A method for quickly locating leakage in long-distance pressurized water pipelines

Abstract

The invention discloses a method for quickly locating leakage of a long-distance pressurized water pipeline, comprising: collecting real-time flow and water pressure monitoring data during water delivery in the long-distance water pipeline; performing transfer entropy calculation on the acquired real-time flow and water pressure monitoring data to obtain a three-dimensional transfer entropy matrix of real-time data; performing dimensionality reduction processing on the three-dimensional transfer entropy matrix of real-time data to obtain a two-dimensional transfer entropy matrix of real-time data, and performing principal component analysis (PCA) processing on the two-dimensional transfer entropy matrix to obtain each principal component. The invention adopts the method of deploying sensors in the pipe network system to monitor the pressure and vibration signals of the pipe network. The noise of the collected signals is greatly reduced compared with the signals collected by traditional manual listening. The invention also uses transfer entropy to extract features from the data signals. The transfer entropy itself also has strong noise resistance. Whether it is a major accident such as a pipe burst or a minor accident such as a leakage, the method provided by the invention has the characteristics of high sensitivity, high precision, and accurate positioning.

Description

Technical Field

[0001] The present invention relates to the technical field of pressurized water transmission monitoring, and more specifically, to a method for quickly locating leakage in a long-distance pressurized water transmission pipeline. Background Art

[0002] In order to alleviate the shortage of water resources in cities or river basins, meet the water demand of cities or river basins, provide water resources guarantee, promote the optimal allocation of regional water resources and the sustainable development of local economy and society, in accordance with the comprehensive planning of national and river basin water resources, various river basins or regions have vigorously prepared large-scale cross-regional or cross-basin water transfer and diversion projects in recent years. Most of these projects use long-distance pressurized pipelines for water transfer and distribution. Such long-distance pressurized water pipelines have the characteristics of large diameter and long distance, and most of them are in unmanned areas. Accidents are not easy to be discovered, which not only seriously affects the safe and stable operation of the water supply system, but also causes waste of water resources, bringing immeasurable losses to the social economy. According to the report of the Geological Environment Supervision Institute of the China Geological Survey, we have pipeline leakage accidents every year, and most of them are concentrated in the peak water use period and spring and winter, which has a major impact on the lives of residents in various places. According to the results of leakage investigations conducted in 76 cities in China, 1,828 leaks were found in the 7,671.82 km pipeline length investigated (an average of 0.24 cases / km), and the total leakage was 10,557.64 / h (an average of 1.38 / km / h). If the leakage per unit length reduced by leakage control is calculated at 1.38 / km / h, then the total leakage will be reduced by 1.508 billion per year, which is equivalent to saving 1.357 billion yuan in direct economic losses caused by leakage each year. Therefore, it is very important for us to monitor the leakage of long-distance pressurized water pipelines.

[0003] Existing leakage detection technologies are mostly divided into two categories. In remote areas, PCCP pipes are used as the research object, mainly for the detection of broken prestressed steel wires. The disadvantage is that it can only detect pipe burst accidents, but cannot detect minor leakage accidents. In urban pipe network systems, most of them are based on manual inspections combined with listening instruments. The collected sound signals contain a lot of noise and need to be inspected regularly. This is not only costly but also difficult to timely identify and accurately locate leaks. Summary of the invention

[0004] The invention provides a method for quickly locating leakage of a long-distance pressurized water pipeline with high sensitivity, high precision and accurate positioning, so as to solve the above-mentioned technical problems in the prior art.

[0005] According to one aspect of the present invention, a method for quickly locating leakage in a long-distance pressurized water pipeline is provided, comprising the following steps:

[0006] Step A1, collecting real-time flow and water pressure monitoring data during water delivery in long-distance water delivery pipelines;

[0007] Step A2, performing transfer entropy calculation on the real-time flow and water pressure monitoring data obtained in step A1 to obtain a three-dimensional transfer entropy matrix of real-time data;

[0008] Step A3, performing dimensionality reduction processing on the three-dimensional transfer entropy matrix of the real-time data to obtain a two-dimensional transfer entropy matrix of the real-time data, and performing principal component analysis (PCA) processing on the two-dimensional transfer entropy matrix to obtain each principal component component;

[0009] Step A4, calculate the Hotelling statistical index T of the real-time monitoring data during water delivery in long-distance water pipelines ∝2 Sum of squared prediction errors SPE α and compare it with the leakage alarm value obtained during normal water delivery to determine if a long-distance water delivery pipeline is leaking;

[0010] Step A5, calculating the contribution of the main component of the signal at each measuring point to determine the leakage location.

