CN110454684A - A kind of fault location test method of city natural gas pipe network leakage - Google Patents
A kind of fault location test method of city natural gas pipe network leakage Download PDFInfo
- Publication number
- CN110454684A CN110454684A CN201910283926.9A CN201910283926A CN110454684A CN 110454684 A CN110454684 A CN 110454684A CN 201910283926 A CN201910283926 A CN 201910283926A CN 110454684 A CN110454684 A CN 110454684A
- Authority
- CN
- China
- Prior art keywords
- svm
- detection
- pipe network
- classification
- fault location
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F17—STORING OR DISTRIBUTING GASES OR LIQUIDS
- F17D—PIPE-LINE SYSTEMS; PIPE-LINES
- F17D5/00—Protection or supervision of installations
- F17D5/005—Protection or supervision of installations of gas pipelines, e.g. alarm
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F17—STORING OR DISTRIBUTING GASES OR LIQUIDS
- F17D—PIPE-LINE SYSTEMS; PIPE-LINES
- F17D5/00—Protection or supervision of installations
- F17D5/02—Preventing, monitoring, or locating loss
Abstract
The invention discloses a kind of fault location test methods of city natural gas pipe network leakage, it is related to Discussion on Pipe Leakage Detection Technology field, solve the problems, such as that existing pipeline network leak detection accuracy is lower, its key points of the technical solution are that: the following steps are included: obtaining the geographical location information and pipe network attribute data of gas distributing system, extract the pressure data and data on flows of each node operation;SVM is constructed according to the Structural risk minization principle in statistics, and determines the kernel functional parameter of SVM;The feature vector that each node is concentrated in detection is extracted, and feature vector is inputted in SVM;Classification and Detection is carried out to feature vector using the classification feature of SVM;According to the fault location information of geographical location information and the output gas distributing system leakage of classification and Detection result, with the uncertain influence to pipeline network leak detection for reducing city natural gas pipe network operating condition, the effect of the fault location detection accuracy of city natural gas pipe network leakage is improved.
Description
Technical field
The present invention relates to Discussion on Pipe Leakage Detection Technology fields, more specifically, it relates to which a kind of city natural gas pipe network is let out
The fault location test method of leakage.
Background technique
Intelligent diagnostics are an important development directions in current Engineering Diagnosis field, and research is advised from test data set off in search
Rule is identified and is predicted to malfunction using these rules.The existing machine learning based on data, including neural network
Inside, common most important theories basis first is that statistics.Traditional statistics research be that number of samples tends to infinity when
Asymptotic theory, but in practical problem, sample number is often limited.Statistical Learning Theory (Statistical Learning
Theory or SLT) it is a kind of theory for specializing in machine learning rule under Small Sample Size.V.Vapnik et al. is from six, seven
The ten's started to be dedicated to study in this respect, and to the mid-90, Statistical Learning Theory starts by more and more extensive attention.
The VC dimension that SVM (support vector machine or SVM) is built upon SLT is theoretical former with Structural risk minization
On the basis of reason, for the neural network that compares, there is good popularization performance, SVM is asked in solution small sample, non-linear and higher-dimension
Many advantages are showed in topic, and can be generalized in other Machine Learning Problems, and machine Learning Theory and skill will be effectively pushed
The development of art.
Currently, gas distributing system is because of factors such as pipe body corrosion, temperature change, pipeline soil deformation and product qualities,
It occur frequently that pipeline gas leakage leakage and bursting accident, bring about great losses.Therefore, city natural gas pipe network pipe leakage is pre-
Surveying with diagnosis is always to the important topic of city natural gas pipe network safe operation research.Thermal infrared imaging method, Magnetic Flux Leakage Inspecting method,
Though fiber laser arrays method, acoustic emission etc. have obtained good result in fuel gas pipeline leakage detection.But city gas pipeline is set
Meter is generally from loop network, and pipe network node is more, mostly mesolow pipeline, is located at by urban transport artery or in residential block,
Ambient enviroment is complex, pipeline leakage signal be often submerged in complex working condition operation and interference noise in, it is difficult to identification and
Extract leakage signal.Based on artificial neural network technology intelligent diagnosing method, by the air pressure for testing adjacent gas ductwork node
And data on flows, position, fault degree and the fault incidence of explosion are diagnosed with artificial neural network technology, still,
Artificial Neural Network solve gas distributing system operation in nonlinear data mapping and convergence problem in existing defects.
