, and then carrying out visualization and performance evaluation on the prediction results of the CatBoost regression prediction model based on the prediction effect of the CatBoost regression prediction model by testing set data.

Inventors

  • LIU XIAOFEI
  • CHEN BAOJUN
  • YU BOYUAN
  • Ku Xuwen
  • Chao Yutong
  • ZHANG QIMING
  • ZHOU XIN
  • XUE SHENG
  • LI YAOBIN
  • Qian Jifa
  • WANG HAO
  • MA YINGFEI

Assignees

  • 中国矿业大学
  • 安徽理工大学
  • 上海应用技术大学
  • 上海海事大学

Dates

Publication Date
20260508
Application Date
20260128

Claims (10)

  1. 1. The method for monitoring leakage of the buried gas pipeline under the action of external disturbance is characterized by comprising the following steps of: S10, collecting multi-source data of a gas pipeline, wherein the multi-source data comprises a temperature signal, a disturbance sound wave signal and a leakage sound signal, and preprocessing the multi-source data to obtain input characteristics; S20, constructing a data set according to the input characteristics, and dividing the data set into a training set and a testing set; S30, three CatBoost regression prediction models, namely a first model, a second model and a third model, are established, and training is carried out on the Catboost regression prediction models by utilizing training set data, wherein the first model is used for judging whether leakage occurs in a pipeline, the second model is used for identifying external disturbance, and the third model is used for identifying pipeline leakage points/external disturbance positions; And S40, performing Bayesian optimization iterative processing on the CatBoost regression prediction model to obtain an optimal super-parameter combination, and then checking the prediction effect of the CatBoost regression prediction model based on the test set data, and performing visualization and performance evaluation on the prediction result of the CatBoost regression prediction model.
  2. 2. The method for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 1, wherein, In the step S10, preprocessing is performed on the multi-source data, including the steps of: s11, applying Min-Max normalization to all input variables, wherein the mathematical definition is as follows: in the formula, Is a normalized value, and is generally between 0 and 1; maximum value of the feature in the original dataset; Is the minimum value of the feature in the original dataset; In the reasoning stage, to restore the model output to the actual unit dimension, an inverse normalization transformation is performed on the normalized predicted values: S12, further introducing a second-order polynomial characteristic expansion and cross term construction mechanism, and setting an input vector as follows: the extended expression is: where x represents the original input feature vector, For the i-th feature component, As a dimension of the features, Is a second order cross term between different features.
  3. 3. The method for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 1, wherein, In step S40, the training process of the CatBoost regression prediction model is as follows: S41, initializing a weak learner, typically a decision tree, denoted as The predicted result is the initial predicted value Wherein an error exists between the initial predicted value and the true value; s42, in the regression task, calculating the residual error of each sample, i.e. the true value Predicted value of current model Is the difference of (2) , wherein, Representing the number of iterative rounds, calculating a negative gradient of the loss function with respect to the current model predictive value in the classification task The expression is: S43, constructing a new decision tree by using the residual error or negative gradient obtained by calculation as a new target value and using a symmetrical tree structure mode ; S44, updating the current model according to the newly trained decision tree, and updating the formula into the following formula , wherein, The learning rate is used for controlling the contribution degree of each tree to model updating; S45, repeating the steps S42-S44, continuously training a new decision tree and updating the model until the preset iteration times are reached, and converging the loss function to a certain degree.
  4. 4. The method for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 1, wherein, The method further comprises the step of preprocessing the original data by the improved GTBS method before the step S41, and specifically comprises the following steps: Using weighting coefficients Prior distribution term Smoothing is performed to obtain an improved GTBS method and to address GBDT discrete feature problems, the expression for the improved GTBS method is as follows: in the formula, To smooth the feature values of the kth sample in the ith feature dimension, Is the first The first sample is at The values in the dimensions of the individual features, Is the first The target variable corresponding to the individual samples is, In order to smooth the coefficient of the coefficient, Is a priori value.
  5. 5. The method for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 1, wherein, In the step S40, performing bayesian optimization iteration on the CatBoost regression prediction model to obtain an optimal super-parameter combination includes: s401, defining a search space of super parameters of CatBoost; S402, establishing a Gaussian process as a proxy model for approximating the mapping relation between the objective function of CatBoost model and the super-parameters; s403, taking root mean square error of CatBoost model under K-fold cross validation as an objective function, and randomly selecting K groups of super parameter combinations as initial evaluation points; s404, carrying out L rounds of iterative optimization on the super parameters of CatBoost, in each round of iterative optimization, updating a proxy model based on the data of all the current estimated points, calculating expected improved acquisition function EI values of all the points in a search space by using the expected improved acquisition function, selecting a super parameter combination with the maximum EI value as a point to be estimated in the next round, training CatBoost by using the super parameter combination with the maximum EI value, and simultaneously calculating root mean square error of the super parameter combination with the maximum EI value under five-fold cross validation; And S405, finishing K+L times of objective function evaluation including K initial evaluation points and L rounds of iterative optimization, and selecting a superparameter combination with the minimum root mean square error as an optimal superparameter combination.