[0011] Preferably, the leakage alarm value in step A4 is obtained by following the steps:

[0012] Step B1, according to the pipeline layout, using a flow sensor and a water pressure sensor to obtain flow and water pressure monitoring data of a long-distance water pipeline during normal water delivery;

[0013] Step B2, the flow data and pressure monitoring data obtained in step B1 are respectively formed into a signal matrix according to the positions of sensors along the line, and a sliding window algorithm is introduced to calculate the transfer entropy value between each measuring point based on the sliding window to obtain a three-dimensional transfer entropy matrix;

[0014] Step B3, reducing the dimension of the three-dimensional transfer entropy matrix in step B2 to obtain a two-dimensional transfer entropy matrix and standardizing it, performing principal component analysis (PCA) processing to obtain each principal component component;

[0015] Step B4: Calculate the Hotelling statistical index for normal water delivery in long-distance water pipelines Sum of squared prediction errors SPE α , compared with Hotelling statistical indicators Sum of squared prediction errors SPE α The sensitivity to faults will use the index with high sensitivity to faults as the leakage alarm value of long-distance water pipelines.

[0016] Preferably, in step A2, the real-time flow and water pressure monitoring data obtained in step A1 are used to calculate the transfer entropy, and the specific calculation formula is as follows:

[0017]

[0018] Among them, P(Xn+1,Xn,Yn) represents the joint probability of Xn+1, Xn, and Yn; P(Xn,Yn) represents the joint probability of Xn and Yn; P(Xn+1|Xn) represents the conditional probability of Xn+1 when Xn is known at time n, X represents flow rate, Y represents water pressure, Xn represents the sampled value of flow rate at the current moment, Yn represents the sampled value of water pressure at the current moment, and Xn+1 represents the sampled value of flow rate at the next moment.

[0019] Preferably, the step A3 specifically includes:

[0020] Step A31, calculate the covariance matrix of the two-dimensional transfer entropy matrix T of the real-time data, and obtain the eigenvectors and eigenvalues ​​of the covariance matrix corresponding to the real-time data,

[0021] Step A32, sort the eigenvalues ​​from large to small, select the eigenvalues ​​of the specified contribution rate accumulation, recombine the eigenvectors corresponding to the eigenvalues ​​into a new matrix, and rearrange the corresponding eigenvectors in the order of eigenvalues ​​from large to small to form a reduced-dimensional eigenvector.

[0022] Preferably, the step of calculating the covariance matrix is:

[0023] Step A311, arranging the real-time data two-dimensional transfer entropy matrix by rows;

[0024] Step A312, calculating the average value of each dimension of the two-dimensional transfer entropy matrix of real-time data;

[0025] Step A313, subtract the average value obtained in 2 from each column in the two-dimensional transfer entropy matrix of real-time data, and record it as F;

[0026] Step A314, calculate the covariance matrix of matrix T

[0027] Among them, where: m is the number of rows of matrix T.

[0028] Preferably, the Hotelling statistical index of the real-time monitoring data The calculation formula is:

[0029]

[0030] Where: pc is the eigenvalue of the covariance matrix of the real-time data calculated in step A4, and t is the eigenvector corresponding to the covariance matrix of the real-time data;

[0031] The prediction error square value SPE of the real-time monitoring data α The calculation formula is:

[0032]

[0033] Where: e is the residual.

[0034] Preferably, the above Hotelling statistical index is calculated The variable contribution rate formula is:

[0035]

[0036] Where C=I-PPT, ξi is the i-th column of the identity matrix I.

[0037] Preferably, T 2 Corresponding and prediction error square value SPE α The variable contribution rate formula is:

[0038]

[0039] The invention discloses a method for quickly locating leakage in a long-distance pressurized water pipeline. The method adopts sensors arranged in the pipeline network system to monitor the pressure and vibration signals of the pipeline network. The noise of the collected signals is greatly reduced compared with the signals collected by traditional manual listening. In addition, the invention uses transfer entropy to extract features of the data signal. The transfer entropy itself also has strong noise resistance. Whether it is a major accident such as a pipe burst or a minor accident such as a leakage, the method provided by the invention has the characteristics of high sensitivity, high precision and accurate positioning.