Therefore, how to design a kind of fault location test method of city natural gas pipe network leakage is that we compel to be essential at present
It solves the problems, such as.
Summary of the invention
The object of the present invention is to provide a kind of fault location test methods of city natural gas pipe network leakage, have and reduce city
The uncertain influence to pipeline network leak detection of city's gas distributing system operating condition, improves the event of city natural gas pipe network leakage
Hinder the effect of detection and localization precision.
Above-mentioned technical purpose of the invention has the technical scheme that a kind of city natural gas pipe network is let out
The fault location test method of leakage, comprising the following steps:
S1: the geographical location information and pipe network attribute data of gas distributing system are obtained, is extracted in the pipe network attribute data
The pressure data and data on flows of each node operation, and the integrated detection collection of pressure data and data on flows that each node is run;
S2: SVM is constructed according to the Structural risk minization principle in statistics, and determines the core of SVM
Function parameter;
S3: the feature vector that each node is concentrated in detection is extracted, and feature vector is inputted in SVM;
S4: classification and Detection is carried out to feature vector using the classification feature of SVM;
S5: according to the fault location information of geographical location information and the output gas distributing system leakage of classification and Detection result.
The present invention is further arranged to: the classification and Detection specific steps of the SVM are as follows:
Setting detection collection are as follows: xi∈Rn,yi∈ { -1,1 }, i=1 ..., l;Kernel function is k (xi,yi), k corresponds to certain feature sky
Between inner product in Z, i.e.,It is greater than
TransformationIt will test sample and be mapped to feature space from the input space;
Two classifiers based on SVM are designed, the optimal classification surface under definite meaning is found in Z
When detection collection in Z linearly can timesharing, keep class interval maximum, that is, solve:
When sample concentrates on linearly inseparable in Z, class interval and classification error is made to reach certain compromise, that is, solved:
Wherein, ξiIt is slack variable, ξi>=0, C are regularization parameter;
According to (1) and (2), its dual problem is solved:
Wherein, α=(a1,a2,…,al)T, αiIt is the corresponding Lagrangian Product-factor of inequality constraints in formula (1);
Hessian matrix Q is positive semi-definite:
Above-mentioned planning problem is solved, two classifiers are obtained:
Wherein, y (x)=sign (u (x));If ai=0, detect sample xiReferred to as non-supported vector;If ai> 0 detects sample
This xiReferred to as support vector;If ai=C detects sample xiReferred to as bounded support vector;If 0 < ai< C detects sample xiIt is referred to as non-
Bounded support vector.
The present invention is further arranged to: the kernel function is Radial basis kernel function:
K(xi,xj)=exp-γ | | xi-xj||2Formula (5)
Wherein, xi、xjTwo detection vectors are respectively indicated, γ indicates the width of radial base;Since the kernel function is negative exponent
Function, thus its index value can not ether it is big, otherwise kernel function is insensitive to the variation of index;For this purpose, enabling:
E(γ||xi-xj||2=1, can use γ=1/E (| | xi-xj||2;
Wherein, E indicates mathematic expectaion.
The present invention is further arranged to: the width parameter γ of the regularization parameter C and Radial basis kernel function is using intersection
Effective verification method is determined.
In conclusion the invention has the following advantages: using SVM, in the fault location to pipe leakage
In detection process, weakens and small sample, the processing of non-linear and higher-dimension are asked brought by the uncertainty because of pipe network operation operating condition
The influence to testing result is inscribed, the fault location detection accuracy of city natural gas pipe network leakage is improved.
Detailed description of the invention
Fig. 1 is the flow chart in the embodiment of the present invention.
Specific embodiment
Below in conjunction with attached drawing, invention is further described in detail.
A kind of embodiment: fault location test method of city natural gas pipe network leakage, comprising the following steps:
Step 1 obtains the geographical location information and pipe network attribute data of gas distributing system, extracts the pipe network attribute number
The pressure data and data on flows of each node operation in, and the integrated detection of pressure data and data on flows that each node is run
Collection.
Step 2 constructs SVM according to the Structural risk minization principle in statistics, and determines SVM
Kernel functional parameter.
Step 3 extracts the feature vector that each node is concentrated in detection, and feature vector is inputted in SVM.
Step 4 carries out classification and Detection to feature vector using the classification feature of SVM.
Step 5 is believed according to the fault location of geographical location information and the output gas distributing system leakage of classification and Detection result
Breath.
