Water supply network leakage analysis method based on water pressure base line

Abstract

The invention discloses a water supply pipe network leakage analysis method based on a water pressure base line, which combines basic information of an urban water supply pipe network and water pressure data monitored in real time to perform real-time analysis on newly added leakage conditions of each subarea in the water supply pipe network and accurately identify leakage points in the water supply pipe network. Comprising the steps of water pressure equipment layout and message analysis, water pressure data preprocessing, pressure baseline calculation based on long-series data, water pressure anomaly analysis based on a pressure baseline and regional newly-added leakage state analysis. The water pressure data abnormity of the water supply pipe network can be efficiently and accurately analyzed, a scientific basis is provided for pipe network management and maintenance of a water supply enterprise, the leakage detection cost is reduced, the pipe network operation and maintenance efficiency is improved, and the stability and safety of urban water supply can be guaranteed.

Description

Technical Field

[0001] The invention relates to water supply network management, fault identification and data processing technology, and in particular to a water supply network leakage analysis method based on a water pressure baseline. Background Art

[0002] In urban water projects, pipe network leakage problems caused by geological changes, material aging, thermal expansion and contraction continue to exist. With the continuous acceleration of urbanization, more and more cement roads and high-rise buildings make it difficult to detect leakage in deep underground pipe networks, and pipe network leaks cannot be discovered in time. According to the "China Urban and Rural Construction Statistical Yearbook" (2022) published by the Ministry of Housing and Urban-Rural Development, the total public water supply in cities and counties across the country in 2022 is 75.167 billion m 3 The water loss is 9.649 billion m 3 , the comprehensive leakage rate is 12.84%. To deal with this problem, water supply companies usually use pipe network modeling, zoning metering, and leak detection technology to locate leakage points. However, these methods are limited by the process of scheme design and equipment procurement, and the implementation cycle is long, so they are not suitable as initial measures for leakage control.

[0003] Pressure is a key hydraulic parameter in the water supply network, and its changes can directly reflect the water flow status and potential problems in the network. Especially in the zoned metering management system, water pressure data is highly sensitive to identifying new leakages, and the measurement is stable and reliable, and easy to obtain in real time. Therefore, it can be used as an important data indicator for evaluating the health status of the network. However, the existing zoned pressure control system cannot directly judge new leakages and has not yet formed a standard method system, which is not suitable for large-scale promotion and use. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to solve the deficiencies in the prior art and to provide a water supply network leakage analysis method based on a water pressure baseline, making full use of the basic information of the urban water supply network and the real-time monitored water pressure data, aiming to conduct real-time analysis of new leakage situations in each zone of the water supply network.

[0005] Technical solution: A water supply network leakage analysis method based on a water pressure baseline of the present invention comprises the following steps:

[0006] Step 1: Hydraulic equipment layout and message analysis;

[0007] The water pressure monitoring equipment in the analysis and study area is defined as monitoring points, and the original messages of each monitoring point are obtained and parsed in real time. The water pressure data in the original message is defined as message pressure, and the time data is defined as message time; n monitoring points are encoded to generate the corresponding monitoring point set {monitor}, the original message data of all monitoring points are obtained and recorded as the message data set {report_data}, and then the message pressure set {report_p} and the message time set {report_t} are stored in the data dictionary {report_data} in the form of {key:value};

[0008] {monitor}={monitor 1 , monitor 2 , ..., monitor n};

[0009] {report_p}={report_p 1 , report_p 2 , ..., report_p l};

[0010] {report_t}={report_t 1 , report_t 2 , ..., report_t l};

[0011]

[0012] In the above formula, monitor n Indicates the nth monitoring point, report_p l and report_t l Respectively represent the message pressure and message time of the lth message data;

[0013] Step 2: pre-process the water pressure data in the data dictionary obtained in step 1 to obtain the corresponding maximum half-range amplitude H of the stable water supply pressure and the water pressure 1 / 2 quantile M, which specifically includes the following contents:

[0014] Step 2.1, data cleaning, including discarding error messages, removing invalid data and supplementing missing data, to obtain the cleaned data set {result_data};

[0015]

[0016] Among them, the message pressure set after data cleaning is: {p′}={p′ 1 , p′ 2 , ..., p′k′};

[0017] The message time set after data cleaning is: {t′}={t′ 1 , t′ 2 , ..., t′ k′};