[0040] Compared with the prior art, the present invention overcomes the traditional manual inspection method with listening devices, saving a lot of manpower, material and financial resources.

[0041] The traditional regular inspection and monitoring method is not only costly, but also has problems such as delayed detection. The present invention adopts real-time monitoring and dynamic analysis methods to achieve real-time monitoring of accidents and timely detection. The present invention also designs a fault feedback system connected to mobile phones and computer terminals, so that once a fault occurs, it can be discovered and handled in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a schematic diagram of signal collection of a long-distance pressurized water pipeline of the present invention;

[0043] Figure 2 It is a flow chart of the method for quickly locating leakage in a long-distance pressurized water pipeline of the present invention;

[0044] Figure 3 It is a schematic diagram of the transfer entropy sequence analysis of the present invention. DETAILED DESCRIPTION

[0045] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0046] See also Figure 1 , and combined with Figure 2 and Figure 3 As shown, the present invention provides a method for quickly locating leakage in a long-distance pressurized water pipeline with high sensitivity, high precision and accurate positioning.

[0047] A method for quickly locating leakage in a long-distance pressurized water pipeline of the present invention comprises the following steps:

[0048] Step A1, deploying flow sensors 3 and water pressure sensors 3 along the long-distance pressurized water pipeline 1 to collect real-time flow and water pressure monitoring data during water delivery in the long-distance water pipeline;

[0049] Step A2, performing transfer entropy calculation on the real-time flow and water pressure monitoring data obtained in step A1 to obtain a three-dimensional transfer entropy matrix 7 of real-time data;

[0050] Step A3, performing dimensionality reduction processing on the three-dimensional transfer entropy matrix 7 of the real-time data to obtain a two-dimensional transfer entropy matrix of the real-time data, and performing principal component analysis (PCA) processing on the two-dimensional transfer entropy matrix to obtain each principal component component;

[0051] Step A4: Calculate the Hotelling statistical index of the real-time monitoring data during long-distance water pipeline water delivery Sum of squared prediction errors SPE α and compare it with the leakage alarm value obtained during normal water delivery to determine if a long-distance water delivery pipeline is leaking;

[0052] Step A5, calculate the contribution of the main component of the signal at each measuring point and determine the leakage position 2.

[0053] Wherein, the leakage alarm value of step A4 of the present invention is obtained by the following steps:

[0054] Step B1, according to the pipeline layout, the flow sensor 3 and the water pressure sensor 3 are used to obtain the flow rate and water pressure monitoring data of the long-distance water pipeline during normal water delivery;

[0055] Step B2, the flow data and pressure monitoring data obtained in step B1 are respectively formed into a signal matrix 5 according to the positions of the sensors 3 along the line, and the sliding window 8 algorithm is introduced to calculate the transfer entropy value 6 between each measuring point based on the sliding window 8, and the corresponding transfer entropy matrix is ​​obtained, and the measuring points of the long-distance pressurized water pipeline 1 are divided into intervals according to the influence of the states between the measuring points.

[0056] The measured flow and pressure monitoring signals 4 are formed into a signal matrix 5, wherein the number of columns is the number of sensors 3, and the number of rows is the number of signals measured by the sensors 3. All the sensor 3 signals of a certain length are intercepted by a sliding window 8 algorithm, and the transfer entropy values ​​6 between the signals of each sensor 3 are mutually calculated to obtain a three-dimensional transfer entropy matrix 7;

[0057] The schematic diagram of the signal matrix 5 and the calculation transfer entropy matrix is ​​as follows:

[0058]

[0059] Step B3, reducing the dimension of the three-dimensional transfer entropy matrix 7 in step B2 to obtain a two-dimensional transfer entropy matrix and standardizing it, performing principal component analysis (PCA) processing on it to obtain each principal component component;

[0060] The specific steps of dimensionality reduction processing are as follows: assuming that there are p measurement points in an interval and the sequence length is X, it can be seen from S2 that a P*P transfer entropy matrix can be calculated in a sliding window 8. Combining the sequence length and the sliding step size, it can be seen that a three-dimensional transfer entropy matrix 7 of P*P*(XL) / 1 will be obtained in the end. Each P*P matrix is ​​expanded row by row into a 1*(P*P) vector. Since the transfer entropy of a variable to itself is always 0, 6 groups of vectors with transfer entropy of 0 are removed, so the three-dimensional transfer entropy matrix 7 is expanded into a two-dimensional transfer entropy matrix T(P*PP)*((XL) / 1).

