Water supply network pipe burst positioning method based on high-frequency pressure data driving

Abstract

The invention discloses a water supply network pipe burst positioning method based on high-frequency pressure data driving, which comprises the following steps: when a high-frequency pressure monitor gives an abnormal alarm, carrying out data cleaning and wavelet noise reduction on real-time pressure signals 6 minutes before and after the alarm, and generating an effective high-frequency pressure data set; extracting a time characteristic value in the effective high-frequency pressure data set by adopting a bilateral cumulative sum algorithm based on a sliding window; in combination with an early-stage wave velocity statistical value, a negative pressure wave method is adopted to position an abnormal pressure wave starting point; if a valve or pump operation point is positioned, normal pipe network operation is listed, meanwhile, the wave velocity of each pipeline is counted and corrected, and if other pipeline nodes are positioned, a pipe explosion positioning alarm is given out. The method has the advantages that the error is accurate to millisecond level by using real-time high-frequency pressure data and adopting a negative pressure wave method, and meanwhile, the wave velocity of each pipeline is corrected by using normal pipe network operation so as to adapt to the influence of factors such as pipeline aging and temperature change, so that the pipe explosion positioning is quicker and more accurate.

Description

Technical Field

[0001] The invention relates to the technical field of water supply pipe network burst positioning technology, and in particular to a water supply pipe network burst positioning method driven by high-frequency pressure data. Background Art

[0002] Some urban water supply network systems have been in use for several years or even decades. Due to the cumulative impact of external environmental factors such as pipeline aging, corrosion, and third-party influences (ground subsidence, surface load), the structural fragility of the water supply pressure network has increased, and pipe bursts often occur. The high incidence of pipe bursts not only causes a large amount of water resources to be wasted, but also causes secondary disasters such as land subsidence and traffic congestion, seriously affecting the maintenance of urban infrastructure and the smooth operation of social order. Therefore, it is necessary to strengthen the control of leakage in the public water supply network. On the basis of completing the informatization of the water supply network, implement intelligent transformation, lay relevant sensors, monitor the operating status of the water supply network in real time, and accurately identify the leakage points of the network.

[0003] It can be seen that conducting research on the precise positioning method of pipe bursts in water supply networks is of great significance for timely repairing pipe bursts and ensuring water supply safety. At present, the research on the positioning of pipe bursts in water supply networks is mainly simulated by hydraulic models or transient flows, which greatly depends on the accuracy of hydraulic model construction and the actual operating conditions of the pipe network. However, the operating conditions of water supply networks in actual projects are complex, and there are certain deviations from computer numerical simulations, that is, the real-time and accuracy of pipe burst positioning is insufficient and needs to be improved urgently. Summary of the invention

[0004] A technical problem to be solved by the present application is to overcome the defects of the above-mentioned related technologies and provide a method for locating burst pipes in a water supply network driven by high-frequency pressure data, the purpose of which is to improve the real-time and accuracy of locating burst pipes in a water supply network.

[0005] The technical solution adopted by the present invention to solve the technical problem is: a method for locating a burst pipe in a water supply network driven by high-frequency pressure data, which specifically includes the following steps:

[0006] Step 1: When the high-frequency pressure monitor issues an abnormal alarm, data cleaning and wavelet noise reduction are performed on the fixed-frequency real-time pressure signal 6 minutes before and after the alarm to generate a high-frequency pressure data set;

[0007] Step 2: Use the sliding window-based bilateral cumulative sum algorithm to extract the time characteristic values ​​of the high-frequency pressure data set;

[0008] Step 3: Based on the previous wave velocity statistics, the pressure wave propagation principle is used to locate the starting point of the abnormal pressure wave;

[0009] Step 4: If the location is a valve or pump operation point, it is included in the normal pipe network operation, and the wave velocity of each pipeline is statistically corrected; if the location is other pipeline nodes, a pipe burst location alarm is issued.

[0010] Preferably, step 1 performs data cleaning and wavelet denoising on a specific high-frequency real-time pressure signal and specifically includes the following steps:

[0011] Step 1.1: When the high-frequency pressure monitor issues a pipe burst warning, the fixed-frequency original pressure signal of each high-frequency pressure monitor 6 minutes before and after the alarm is automatically read. The fixed frequency is the set high-frequency pressure monitor acquisition frequency of 128Hz;

[0012] Step 1.2: Clean the original pressure signal and check that there are no gaps, misalignments or obvious abnormalities in the cleaned pressure signal to ensure that the signal time of multiple pressure monitoring points is synchronized;

[0013] Step 1.3: Use discrete wavelet transform to reduce the noise of the cleaned pressure signal and generate a high-frequency pressure data set.