[0018] k′ is the amount of data in the daily pressure set of each monitoring station, a total of r days,

[0019] Step 2.2, extract the feature values ​​of the cleaned data set {result_data} to obtain the maximum half-range amplitude set {H} and 1 / 2 quantile set {M} of the daily stable water supply of n′ stable monitoring points;

[0020]

[0021]

[0022] Among them, the maximum half-range fluctuation value H of the stable water supply of a certain stable monitoring point on a certain day is equal to half of its maximum pressure fluctuation value p_dif, and the maximum pressure fluctuation value p_dif is the difference between the maximum pressure and the minimum pressure of the stable monitoring point on that day;

[0023] Among them, the 1 / 2 quantile M of a certain stable monitoring point on a certain day is the boundary value that divides the water pressure data of all monitoring points into two parts;

[0024] Step 3: Calculate the pressure baseline based on the long series data, that is, calculate the weighted mean of the water pressure data of each stable monitoring point, which specifically includes the following contents:

[0025] Calculate the weight value normal_ω of each stable monitoring point. First, standardize the 1 / 2 quantile M of each stable monitoring point to obtain the corresponding standardized Then standardize the obtained The value is calculated based on the stability weight ω, and then each weight ω is normalized to obtain the corresponding weight value normal_ω, forming the weight value set {normal_ω} of all stable monitoring points;

[0026]

[0027] For the n′th stable monitoring point monitor n′ The weight value on day r is normal_ω r ;

[0028] Next, for the 1 / 2 quantile M and the weight value normal_ω of each stable monitoring point every day, the daily pressure baseline day_line is calculated by the weighted mean of the weight value, and the monthly pressure baseline month_line and the annual pressure baseline year_line are summarized in turn;

[0029] Step 4: Perform water pressure anomaly analysis based on the pressure baseline;

[0030] For each monitoring point, first calculate the corresponding pressure deviation p_d based on the measured pressure value p, the annual baseline year_line and the maximum half-range value H of a single monitoring point; then calculate the continuous duration of the pressure deviation p_d continuously greater than the pressure deviation limit p_d_limit, recorded as the continuous deviation duration t_d; and calculate the average pressure deviation p_d_avg within the continuous deviation duration t_d; finally, perform water pressure anomaly analysis on a certain monitoring point based on the pressure deviation limit p_d, the continuous deviation duration t_d and the average pressure deviation p_d_avg;

[0031] Step 5: Add leakage status analysis to the area in real time.

[0032] Furthermore, the specific method of discarding erroneous messages in the data cleaning process in step 2.1 is: verifying the data integrity, field consistency, and data redundancy of the perception data messages obtained in real time, identifying and discarding incomplete or damaged messages, messages that do not conform to the expected format, or messages that are repeated or redundant;

[0033] The error message identification rule is defined as rule, and the message discarding method is defined as dis(); after discarding the error message, the original data set {origin_data} is obtained. The original data set is to extract the monitoring point pressure p and time t within T hours from the processed perception data message, and form a message pressure set {p} and a message time set {t} containing k data in chronological order, and store them in the data dictionary {origin_data} in the form of {key: value};

[0034] {p}=dis({report_p}, rule);

[0035] {t}=dis({report_t}, rule);

[0036]

[0037] The specific method of eliminating invalid data in the data cleaning process of step 2.1 is: the values ​​that obviously deviate from the normal data range are collectively referred to as invalid data. Invalid data may be caused by equipment failure, sensor error or data transmission error. The box plot is a statistical chart used to display the dispersion of a set of data. The box plot is used to identify invalid data and eliminate it. The steps are as follows:

[0038] First, calculate the upper quantile Q3, median Q2, and lower quantile Q1 of the water pressure data {p} in the original data set {origin_data}; and calculate the interquartile range IQR as an indicator to measure the width of data distribution;

[0039] IQR = Q3-Q1;

[0040] Then, the upper limit and lower limit of the data are calculated based on IQR. Data exceeding the upper limit / lower limit are regarded as invalid data and are eliminated.