[0061] Step B4: Calculate the Hotelling statistical index for normal water delivery in long-distance water pipelines Sum of squared prediction errors SPE α , compared with Hotelling statistical indicators Sum of squared prediction errors SPE α The sensitivity to faults will use the index with high sensitivity to faults as the leakage alarm value of long-distance water pipelines.

[0062] Among them, step A2 of the present invention calculates the transfer entropy of the real-time flow and water pressure monitoring data obtained in step A1, and the specific calculation formula is as follows:

[0063]

[0064] Among them, P(Xn+1,Xn,Yn) represents the joint probability of Xn+1, Xn, and Yn; P(Xn,Yn) represents the joint probability of Xn and Yn; P(Xn+1|Xn) represents the conditional probability of Xn+1 when Xn is known at time n, X represents flow rate, Y represents water pressure, Xn represents the sampled value of flow rate at the current moment, Yn represents the sampled value of water pressure at the current moment, and Xn+1 represents the sampled value of flow rate at the next moment.

[0065] Wherein, step A3 of the present invention specifically comprises:

[0066] Step A31, calculate the covariance matrix of the two-dimensional transfer entropy matrix T of the real-time data, and obtain the eigenvectors and eigenvalues ​​of the covariance matrix corresponding to the real-time data,

[0067] Step A32, sort the eigenvalues ​​from large to small, select the eigenvalues ​​of the specified contribution rate accumulation, recombine the eigenvectors corresponding to the eigenvalues ​​into a new matrix, and rearrange the corresponding eigenvectors in the order of eigenvalues ​​from large to small to form the eigenvector after dimensionality reduction.

[0068] The steps of calculating the covariance matrix of the present invention are:

[0069] Step A311, arranging the real-time data two-dimensional transfer entropy matrix by rows;

[0070] Step A312, calculating the average value of each dimension of the two-dimensional transfer entropy matrix of real-time data;

[0071] Step A313, subtract the average value obtained in 2 from each column in the two-dimensional transfer entropy matrix of real-time data, and record it as F;

[0072] Step A314, calculate the covariance matrix of matrix T

[0073] Among them, where: m is the number of rows of matrix T.

[0074] Among them, the Hotelling statistical indicators of real-time monitoring data The calculation formula is:

[0075]

[0076] Where: pc is the eigenvalue of the covariance matrix of the real-time data calculated in step A4, and t is the eigenvector corresponding to the covariance matrix of the real-time data;

[0077] The squared prediction error SPE of real-time monitoring data α The calculation formula is:

[0078] SPE α =e T e

[0079] Where: e is the residual.

[0080] Among them, the calculation of the above Hotelling statistical indicators The variable contribution rate formula is:

[0081]

[0082] Where C=I-PPT, ξi is the i-th column of the identity matrix I.

[0083] Among them, T 2 Corresponding and prediction error square value SPE α The variable contribution rate formula is:

[0084]

[0085] The present invention provides a method for quickly locating leakage in a long-distance pressurized water pipeline. The method uses a sensor 3 arranged in the pipeline system to monitor the pipeline pressure and vibration signals. The noise of the collected signal is greatly reduced compared with the signal collected by traditional manual listening. In addition, the present invention uses transfer entropy to extract features from the data signal. The transfer entropy itself also has strong noise resistance. Whether it is a major accident such as a pipe burst or a minor accident such as a leakage, the method provided by the present invention has the characteristics of high sensitivity, high precision and accurate positioning.

[0086] Compared with the prior art, the present invention overcomes the traditional manual inspection method with listening devices, saving a lot of manpower, material and financial resources.

[0087] The traditional regular inspection and monitoring method is not only costly, but also has problems such as delayed detection. The present invention adopts real-time monitoring and dynamic analysis methods to achieve real-time monitoring of accidents and timely detection. The present invention also designs a fault feedback system connected to mobile phones and computer terminals, so that once a fault occurs, it can be discovered and handled in time.