[0014] As an improvement, step 1.2 performs data cleaning on the original pressure signal, specifically including the following steps:

[0015] Step 1.2.1: If there is a short-term missing in the pressure monitoring point data, the pressure monitoring data within this time period is linearly interpolated to fill the gap. Short time means less than <1s;

[0016] Step 1.2.2: If the pressure monitoring point data is missing for a long time, a high-frequency pressure monitor failure warning is issued and the pressure monitoring point data is discarded;

[0017] Step 1.2.3: Using the steady-state pressure under normal operating conditions as a benchmark, set high and low thresholds to exclude obviously abnormal data from the pressure monitoring point data.

[0018] Preferably, the noise reduction process in step 1.3 specifically includes:

[0019] Step 1.3.1: Perform multi-level discrete wavelet decomposition on the pressure signal after cleaning, select db4 wavelet basis, universal threshold and soft threshold function,

[0020]

[0021]

[0022] In the formula, λ is the threshold, σ is the standard deviation of the noise, N is the length of the original signal, and d j，k represents the kth wavelet coefficient at the jth decomposition scale in the wavelet transform, median(|d j，k |) is the median value of the amplitude of all high-frequency sub-band wavelet coefficients, is the wavelet coefficient after threshold processing;

[0023] Step 1.3.2: Discrete wavelet multi-level decomposition, observe the decomposed signal, ensure that the noise signal is uniform, the effective signal is smooth, no high-frequency signal is decomposed, and determine the wavelet decomposition level k;

[0024] Step 1.3.3: Discrete wavelet transform k-level reconstruction to obtain the effective high-frequency pressure smoothing signal after noise reduction.

[0025] Preferably, step 2 of extracting the time characteristic value of the high-frequency pressure data set is implemented by the following steps:

[0026] Step 2.1: The high-frequency pressure time series of the monitoring point is P = {p(k)}, k = 1, 2, 3, ... N, and the statistical value in the bilateral cumulative sum algorithm is calculated

[0027]

[0028] In the formula, is the cumulative sum of upper offsets, is the cumulative sum of the downward offset, μ0 is the average steady-state pressure before the change point, and β is the drift parameter;

[0029] Step 2.2: Select the data 4 seconds before and 1 second after the reading point as a sliding window with a fixed length of d, calculate the sliding window standard deviation Std and mean μ0, take the inverse and normalize it so that β is in the range [0,0.07]; when the pressure is in a steady state, and Random fluctuations around 0. If the pressure suddenly changes, a positive shift occurs and the mean after the shift is greater than μ0+β, then these shifts are accumulated to the statistic middle, Similarly, when a negative offset occurs and the mean after the offset is less than μ0-β, the offset accumulates to the statistic middle, Continuously increasing;

[0030]

[0031] Where Std is the standard deviation of the sliding window, β is the drift parameter, fre is the signal frequency, and p is the mean of the pressure data in the sliding window;

[0032] Step 2.3: Fluctuation returns to normal, calculate sampling point p k The average of the first half minute, when the pressure value is near the average, the accumulation and reset to zero;

[0033] Step 2.4: Set the threshold value H = max(3σ, 1), where σ is the standard deviation of the actual monitoring pressure in the previous three minutes. That is, when the current pressure fluctuates gently, σ is small and the threshold value is 1; when the current pressure fluctuates greatly, σ is large and the threshold value is 3σ; when or When it is greater than 0, record the time T and continue to calculate. or If the accumulated value is always less than the threshold value H, it will start to accumulate again from 0; or When the accumulated value is greater than the threshold value H, a sudden change event is detected, and time T is the moment of sudden pressure change.