The classification and Detection specific steps of SVM are as follows:
Setting detection collection are as follows: xi∈Rn,yi∈ { -1,1 }, i=1 ..., l;Kernel function is k (xi,yi), k corresponds to certain feature sky
Between inner product in Z, i.e.,It is greater than
TransformationIt will test sample and be mapped to feature space from the input space.
Two classifiers based on SVM are designed, the optimal classification surface under definite meaning is found in Z
When detection collection in Z linearly can timesharing, keep class interval maximum, that is, solve:
When sample concentrates on linearly inseparable in Z, class interval and classification error is made to reach certain compromise, that is, solved:
Wherein, ξiIt is slack variable, ξi>=0, C are regularization parameter.
According to (1) and (2), its dual problem is solved:
Wherein, α=(a1,a2,…,al)T, αiIt is the corresponding Lagrangian Product-factor of inequality constraints in formula (1);
Hessian matrix Q is positive semi-definite:
Above-mentioned planning problem is solved, two classifiers are obtained:
Wherein, y (x)=sign (u (x));If ai=0, detect sample xiReferred to as non-supported vector;If ai> 0 detects sample
This xiReferred to as support vector;If ai=C detects sample xiReferred to as bounded support vector;If 0 < ai< C detects sample xiIt is referred to as non-
Bounded support vector.
Formula (1), (2), (3) are all convex programming problems, and locally optimal solution, that is, globally optimal solution of convex programming.Therefore,
SVM method avoids locally optimal solution problem existing for the methods of neural network.SVM method has clearly geometry meaning, In
Formula (1) and (2) be equivalent to respectively solve feature space in two classes detection sample formed 2 convex closures or diminution convex closure it
Between distance.Method of geometry is exactly to convert the training problem of SVM to classical geometrical issues using this explanation.As one
The solution of a QP problem, formula (4) is not theoretically difficult.For some special circumstances, usually gradually approached by iteration
Optimal solution.When sample is smaller, existing algorithm, such as interior-point algohnhm, gradient projection method can be directly utilized.
For nonlinear problem, optimal hyperlane can be constructed in the higher-dimension inner product space, carry out inner product calculating.According to functional
It is related theoretical, as long as a kind of kernel function R (xi,xj) meeting Merce condition, it just corresponds to the inner product of a certain transformation space.Therefore,
Using kernel function R (x appropriatei,xj) achieve that the linear classification after a certain nonlinear transformation.Briefly, SVM
It is exactly that the input space is transformed to by a higher dimensional space by inner product kernel function first, then seeks optimal classification in this space
Face.
In the present embodiment, kernel function is Radial basis kernel function:
K(xi,xj)=exp-γ | | xi-xj||2Formula (5)
Wherein, xi、xjTwo detection vectors are respectively indicated, γ indicates the width of radial base;Since the kernel function is negative exponent
Function, thus its index value can not ether it is big, otherwise kernel function is insensitive to the variation of index;For this purpose, enabling:
E(γ||xi-xj||2=1, can use γ=1/E (| | xi-xj||2;
Wherein, E indicates mathematic expectaion.
When carrying out fault diagnosis using SVM method, preferable precision in order to obtain, it is necessary to the hyper parameter of selection SVM meticulously.
In the SVM using Radial basis kernel function, hyper parameter includes several following:
(1) regularization parameter C decides the specific gravity of model complexity and detection error in objective function.
(2) what the width parameter γ of Radial basis kernel function was implicit defines non-thread from the input space to high-dimensional feature space
Property mapping, determine the structure of feature space, thus control the complexity of last solution.
In the present embodiment, the width parameter γ of regularization parameter C and Radial basis kernel function is using the effective authentication of intersection
Method is determined.
Sample calculation analysis: herein based on many experiments test data, to the gas distributing system fault diagnosis side based on SVM
Method is tested, and is compared under the premise of identical experiment test data with neural network (ANN) method, to prove it effectively
Property.
Choose average relative error (ΔMRE) and root-mean-square error (ΔRMSE) it is performance indicator, it is defined as follows:
Wherein, (7), n is diagnosis number in (8) formula, r is pipe network number of nodes, xi,jIndicate fault diagnosis at i-th jth point
Value,Indicate physical fault value.When booster failure occurs, xi,j、Value be 1;When booster failure not occurring, xi,j,
Value be 0.