  6. 6. The method for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 5, wherein, The expression for the desired improved acquisition function EI is: In the formula { In the hope of the improvement, In the case of a combination of super-parameters, Is the predicted mean value of the proxy model, For the best performance value currently observed, For the currently optimal combination of super-parameters, In order to be able to adjust the exploration parameters, As an intermediate parameter, a parameter which is a function of the parameter, Is the standard deviation of the proxy model, Is a cumulative distribution function of a standard normal distribution, Is a probability density function of a standard normal distribution.
  7. 7. A monitoring system for buried gas pipeline leakage under external disturbance, comprising: The data acquisition module is configured to acquire multi-source data of the gas pipeline, wherein the multi-source data comprises a temperature signal, a disturbance sound wave signal and a leakage sound signal, and the multi-source data is preprocessed to obtain input characteristics; the data dividing module is configured to construct a data set according to the input characteristics and divide the data set into a training set and a testing set; The model training module is configured to be used for building three CatBoost regression prediction models, namely a first model, a second model and a third model, training the Catboost regression prediction model by utilizing training set data, wherein the first model is used for judging whether leakage occurs in a pipeline, the second model is used for identifying external disturbance, and the third model is used for identifying pipeline leakage points/external disturbance positions; the model prediction module is configured to perform bayesian optimization iterative processing on the CatBoost regression prediction model to obtain an optimal super-parameter combination, and then, based on the test set data, examine the prediction effect of the CatBoost regression prediction model, and perform visualization and performance evaluation on the prediction result of the CatBoost regression prediction model.
  8. 8. The system for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 7, The data dividing module comprises: a normalization unit configured to apply Min-Max normalization to all input variables, the mathematical definition of which is as follows: in the formula, Is a normalized value, and is generally between 0 and 1; maximum value of the feature in the original dataset; Is the minimum value of the feature in the original dataset; In the reasoning stage, to restore the model output to the actual unit dimension, an inverse normalization transformation is performed on the normalized predicted values: The data expansion unit is configured to further introduce a second-order polynomial feature expansion and cross term construction mechanism, and set the input vector as: the extended expression is: where x represents the original input feature vector, For the i-th feature component, As a dimension of the features, Is a second order cross term between different features.
  9. 9. The system for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 7, The model prediction module includes: An initial prediction unit configured to initialize a weak learner, denoted as The predicted result is the initial predicted value Wherein an error exists between the initial predicted value and the true value; a loss function unit configured to calculate a residual, i.e. a true value, of each sample in the regression task Predicted value of current model Is the difference of (2) , wherein, Representing the number of iterations, and in the classification task, calculating the negative gradient of the loss function with respect to the current model predictive value: The decision tree unit uses the residual error or negative gradient obtained by calculation as a new target value and constructs a new decision tree by using a symmetrical tree structure mode ; A model updating unit configured to update the current model according to the newly trained decision tree, and update the formula as , wherein, The learning rate is used for controlling the contribution degree of each tree to model updating; And the loop iteration unit is configured to train a new decision tree continuously and update the model until the preset iteration times are reached and the loss function converges to a certain degree.
  10. 10. The system for monitoring leakage of a buried gas pipeline under the action of external disturbance according to claim 7, The model prediction module further includes: an improvement GTBS unit configured to employ weight coefficients Prior distribution term Smoothing is performed to obtain an improved GTBS method and to address GBDT discrete feature problems, the expression for the improved GTBS method is as follows: in the formula, To smooth the feature values of the kth sample in the ith feature dimension, Is the first The first sample is at The values in the dimensions of the individual features, Is the first The target variable corresponding to the individual samples is, In order to smooth the coefficient of the coefficient, Is a priori value.