[0034] Preferably, step 3 uses the pressure wave propagation principle to locate the starting point of the abnormal pressure wave through the following steps:

[0035] Step 3.1: Discretize the network topology with the high-frequency pressure monitor points in the pipeline network to be monitored as nodes, and use graph theory to preliminarily locate a certain pipe section;

[0036] Step 3.2: Based on the time difference of the high-frequency pressure monitors upstream and downstream of the pipe section, further accurate positioning is performed using the principle of pressure waves

[0037]

[0038] In the formula, X L is the distance between the burst point and the upstream monitoring point, L is the length of the burst pipe section, a is the pressure wave velocity of the pipe section, and v is the liquid flow velocity in the pipe;

[0039] Step 3.4: Select different upstream and downstream high-frequency pressure monitors and calculate X L ={X L1 , X L2 , X L3 ...X LN}, calculate the final positioning point;

[0040] Step 3.5: At the same time, based on the pipeline network wave velocity database, calculate the 68% and 95% confidence intervals to obtain the high and low risk pipe burst location intervals.

[0041] Preferably, the preliminary positioning in step 3.1 is achieved by the following steps:

[0042] Step 3.1.1: The topological structure of the pipe network to be monitored is regarded as a directed graph, and each high-frequency pressure monitoring point is regarded as a node V = {V1, V2, V3, ... V n}, n is the total number of high-frequency pressure monitors, and each pipe section is regarded as an edge of the graph theory E = {e1, e2, e3, ...e m}, m is the total number of pipe segments, and any pipe segment ek =(V i , V j )∈E and node V i ∈V and V j ∈V are related to each other, then the sets V and E form a pipe network diagram G = (V, E); and are represented by the association matrix A

[0043]

[0044] Among them, A is composed of element a ij The n×m-order matrix formed;

[0045] Step 3.1.2: Set virtual nodes on each pipe segment i Equal i , add each virtual node to the virtual set S, and calculate each virtual node s i The time difference between different pressure monitoring points when a pressure fluctuation event occurs, and the node time difference built-in library of the algorithm is established;

[0046] Step 3.1.3: When a pipeline network pressure fluctuation event occurs, read the actual arrival time of the pressure wave at each pressure monitoring point and apply the following formula:

[0047]

[0048] In the formula, j and k are two different high-frequency pressure monitors, Δt jk is the actual arrival time difference of the pressure waves at the two monitoring points, Δτ jk The theoretical arrival time difference of pressure waves at two monitoring points in the built-in library;

[0049] Step 3.1.4: Find a solution such that E i The virtual node with the smallest value i , initially locate the burst pipe on a certain pipe section.

[0050] Preferably, step 4 is achieved by the following steps:

[0051] Step 4.1: Calculate the pipeline wave velocity between each pressure monitoring point through the pressure fluctuation events at known historical points, remove abnormal values ​​and outliers, and obtain the initial wave velocity;

[0052] Step 4.2: When the actual position is at the valve or pump operating point, continuously correct the pipeline wave velocity

[0053]

[0054] In the formula, v i is the wave velocity of the pressure fluctuation event pipe section i when the valve or pump is located at the operating point, l i is the length of pipe section i, t iis the propagation time of the wave in pipe section i, σ i is the mean square error of the wave velocity statistics of pipe section i, is the average wave velocity in pipe section i;

[0055] Step 4.3: When the position is located at other pipeline points, a pipe burst location alarm is issued.

[0056] Compared with the related art, the present invention has the following advantages: the present invention uses a 128Hz high-frequency pressure sensor as a monitor, and the algorithm no longer resamples the high-frequency pressure signal. Compared with low-frequency steady-state data, the high-frequency data can quickly capture the instantaneous generation and change of pressure waves in the water supply network, ensuring the real-time and accuracy of the burst pipe location, saving precious time for burst pipe repair;

[0057] The present invention completely uses high-frequency pressure data to drive the location of pipe bursts in the water supply network, and does not require a large amount of computing cost to accurately calibrate the hydraulic model of the pipe burst area, which is convenient for the later maintenance and update of pipelines in the water supply network or changes in the network topology structure, and is more suitable for actual engineering applications;

[0058] The present invention continuously updates pipeline wave velocity parameters in subsequent normal operation, adapts to the influence of factors such as pipeline aging and temperature changes, and improves the accuracy of the model method in long-term applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flow chart of the present invention.

[0060] Figure 2 It is a flow chart of a method for extracting time characteristic values ​​of a data set based on a bilateral cumulative sum algorithm (CUSUM) of a sliding window in the present invention.