[0088] Finally, the method of the present application is only a preferred implementation scheme and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for quickly locating leakage in a long-distance pressurized water pipeline, characterized in that: The following steps are involved: Step A1, collecting real-time flow and water pressure monitoring data during water delivery in long-distance water delivery pipelines; Step A2, performing transfer entropy calculation on the real-time flow and water pressure monitoring data obtained in step A1 to obtain a three-dimensional transfer entropy matrix of real-time data; Step A3, performing dimensionality reduction processing on the three-dimensional transfer entropy matrix of the real-time data to obtain a two-dimensional transfer entropy matrix of the real-time data, and performing principal component analysis (PCA) processing on the two-dimensional transfer entropy matrix to obtain each principal component component; Step A4, calculate the Hotelling statistical index T of the real-time monitoring data during water delivery in long-distance water pipelines ∝2 Sum of squared prediction errors SPE α and compare it with the leakage alarm value obtained during normal water delivery to determine if a long-distance water delivery pipeline is leaking; The leakage alarm value obtained during normal water delivery in step A4 is obtained by the following steps: Step B1, according to the pipeline layout, using a flow sensor and a water pressure sensor to obtain flow and water pressure monitoring data of a long-distance water pipeline during normal water delivery; Step B2, the flow data and pressure monitoring data obtained in step B1 are respectively formed into a signal matrix according to the positions of sensors along the line, and a sliding window algorithm is introduced to calculate the transfer entropy value between each measuring point based on the sliding window to obtain a three-dimensional transfer entropy matrix; Step B3, reducing the dimension of the three-dimensional transfer entropy matrix in step B2 to obtain a two-dimensional transfer entropy matrix and standardizing it, performing principal component analysis (PCA) processing to obtain each principal component component; Step B4, calculate the Hotelling statistical index T when the long-distance water pipeline is normally delivering water ∝2 Sum of squared prediction errors SPE α , compared with Hotelling statistical index T ∝2 Sum of squared prediction errors SPE α Sensitivity to faults: the index with high sensitivity to faults is used as the leakage alarm value for long-distance water pipelines; Step A5, calculating the contribution of the main component of the signal at each measuring point to determine the leakage location.

2. A method for quickly locating leakage in a long-distance pressurized water pipeline according to claim 1, characterized in that: In step A2, the real-time flow and water pressure monitoring data obtained in step A1 are used to calculate the transfer entropy. The specific calculation formula is as follows: Among them, P(Xn+1,Xn,Yn) represents the joint probability of Xn+1, Xn, and Yn; P(Xn,Yn) represents the joint probability of Xn and Yn; P(Xn+1|Xn) represents the conditional probability of Xn+1 when Xn is known at time n, X represents flow rate, Y represents water pressure, Xn represents the sampled value of flow rate at the current moment, Yn represents the sampled value of water pressure at the current moment, and Xn+1 represents the sampled value of flow rate at the next moment.

3. A method for quickly locating leakage in a long-distance pressurized water pipeline according to claim 1, characterized in that: The step A3 specifically includes: Step A31, calculate the covariance matrix of the two-dimensional transfer entropy matrix T of the real-time data, and obtain the eigenvectors and eigenvalues ​​of the covariance matrix corresponding to the real-time data, Step A32, sort the eigenvalues ​​from large to small, select the eigenvalues ​​of the specified contribution rate accumulation, recombine the eigenvectors corresponding to the eigenvalues ​​into a new matrix, and rearrange the corresponding eigenvectors in the order of eigenvalues ​​from large to small to form a reduced-dimensional eigenvector.

4. A method for quickly locating leakage in a long-distance pressurized water pipeline according to claim 3, characterized in that: The steps of calculating the covariance matrix are: Step A311, arranging the real-time data two-dimensional transfer entropy matrix by rows; Step A312, calculating the average value of each dimension of the two-dimensional transfer entropy matrix of real-time data; Step A313, subtract the average value obtained in step A312 from each column in the two-dimensional transfer entropy matrix of real-time data, denoted as F; Step A314, calculate the covariance matrix of the two-dimensional transfer entropy matrix T of the real-time data Among them, where: m is the number of rows of the two-dimensional transfer entropy matrix T of real-time data.

5. A method for quickly locating leakage in a long-distance pressurized water pipeline according to claim 4, characterized in that: Hotelling statistical indicators of the real-time monitoring data The calculation formula is: Where: pc is the eigenvalue of the covariance matrix of the real-time data calculated in step A4, and t is the eigenvector corresponding to the covariance matrix of the real-time data; The prediction error square value SPE of the real-time monitoring data α The calculation formula is: SPE α =and T And Where: e is the residual.

Images