It in identical experiment data basis, carries out giving gas distributing system fault diagnosis with neural network (ANN) method, lead to
Excessive amount calculating, which is compared, to be learnt, the Δ that SVM is diagnosed in experimental dataMREFor the fault diagnosis precision average specific ANN of 1.3%, SVM
Improve 0.6% or so.
Working principle: it is weakened in the fault location detection process to pipe leakage because of pipe network using SVM
Influence brought by the uncertainty of operating condition to small sample, non-linear and higher-dimension processing problem to testing result, is improved
The fault location detection accuracy of city natural gas pipe network leakage.
This specific embodiment is only explanation of the invention, is not limitation of the present invention, those skilled in the art
Member can according to need the modification that not creative contribution is made to the present embodiment after reading this specification, but as long as at this
All by the protection of Patent Law in the scope of the claims of invention.
Claims (4)
1. a kind of fault location test method of city natural gas pipe network leakage, it is characterized in that: the following steps are included:
S1: the geographical location information and pipe network attribute data of gas distributing system are obtained, extracts and is respectively saved in the pipe network attribute data
The pressure data and data on flows of point operation, and the integrated detection collection of pressure data and data on flows that each node is run;
S2: SVM is constructed according to the Structural risk minization principle in statistics, and determines the kernel function of SVM
Parameter;
S3: the feature vector that each node is concentrated in detection is extracted, and feature vector is inputted in SVM;
S4: classification and Detection is carried out to feature vector using the classification feature of SVM;
S5: according to the fault location information of geographical location information and the output gas distributing system leakage of classification and Detection result.
2. the fault location test method of a kind of city natural gas pipe network leakage according to claim 1, it is characterized in that: institute
State the classification and Detection specific steps of SVM are as follows:
Setting detection collection are as follows: xi∈Rn,yi∈ { -1,1 }, i=1 ..., l;Kernel function is k (xi,yi), k corresponds to certain feature space Z
In inner product, i.e.,It is greater than
TransformationX → Z will test sample from the input space and be mapped to feature space;
Two classifiers based on SVM are designed, the optimal classification surface under definite meaning is found in Z
When detection collection in Z linearly can timesharing, keep class interval maximum, that is, solve:
When sample concentrates on linearly inseparable in Z, class interval and classification error is made to reach certain compromise, that is, solved:
Wherein, ξiIt is slack variable, ξi>=0, C are regularization parameter;
According to (1) and (2), its dual problem is solved:
Wherein, α=(a1,a2,…,al)T, αiIt is the corresponding Lagrangian Product-factor of inequality constraints in formula (1);Hessian
Matrix Q is positive semi-definite:
Above-mentioned planning problem is solved, two classifiers are obtained:
Wherein, y (x)=sign (u (x));If ai=0, detect sample xiReferred to as non-supported vector;If ai> 0 detects sample xiClaim
For support vector;If ai=C detects sample xiReferred to as bounded support vector;If 0 < ai< C detects sample xiReferred to as non-bounded
Support vector.
3. the fault location test method of a kind of city natural gas pipe network leakage according to claim 2, it is characterized in that: institute
Stating kernel function is Radial basis kernel function:
K(xi,xj)=exp-γ | | xi-xj||2Formula (5)
Wherein, xi、xjTwo detection vectors are respectively indicated, γ indicates the width of radial base;Since the kernel function is negative exponent letter
Number, thus its index value can not ether it is big, otherwise kernel function is insensitive to the variation of index;For this purpose, enabling:
E(γ||xi-xj||2=1, can use γ=1/E (| | xi-xj||2;
Wherein, E indicates mathematic expectaion.