[0061] Figure 3 It is a schematic diagram of reading alarm points and mutation points by the bilateral cumulative sum algorithm of the present invention.

[0062] Figure 4 It is a 128 Hz high frequency pressure monitoring data curve after noise reduction in a water release experiment in the verification example of the present invention.

[0063] Figure 5 It is the wave velocity distribution diagram of each pipe section in the verification example of the present invention.

[0064] Figure 6 It is the final positioning result of the water release experiment in the verification example of the present invention. DETAILED DESCRIPTION

[0065] First, those skilled in the art should understand that these implementations are only used to explain the technical principles of the embodiments of the present application, and are not intended to limit the protection scope of the embodiments of the present application. Those skilled in the art can make adjustments to them as needed to adapt to specific application scenarios.

[0066] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0067] The preferred embodiment is as follows Figure 1 A method for locating a burst pipe in a water supply network driven by high-frequency pressure data is shown, and specifically comprises the following steps:

[0068] Step 1: When the high-frequency pressure monitor issues an abnormal alarm, the fixed-frequency real-time pressure signal 6 minutes before and after the alarm is cleaned and denoised by wavelet to generate a high-frequency pressure data set P = {p(k)}, k = 1, 2, 3, ... N; here, the fixed frequency is the set high-frequency pressure monitor acquisition frequency of 128Hz, and the 6 minutes before and after the alarm refers to the 3 minutes before the abnormal alarm and the 3 minutes after the abnormal alarm;

[0069] Step 2: Use the sliding window-based bilateral cumulative sum algorithm to extract the time characteristic values ​​of the high-frequency pressure data set;

[0070] Step 3: Based on the previous wave velocity statistics, the pressure wave propagation principle is used to locate the starting point of the abnormal pressure wave;

[0071] Step 4: If the location is a valve or pump operation point, it is included in the normal pipe network operation, and the wave velocity of each pipeline is statistically corrected; if the location is other pipeline nodes, a pipe burst location alarm is issued.

[0072] In this embodiment, step 1 performs data cleaning and wavelet denoising on a specific high-frequency real-time pressure signal, specifically comprising the following steps:

[0073] Step 1.1: When the high-frequency pressure monitor issues a pipe burst warning, the fixed-frequency original pressure signal of each high-frequency pressure monitor 6 minutes before and after the alarm is automatically read;

[0074] Step 1.2: Clean the original pressure signal and check that there are no gaps, misalignments or obvious abnormalities in the cleaned pressure signal to ensure that the signal time of multiple pressure monitoring points is synchronized;

[0075] Step 1.3: Use discrete wavelet transform to perform noise reduction on the cleaned pressure signal to generate a high-frequency pressure data set P = {p(h)}, k = 1, 2, 3, ...N.

[0076] In this embodiment, step 1.2 of performing data cleaning on the original pressure signal specifically includes:

[0077] Step 1.2.1: If there is a short-term (<1s) missing in the pressure monitoring point data, linear interpolation is used to fill the gap in the pressure monitoring data within that time period;

[0078] Step 1.2.2: If the pressure monitoring point data is missing for a long time, a high-frequency pressure monitor failure warning is issued and the pressure monitoring point data is discarded;

[0079] Step 1.2.3: Using the steady-state pressure under normal operating conditions as a benchmark, set high and low thresholds to exclude obvious abnormal data in the pressure monitoring signal.

[0080] In this embodiment, step 1.3 of the noise reduction process specifically includes:

[0081] Step 1.3.1: Perform multi-level discrete wavelet decomposition on the pressure signal after cleaning, select db4 wavelet basis, universal threshold and soft threshold function,

[0082]

[0083] In the formula, λ is the threshold, σ is the standard deviation of the noise, N is the length of the original signal, and d j，k represents the kth wavelet coefficient at the jth decomposition scale in the wavelet transform, median(|d j，k |) is the median value of the amplitude of all high-frequency sub-band wavelet coefficients, is the wavelet coefficient after threshold processing;

[0084] Step 1.3.2: Discrete wavelet multi-level decomposition, observe the decomposed signal, ensure that the noise signal is uniform, the effective signal is smooth, no high-frequency signal is decomposed, and determine the wavelet decomposition level k;

[0085] Step 1.3.3: Discrete wavelet transform k-level reconstruction to obtain the effective high-frequency pressure smoothing signal after noise reduction.