[0041] Upperlimit = Q3 + 15 * IQR;

[0042] Lowerlimit = Q1 + 15 * IQR;

[0043] The specific method of supplementing missing data in the data cleaning process of step 2.1 is: judging whether there are missing values ​​in the data set based on time series analysis, which will affect the accuracy and reliability of subsequent analysis; using cubic spline interpolation to fill in the missing data, which can provide a smoother data curve and ensure the continuity of the first-order and second-order derivatives of the data at the interpolation point;

[0044] The time interval [t 0 , t u ](year-month-day-hour-minute-second) is divided into u subintervals, each subinterval is [t i-1 , t i ], where i = 1, 2, ..., u; in each subinterval [t i-1 , t i ], construct a cubic polynomial S i (t), which has the form:

[0045] S i (t) = a i +b i (tt i )+c i (tt i ) 2 +d i (tt i ) 3

[0046] Among them, a i 、b i 、c i and d i is the coefficient to be determined;

[0047] At any time t∈[t i-1 , t i ], using the obtained cubic polynomial S i (t) is used to calculate the corresponding pressure value p′, and the calculated results are divided by day and stored in the result data set {result_data}, for a total of r days, where k′ is the amount of data in the daily pressure set of each monitoring station.

[0048] The detailed process of feature value extraction in step 2 above is:

[0049] Step 1: Stabilize the maximum half-range amplitude of water supply H:

[0050] The stable water supply state (ss) refers to the state in which the water pressure value at the monitoring point fluctuates within a certain range under normal working conditions of the water supply network, and the fluctuation amplitude is small, and there is no significant abnormal change. The fluctuation amplitude limit is defined as ss_limit. The maximum pressure fluctuation amplitude {p_dif} of each monitoring point is compared with the fluctuation amplitude limit. If it is less than the limit, it is in a stable water supply state.

[0051] First, compare the daily pressure concentration p′ of each monitoring station, calculate the maximum pressure in the pressure concentration, form the maximum pressure set {p_max} and the maximum pressure set {p_min}, and store them in the pressure feature set {p_feature} in the form of {key:value}. The calculation formula is as follows:

[0052] {p_max}={p_max 1 , ..., p_max r}

[0053] ={max(p′ 1 , p′ 2 , ..., p′ k′ ) 1 , ..., max(p′ 1 , p′ 2 , ..., p′ k′ ) r}

[0054] {p_min}={p_min 1 , ..., p_min r}

[0055] ={min(p′1 , p′ 2 , ..., p′ k′ ) 1 , ..., min(p′ 1 , p′ 2 , ..., p′ k′ ) r}

[0056]

[0057] Then calculate the maximum pressure variation p_dif, where p_dif is the difference between the maximum pressure and the minimum pressure, and store all p_dif in the pressure feature set {p_feature};

[0058] {p_max}={p_dif 1 , ..., p_dif r}={p_max 1 -p_min 1 , ..., p_max r -p_min r}

[0059]

[0060] Next, the stable water supply status is judged, and the rules are as follows:

[0061] 1) When the maximum pressure variation value p_dif is less than or equal to ss_limit, it is judged to be in a stable water supply state.

[0062] 2) When the maximum pressure variation value p_dif is greater than ss_limit, it is judged to be in a fluctuating water supply state.

[0063] Eliminate the monitoring points with fluctuating water supply status ss=1, select the monitoring points with stable water supply status ss=0, and save them as the stable monitoring point set {stable_monitor}, the number is n′;

[0064] Finally, the maximum half-range fluctuation value H of stable water supply is calculated based on the maximum fluctuation value of pressure.

[0065] The maximum half-range fluctuation of daily stable water supply at all stable monitoring points is integrated into the maximum half-range fluctuation set {H}, which is expressed as:

[0066] After obtaining the maximum half-range amplitude set {H}, the boundary value of the water pressure data of each monitoring point is calculated to be divided into two parts, that is, the 1 / 2 quantile M. The 1 / 2 quantile M reflects the average level of pressure at each monitoring point. The specific calculation method is:

[0067] First, the daily pressure set {k′} of each monitoring point is sorted from small to large using the quick sort method to ensure that the sorted data is in ascending order;

[0068] If p ′ If is an odd number, the 1 / 2 quantile M is the data point in the middle after sorting;

[0069] If k ′ If is an even number, the 1 / 2 quantile M is the average of the two middle data points after sorting;

[0070] Then, the daily 1 / 2 quantiles M of all stable monitoring points are integrated into the 1 / 2 quantile set {M}; the calculation formula is as follows:

[0071] k ′ When it is an odd number:

[0072] k ′ When it is an even number:

[0073]

[0074] Furthermore, the specific method for calculating the weight of the monitoring point in step 3 is:

[0075] First, the 1 / 2 quantile M of each monitoring point is standardized: the mean μ and standard deviation σ of the 1 / 2 quantile M of all monitoring points are calculated to obtain the standardized 1 / 2 quantile of each station. The 1 / 2 quantile Transformed to standard normal distribution data with mean 0 and standard deviation 1, so that the 1 / 2 quantiles of different sites It can be compared and divided on the same scale, and the calculation formula is as follows:

[0076]

[0077] Due to the standardized is a dimensionless value, indicating the degree of deviation of the original data M from the mean value μ. When , it means that the M of the site is higher than μ; when When , it means that the M of the site is lower than μ; when When , it means that the M of the site is equal to μ;

[0078] Then, according to each monitoring point The weight ω of each monitoring point is calculated based on the stability of the monitoring point. The calculation formula is as follows:

[0079]

[0080] in, The constant C is an adjustment factor used to adjust the distribution range of the weight to avoid the denominator being 0 or a negative number; when the monitoring point The smaller the standardized score is, the more stable it is and the larger the weight ω is; conversely, The larger the standardized score of , the more unstable the identification, and the smaller the weight ω;

[0081] Next, to ensure that the sum of the weights of all sites is 1, the calculated weights are normalized to obtain the final weight value normal_ω, which is calculated as follows:

[0082]

[0083]

[0084] The daily pressure baseline day_line is calculated as follows:

[0085]

[0086] Form a daily pressure baseline set {day_line} for each stable monitoring point every day, {day_line} = {day_line 1 , day_line 2 , ..., day_line r};

[0087] Summarize the daily pressure baseline values ​​day_line, calculate the monthly average pressure baseline value as the monthly pressure baseline month_line, and form the monthly pressure baseline set {month_line}.

[0088] {month_line}={month_line 1 , month_line 2 , ..., day_line 12}

[0089] The monthly pressure baseline values ​​month_line obtained each month are summarized and the average pressure baseline value for the whole year is calculated as the annual pressure baseline year_line; this annual baseline will serve as the benchmark value for pipeline pressure monitoring and leakage analysis in the next year.

[0090] Furthermore, the pressure deviation p_d of a certain monitoring point in step 4 is i The calculation formula is:

[0091]

[0092] pi is the measured pressure value of the monitoring point, H and year_line are the maximum half-range value and monthly pressure baseline corresponding to the monitoring point;

[0093] The calculation formula for the average pressure deviation p_d_avg within the continuous deviation duration is:

[0094]

[0095] p_d_count is the number of pressure values ​​within the continuous deviation duration;

[0096] Then, the abnormal pressure situation is judged, and the judgment rules (Table 2) are:

[0097] 1) If the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is less than or equal to the continuous deviation duration limit t_d_limit, it is judged as a first-level abnormality and the pressure abnormality state p_state is set to 1;

[0098] 2) When the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is greater than the continuous deviation duration limit t_d_limit and the pressure average deviation p_d_avg is less than or equal to the pressure deviation limit p_d_limit, it is judged as a secondary abnormality and the pressure abnormality state p_state is set to 2;

[0099] 3) When the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is greater than the continuous deviation duration limit t_d_limit and the pressure average deviation p_d_avg is greater than the pressure deviation limit p_d_limit, it is judged as a level 3 abnormality and the pressure abnormality state p_state is set to 3;

[0100] p_d_limit is the pressure deviation limit and t_d_limit is the continuous deviation time limit;

[0101] Furthermore, the detailed method of performing real-time analysis on the newly added leakage status of the region in step 5 is as follows:

[0102] If the pressure abnormal state p_state is 1, the newly added leakage point is judged as a newly added first-level leakage;

[0103] If the pressure abnormal state p_state is 2, the newly added leakage point is judged as a newly added secondary leakage;

[0104] If the pressure abnormal state p_state is 3, the newly added leakage point is determined as a newly added third-level leakage.