4. the fault location test method of a kind of city natural gas pipe network leakage according to claim 3, it is characterized in that: institute
The width parameter γ for stating regularization parameter C and Radial basis kernel function is determined using effective verification method is intersected.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910283926.9A CN110454684A (en) | 2019-04-10 | 2019-04-10 | A kind of fault location test method of city natural gas pipe network leakage |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910283926.9A CN110454684A (en) | 2019-04-10 | 2019-04-10 | A kind of fault location test method of city natural gas pipe network leakage |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110454684A true CN110454684A (en) | 2019-11-15 |
Family
ID=68480562
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910283926.9A Pending CN110454684A (en) | 2019-04-10 | 2019-04-10 | A kind of fault location test method of city natural gas pipe network leakage |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110454684A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113486950A (en) * | 2021-07-05 | 2021-10-08 | 华能国际电力股份有限公司上安电厂 | Intelligent pipe network water leakage detection method and system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102567742A (en) * | 2010-12-15 | 2012-07-11 | 中国科学院电子学研究所 | Automatic classification method of support vector machine based on selection of self-adapting kernel function |
CN103322415A (en) * | 2013-06-05 | 2013-09-25 | 哈尔滨工程大学 | Two-dimensional reproduction method for petroleum pipeline defects through least squares support vector machines (LS-SVM) |
CN106762340A (en) * | 2016-12-02 | 2017-05-31 | 国家电网公司 | A kind of hydraulic turbine servomotor intelligence leak detection system and its method |
EP3327206A1 (en) * | 2016-11-25 | 2018-05-30 | Tata Consultancy Services Limited | Ranking pipes for maintenance in pipe networks using approximate hydraulic metrics |
-
2019
- 2019-04-10 CN CN201910283926.9A patent/CN110454684A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102567742A (en) * | 2010-12-15 | 2012-07-11 | 中国科学院电子学研究所 | Automatic classification method of support vector machine based on selection of self-adapting kernel function |
CN103322415A (en) * | 2013-06-05 | 2013-09-25 | 哈尔滨工程大学 | Two-dimensional reproduction method for petroleum pipeline defects through least squares support vector machines (LS-SVM) |
EP3327206A1 (en) * | 2016-11-25 | 2018-05-30 | Tata Consultancy Services Limited | Ranking pipes for maintenance in pipe networks using approximate hydraulic metrics |
CN106762340A (en) * | 2016-12-02 | 2017-05-31 | 国家电网公司 | A kind of hydraulic turbine servomotor intelligence leak detection system and its method |
Non-Patent Citations (1)
Title |
---|
朱晓红等: "支撑向量机在给水管网故障诊断中的应用", 《重庆大学学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113486950A (en) * | 2021-07-05 | 2021-10-08 | 华能国际电力股份有限公司上安电厂 | Intelligent pipe network water leakage detection method and system |
CN113486950B (en) * | 2021-07-05 | 2023-06-16 | 华能国际电力股份有限公司上安电厂 | Intelligent pipe network water leakage detection method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | Variational attention-based interpretable transformer network for rotary machine fault diagnosis | |
CN112087443B (en) | Sensing data anomaly detection method under physical attack of industrial sensing network information | |
Zhao et al. | A novel cap-LSTM model for remaining useful life prediction | |
CN106991049A (en) | A kind of Software Defects Predict Methods and forecasting system | |
CN112199890B (en) | System-level fault diagnosis method for integrated nuclear power device | |
Wang et al. | A multi-stage hybrid fault diagnosis approach for operating conditions of nuclear power plant | |
Li et al. | An accurate leakage localization method for water supply network based on deep learning network | |
Yan et al. | Deep learning technology for chiller faults diagnosis | |
CN113612237A (en) | Method for positioning resonance-induced subsynchronous oscillation source in offshore wind farm | |
Wang et al. | A remaining useful life prediction model based on hybrid long-short sequences for engines | |
Fan | Research on deep learning energy consumption prediction based on generating confrontation network | |
Fan et al. | Two-level diagnosis of heating pipe network leakage based on deep belief network | |
CN110454684A (en) | A kind of fault location test method of city natural gas pipe network leakage | |
Kollias et al. | Machine learning for analysis of real nuclear plant data in the frequency domain | |
Xu et al. | Rethink long-tailed recognition with vision transforms | |
CN112101797B (en) | Dynamic fault diagnosis method and system for complex industrial system | |
Yan et al. | Yolo-pd: Abnormal signal detection in gas pipelines based on improved yolov7 | |
CN106295043A (en) | A kind of parameter determination method of gaussian radial basis function agent model | |
Shi et al. | Recognizing the take-off action pattern of basketball players based on fuzzy neural network system | |
Lee et al. | Event diagnosis method for a nuclear power plant using meta-learning | |
Zhao et al. | Industrial fault diagnosis based on few shot learning | |
Jiang et al. | A Novel Algorithm YOLOv4-Mini to Detection Automobile Rim Weld | |
CN113970073B (en) | ResNet-based water supply network leakage accurate positioning method | |
Guo et al. | Operation stage division and RUL prediction of bearings based on 1DCNN-ON-LSTM | |
CN117764169A (en) | Civil aircraft troubleshooting path reasoning method and system based on knowledge graph |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191115 |