[0086] In this embodiment, step 2 of extracting the time characteristic value of the high-frequency pressure data set is implemented by the following steps:

[0087] Step 2.1: The high-frequency pressure time series of the monitoring point is P = {p(k)}, k = 1, 2, 3, ... N, and the statistical value in the bilateral cumulative sum algorithm is calculated

[0088]

[0089] In the formula, is the cumulative sum of upper offsets, is the cumulative sum of the downward offset, μ0 is the average steady-state pressure before the change point, and β is the drift parameter;

[0090] Step 2.2: Select the data 4 seconds before and 1 second after the reading point as a sliding window with a fixed length of d, calculate the sliding window standard deviation Std and mean μ0, take the inverse and normalize it so that β is in the range [0,0.07]; when the pressure is in a steady state, and Random fluctuations around 0. If the pressure suddenly changes, a positive shift occurs and the mean after the shift is greater than μ0+β, then these shifts are accumulated to the statistic middle, Similarly, when a negative offset occurs and the mean after the offset is less than μ0-β, the offset accumulates to the statistic middle, Continuously increasing;

[0091]

[0092] Where Std is the standard deviation of the sliding window, β is the drift parameter, fre is the signal frequency, and p is the mean of the pressure data in the sliding window;

[0093] Step 2.3: Fluctuation returns to normal, calculate sampling point p k The average of the first half minute, when the pressure value is near the average, the accumulation and reset to zero;

[0094] Step 2.4: Set the threshold value H = max(3σ, 1), where σ is the standard deviation of the actual monitoring pressure in the previous three minutes. That is, when the current pressure fluctuates gently, σ is small and the threshold value is 1; when the current pressure fluctuates greatly, σ is large and the threshold value is 3σ; when or When it is greater than 0, record the time T and continue to calculate. or If the accumulated value is always less than the threshold value H, it will start to accumulate again from 0; or When the accumulated value is greater than the threshold value H, a sudden change event is detected, and time T is the moment of sudden pressure change.

[0095] In this embodiment, step 3 uses the pressure wave propagation principle to locate the starting point of the abnormal pressure wave through the following steps:

[0096] Step 3.1: Discretize the network topology with the high-frequency pressure monitor points in the pipeline network to be monitored as nodes, and use graph theory to preliminarily locate a certain pipe section;

[0097] Step 3.2: Use the pressure wave principle to further accurately locate the pipe according to the time difference between the high-frequency pressure monitors upstream and downstream of the pipe section

[0098]

[0099] In the formula, XL is the distance between the burst point and the upstream monitoring point, L is the length of the burst pipe section, a is the pressure wave velocity of the pipe section, and v is the liquid flow velocity in the pipe;

[0100] Step 3.4: Select different upstream and downstream high-frequency pressure monitors and calculate X L ={X L1 , X L2 , X L3 ...X iN}, calculate the final positioning point;

[0101] Step 3.5: At the same time, based on the pipeline network wave velocity database, calculate the 68% and 95% confidence intervals to obtain the high and low risk pipe burst location intervals.

[0102] In this embodiment, step 3.1 preliminary positioning is achieved by the following steps:

[0103] Step 3.1.1: The topological structure of the pipe network to be monitored is regarded as a directed graph, and each high-frequency pressure monitoring point is regarded as a node V = {V1, V2, V3, ... V n}, n is the total number of high-frequency pressure monitors, and each pipe section is regarded as an edge of the graph E = {e1, e2, e3, ..., e m}, m is the total number of pipe segments, and any pipe segment e k =(V i , V j )∈E and node V i ∈V and V j ∈V are related to each other, then the sets V and E form a pipe network diagram G = (V, E); and are represented by the association matrix A

[0104]

[0105] Among them, A is composed of element a ij The n×m-order matrix formed;

[0106] Step 3.1.2: Set “virtual nodes” on each pipe segment i Equal i , add each virtual node to the virtual set S, and calculate each virtual node s i The time difference between different pressure monitoring points when a pressure fluctuation event occurs, and the node time difference built-in library of the algorithm is established;

[0107] Step 3.1.3: When a pipeline network pressure fluctuation event occurs, read the actual arrival time of the pressure wave at each pressure monitoring point and apply the following formula:

[0108]

[0109] In the formula, j and k are two different high-frequency pressure monitors, Δt jk is the actual arrival time difference of the pressure waves at the two monitoring points, Δτ jk The theoretical arrival time difference of pressure waves at two monitoring points in the built-in library;

[0110] Step 3.1.4: Find a solution such that E i The virtual node with the smallest value i , initially locate the burst pipe on a certain pipe section.