Description

Method and system for monitoring leakage of buried gas pipeline under external disturbance Technical Field The invention relates to the technical field of gas pipeline leakage monitoring, in particular to a method and a system for monitoring leakage of a buried gas pipeline under the action of external disturbance. Background During transportation, distribution and use, natural gas may leak due to pipeline aging, construction problems or natural disasters, etc. The buried gas pipeline is used as a life line for urban energy transportation, and the safe and stable operation of the buried gas pipeline is directly related to public safety and social economic development. In recent years, as the speed of the urban process and the development of underground space are increased, pipeline leakage accidents caused by external disturbance are frequent. At present, scholars mostly aim at the overhead pipeline without external disturbance in research of pipeline leakage monitoring technology, and the research of identifying and researching the leakage of the buried pipeline with complex working conditions and more external disturbance is less, so that the research of the buried gas pipeline leakage monitoring technology suitable for the external disturbance environment has higher practical significance. The distributed optical fiber monitoring is used as a monitoring means with high sensitivity and high positioning precision, can realize distributed continuous monitoring of a few meters to tens of kilometers, and can accurately capture local tiny leakage or small-range external disturbance tiny signals by means of millimeter-level to meter-level spatial resolution to construct a non-blind area monitoring network, and meanwhile, the high-precision positioning of abnormal points is realized by an optical signal space-time analysis technology, so that accurate guidance is provided for field investigation. Patent CN114352947a proposes a method, a system, a device and a storage medium for gas pipeline leakage detection, which are based on a first sequence and a preset expert scoring model, and identify a gas pipeline leakage state by using an LSTM network. Patent CN116246659a proposes a gas pipeline leakage acoustic wave signal processing method and system, which is based on blind source separation technology to extract leakage acoustic wave characteristic quantity of low frequency band to realize recognition of gas pipeline leakage state. Patent CN118423622a proposes a method for monitoring leakage of a low-pressure gas pipeline by using a distributed optical fiber, when the low-pressure gas pipeline leaks, sound waves are generated and propagated forward in soil, soil particles vibrate and propagate to the surface of the optical fiber, further the internal optical path of the optical fiber changes, a leakage signal is transmitted back to a computer end, and the leakage position and the pipeline running state of the pipeline can be visually seen after signal processing. The method can fill the gap in the field of gas pipeline leakage monitoring to a certain extent, but has obvious limitations that 1. The monitoring blind area is large, most of the prior art is selected to be parallel to the pipeline and laid at a certain distance, and the effect of omnibearing monitoring of pipeline leakage cannot be achieved. 2. The multi-parameter fusion mechanism is lack of monitoring and analyzing of multi-focus single physical parameters in the prior art, a cooperative correlation model of leakage characteristic parameters and external disturbance multi-parameters is not established, and a 'simple disturbance signal' and a 'disturbance induced leakage signal' are difficult to distinguish, so that missed judgment or misjudgment is easily caused by parameter coupling in a complex disturbance environment. 3. The complex environment has insufficient anti-interference capability, and when the complex environment is subjected to multi-source interference such as soil noise, peripheral traffic vibration, industrial electromagnetic interference and the like in a buried scene, characteristic signals are easy to submerge, and a targeted anti-interference signal processing strategy is lacked. 4. The coupling mechanism is not deeply analyzed, the traditional distributed monitoring mostly adopts a single optical fiber sensing technology, only single parameters of temperature or strain can be obtained, and the coupling mechanism of leakage and external disturbance is difficult to comprehensively reflect. Disclosure of Invention Aiming at the problems and the demands, the proposal provides a method and a system for monitoring the leakage of the buried gas pipeline under the action of external disturbance, the technical object described above can be achieved by adopting the following technical features, and other technical effects are brought about. The invention aims to provide a method for monitoring leakage of a buried gas pipeline under the acti