[0111] In this embodiment, step 4 is implemented by the following steps:

[0112] Step 4.1: Calculate the pipeline wave velocity between each pressure monitoring point through the pressure fluctuation events at known historical points, remove abnormal values ​​and outliers, and obtain the initial wave velocity;

[0113] Step 4.2: When the actual position is at the valve or pump operating point, continuously correct the pipeline wave velocity

[0114]

[0115]

[0116] In the formula, v i is the wave velocity of the pressure fluctuation event pipe section i when the valve or pump is located at the operating point, l i is the length of pipe section i, t i is the propagation time of the wave in pipe section i, σ i is the mean square error of the wave velocity statistics of pipe section i, is the average wave velocity in pipe section i;

[0117] Step 4.3: When the positioning is at other pipeline points, a pipe burst positioning alarm is issued; other pipeline points are points on the pipeline that are not valve or pump operating points.

[0118] Verification example:

[0119] The present invention takes a water release experiment of a long-distance tree-like water supply network of N length on November 23, 2023 as an example to further illustrate the implementation process of the water supply network burst positioning method driven by high-frequency pressure data of the present invention.

[0120] Step 1: When the high-frequency pressure monitor issues an abnormal alarm, the 128Hz real-time pressure signal for 6 minutes before and after the alarm is cleaned and denoised by wavelet to generate a high-frequency pressure data set. In this water release experiment, the high-frequency pressure monitor issued an alarm at 09:05 on November 23, 2023. The six-minute high-frequency pressure raw data from 09:02 to 09:08 were read, and the abnormal pressure monitoring data above 100m and below 1m were removed. It was checked whether there was a data missing problem, and the missing data was filled by linear interpolation or the high-frequency pressure raw data of the faulty high-frequency pressure monitor was removed by alarm. At the same time, the discrete wavelet transform was used to denoise the signal to generate a valid high-frequency pressure data set.

[0121] like Figure 2 The following are the effective high-frequency pressure data curves generated for each pressure monitor. It can be seen that the pressure suddenly dropped, which was different from the pipe burst event, and then slowly and steadily recovered. The water release experiment caused the pressure wave to rise, and there was an oscillation phenomenon afterwards.

[0122] Step 2: Use the sliding window-based bilateral cumulative sum algorithm to extract the time characteristic values ​​of the effective high-frequency pressure data set. Put the effective high-frequency pressure data set into the bilateral cumulative sum algorithm, adjust the appropriate parameters, extract the pressure wave arrival time characteristic values ​​in the pressure signal, and fuse the characteristic values ​​extracted from several monitoring points into step 3.

[0123] Step 3: Discretize the pipe network to be monitored and initially locate the burst point on a certain pipe section. Combined with the previous wave velocity statistics, the pressure wave propagation principle, i.e., the negative pressure wave method, is used to accurately locate the starting point of the abnormal pressure wave. The weighted value of the final burst point position is calculated based on the distance between the monitoring point and the burst point.

[0124] like Figure 3 As shown, in order to verify the accuracy of the positioning results, the present invention uses the positioning results to compare with the actual water release experiment records to verify the positioning accuracy of the method. According to the experimental records, the water release point is located 225m upstream of the 58th and 59th high-frequency pressure monitors, and the algorithm locates it 155m upstream of the 58th and 59th high-frequency pressure monitors, with a difference of 70m. Within a reasonable range, the response from detection to specific positioning is rapid, which successfully saves precious time for the burst pipe repair.

[0125] Step 4: Other pipeline points that are not valve or pump operation points are actually located, and the algorithm issues a positioning alarm.

[0126] In general, the positioning results are basically accurate and the response is rapid. The method described in the present invention is of great significance for timely repairing burst pipes and ensuring water supply safety.

[0127] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for locating burst pipes in a water supply network based on high-frequency pressure data, characterized in that: The specific steps include: Step 1: When the high-frequency pressure monitor issues an abnormal alarm, data cleaning and wavelet noise reduction are performed on the fixed-frequency real-time pressure signal 6 minutes before and after the alarm to generate a high-frequency pressure data set; Step 2: Use the sliding window-based bilateral cumulative sum algorithm to extract the time characteristic values ​​of the high-frequency pressure data set; Step 3: Based on the previous wave velocity statistics, the pressure wave propagation principle is used to locate the starting point of the abnormal pressure wave; Step 4: If the location is a valve or pump operation point, it is included in the normal pipe network operation, and the wave velocity of each pipeline is statistically corrected; if the location is other pipeline nodes, a pipe burst location alarm is issued.

2. According to claim 1, a method for locating burst pipes in a water supply network based on high-frequency pressure data drive is characterized in that: Step 1: Data cleaning and wavelet noise reduction of the fixed-frequency real-time pressure signal specifically include the following steps: Step 1.1: When the high-frequency pressure monitor issues a pipe burst warning, the fixed-frequency original pressure signal of each high-frequency pressure monitor 6 minutes before and after the alarm is automatically read. The fixed frequency refers to the set high-frequency pressure monitor acquisition frequency of 128Hz; Step 1.2: Clean the original pressure signal and check that there are no gaps, misalignments or obvious abnormalities in the cleaned pressure signal to ensure that the signal time of each pressure monitoring point is synchronized; Step 1.3: Use discrete wavelet transform to reduce the noise of the cleaned pressure signal and generate a high-frequency pressure data set.

3. The method for locating burst pipes in a water supply network based on high-frequency pressure data drive according to claim 2 is characterized in that: Step 1.2: Data cleaning of the original pressure signal includes the following steps: Step 1.2.1: If there is a short-term missing in the pressure monitoring point data, the pressure monitoring data within this time period is linearly interpolated to fill the gap. Short time means less than <1s; Step 1.2.2: If the pressure monitoring point data is missing for a long time, a high-frequency pressure monitor failure warning is issued, and the high-frequency pressure monitoring point data is discarded; Step 1.2.3: Using the steady-state pressure under normal operating conditions as a benchmark, set high and low thresholds to exclude obvious abnormal data in the pressure monitoring point data.

4. The method for locating burst pipes in a water supply network based on high-frequency pressure data drive according to claim 3 is characterized in that: Step 1.3: Noise reduction processing specifically includes: Step 1.3.1: Perform multi-level discrete wavelet decomposition on the pressure signal after cleaning, select db4 wavelet basis, universal threshold and soft threshold function, In the formula, λ is the threshold, σ is the standard deviation of the noise, N is the length of the original signal, and d j，k represents the kth wavelet coefficient at the jth decomposition scale in the wavelet transform, median(|d j，k |) is the median value of the amplitude of all high-frequency sub-band wavelet coefficients, is the wavelet coefficient after threshold processing; Step 1.3.2: Discrete wavelet multi-level decomposition, observe the decomposed signal, ensure that the noise signal is uniform, the effective signal is smooth, no high-frequency signal is decomposed, and determine the wavelet decomposition level k; Step 1.3.3: Discrete wavelet transform k-level reconstruction to obtain the effective high-frequency pressure smoothing signal after noise reduction.

5. The method for locating burst pipes in a water supply network based on high-frequency pressure data drive according to claim 1, characterized in that: Step 2: Extracting the time characteristic values ​​of the high-frequency pressure data set is achieved by the following steps: Step 2.1: The high-frequency pressure time series of the monitoring point is P = {p(k)}, k = 1, 2, 3, ... N, and the statistical value in the bilateral cumulative sum algorithm is calculated In the formula, is the cumulative sum of upper offsets, is the cumulative sum of the downward offset, μ0 is the average steady-state pressure before the change point, and β is the drift parameter; Step 2.2: Select the data 4 seconds before and 1 second after the reading point as a sliding window with a fixed length of d, calculate the sliding window standard deviation Std and mean μ0, take the inverse and normalize it so that β is in the range [0,0.07]; when the pressure is in a steady state, and Random fluctuations around 0. If the pressure suddenly changes, a positive shift occurs and the mean after the shift is greater than μ0+β, then these shifts are accumulated to the statistic middle, Similarly, when a negative offset occurs and the mean after the offset is less than μ0-β, the offset accumulates to the statistic middle, Continuously increasing; Where Std is the sliding window standard deviation, β is the drift parameter, fre is the signal frequency, is the mean of the pressure data in the sliding window; Step 2.3: Fluctuation returns to normal, calculate sampling point p k The average of the first half minute, when the pressure value is near the average, the accumulation and reset to zero; Step 2.4: Set the threshold value H = max(3σ, 1), where σ is the standard deviation of the actual monitoring pressure in the previous three minutes. That is, when the current pressure fluctuates gently, σ is small and the threshold value is 1; when the current pressure fluctuates greatly, σ is large and the threshold value is 3σ; when or When it is greater than 0, record the time T and continue to calculate. or If the accumulated value is always less than the threshold value H, it will start to accumulate again from 0; or When the accumulated value is greater than the threshold value H, a sudden change event is detected, and time T is the moment of sudden pressure change.

6. The method for locating burst pipes in a water supply network based on high-frequency pressure data drive according to claim 1, characterized in that: Step 3 uses the pressure wave propagation principle to locate the starting point of the abnormal pressure wave through the following steps: Step 3.1: Discretize the network topology using the high-frequency pressure monitor points in the network to be monitored as nodes, and use graph theory to preliminarily locate a certain pipe section; Step 3.2: Based on the time difference of the high-frequency pressure monitors upstream and downstream of the pipe section, further accurate positioning is performed using the principle of pressure waves Where, X L is the distance between the burst point and the upstream monitoring point, L is the length of the burst pipe section, a is the pressure wave velocity of the pipe section, and v is the liquid flow velocity in the pipe; Step 3.4: Select different upstream and downstream high-frequency pressure monitors and calculate X = {X L1 , X L2 , X L3 , ...X LN }, calculate the final positioning point; Step 3.5: At the same time, based on the pipeline network wave velocity database, calculate the 68% and 95% confidence intervals to obtain the high and low risk pipe burst location intervals.

7. A method for locating burst pipes in a water supply network based on high-frequency pressure data drive according to claim 6, characterized in that: Step 3.1 Preliminary positioning is achieved through the following steps: Step 3.1.1: The topological structure of the pipe network to be monitored is regarded as a directed graph, and each high-frequency pressure monitoring point is regarded as a node V = {V1, V2, V3, ... V n }, n is the total number of high-frequency pressure monitors, and each pipe section is regarded as an edge of the graph theory E = {e1, e2, e3, ...e m }, m is the total number of pipe segments, and any pipe segment e k =(V i , V j )∈E and node V i ∈V and V j ∈V are related to each other, then the sets V and E form a pipe network diagram G = (V, E); and are represented by the association matrix A Among them, A is composed of element a ij The n×m-order matrix formed; Step 3.1.2: Set virtual nodes on each pipe segment i Equal i , add each virtual node to the virtual set S, and calculate each virtual node s i The time difference between different pressure monitoring points when a pressure fluctuation event occurs, and the node time difference built-in library of the algorithm is established; Step 3.1.3: When a pipeline pressure fluctuation event occurs, read the actual arrival time of the pressure wave at each pressure monitoring point and apply the following formula: In the formula, j and k are two different high-frequency pressure monitors, Δt jk is the actual arrival time difference of the pressure waves at the two monitoring points, Δτ jk The theoretical arrival time difference of pressure waves at two monitoring points in the built-in library; Step 3.1.4: Find a solution such that E i The virtual node with the smallest value i , initially locate the burst pipe on a certain pipe section.

8. The method for locating burst pipes in a water supply network based on high-frequency pressure data drive according to claim 1, characterized in that: Step 4 is achieved by following these steps: Step 4.1: Calculate the pipeline wave velocity between each pressure monitoring point through the pressure fluctuation events at known historical points, remove abnormal values ​​and outliers, and obtain the initial wave velocity; Step 4.2: When the actual position is at the valve or pump operating point, continuously correct the pipeline wave velocity In the formula, v i is the wave velocity of the pressure fluctuation event pipe section i when the valve or pump is located at the operating point, l i is the length of pipe section i, t i is the propagation time of the wave in pipe section i, σ i is the mean square error of the wave velocity statistics of pipe section i, is the average wave velocity in pipe section i; Step 4.3: When the position is located at other pipeline points, a pipe burst location alarm is issued.

Images