[0105] Beneficial effects: The present invention can accurately identify abnormal pressure fluctuation patterns, thereby quickly locating possible new leakage points. Compared with the prior art, the present invention can adapt to large-scale, multi-level pipe network systems, effectively analyze leakage problems in complex pipe networks, reduce the number of manual steps, and reduce labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 It is a schematic diagram of the overall analysis process of the present invention;

[0107] Figure 2 It is a schematic diagram of the annual pressure baseline calculation process in the present invention;

[0108] Figure 3 A leakage state analysis flow chart is newly added in the present invention. DETAILED DESCRIPTION

[0109] The technical solution of the present invention is described in detail below, but the protection scope of the present invention is not limited to the embodiments.

[0110] The present invention combines the real-time data of each zone pressure monitoring point with advanced data analysis technology to identify abnormal pressure fluctuation patterns, thereby locating possible new leakage points. It is not only suitable for daily pipe network operation and maintenance, but can also respond quickly in emergency situations, effectively shortening the time window for leak discovery and repair. By reducing the number of unnecessary inspections and the false alarm rate, the present invention effectively reduces the cost of pipe network operation and maintenance. At the same time, it provides water supply companies with a scientific and efficient new leakage management strategy, contributing technical strength to the sustainable use and protection of urban water resources.

[0111] like Figure 1 As shown, the water supply network leakage analysis method based on water pressure data of the present invention comprises the following steps:

[0112] Step 1: Hydraulic equipment layout and message analysis;

[0113] The water pressure monitoring equipment in the analysis and study area is defined as monitoring points, and the original messages of each monitoring point are obtained and parsed in real time. The water pressure data in the original message is defined as message pressure, and the time data is defined as message time; n monitoring points are encoded to generate the corresponding monitoring point set {monitor}, the original message data of all monitoring points are obtained and recorded as the message data set {report_data}, and then the message pressure set {report_p} and the message time set {report_t} are stored in the data dictionary {report_data} in the form of {key:value};

[0114] {monitor}={monitor 1 , monitor 2, ..., monitor n};

[0115] {report_p}={report_p 1 , report_p 2 , ..., report_p l};

[0116] {report_t}={report_t 1 , report_t 2 , ..., report_t l};

[0117]

[0118] In the above formula, monitor n Indicates the nth monitoring point, report_p l and report_t l Respectively represent the message pressure and message time of the lth message data;

[0119] Step 2: pre-process the water pressure data in the data dictionary obtained in step 1 to obtain the corresponding maximum half-range amplitude H of the stable water supply pressure and the water pressure 1 / 2 quantile M, which specifically includes the following contents:

[0120] Step 2.1, data cleaning, including discarding error messages, removing invalid data and supplementing missing data, to obtain the cleaned data set {result_data};

[0121]

[0122] Among them, the message pressure set after data cleaning is: {p′}={p′ 1 , p′ 2 , ..., p′ k′};

[0123] The message time set after data cleaning is: {t′}={t′ 1 , t′ 2 , ..., t′ k′};

[0124] k′ is the amount of data in the daily pressure set of each monitoring station, a total of r days,

[0125] Step 2.2, extract the feature values ​​of the cleaned data set {result_data} to obtain the maximum half-range amplitude set {H} and 1 / 2 quantile set {M} of the daily stable water supply of n′ stable monitoring points;

[0126]

[0127]

[0128] Among them, the maximum half-range fluctuation value H of the stable water supply of a certain stable monitoring point on a certain day is equal to half of its maximum pressure fluctuation value p_dif, and the maximum pressure fluctuation value p_dif is the difference between the maximum pressure and the minimum pressure of the stable monitoring point on that day;

[0129] Among them, the 1 / 2 quantile M of a certain stable monitoring point on a certain day is the boundary value that divides the water pressure data of all monitoring points into two parts;

[0130] Step 3: Calculate the pressure baseline based on the long series data, that is, calculate the weighted mean of the water pressure data of each stable monitoring point, which specifically includes the following contents:

[0131] Calculate the weight value normal_ω of each stable monitoring point. First, standardize the 1 / 2 quantile M of each stable monitoring point to obtain the corresponding standardized Then standardize the obtained The value is calculated based on the stability weight ω, and then each weight ω is normalized to obtain the corresponding weight value normal_ω, forming the weight value set {normal_ω} of all stable monitoring points;

[0132]

[0133] For the n′th stable monitoring point monitor n′ The weight value on day r is normal_ω r ;

[0134] Next, for the 1 / 2 quantile M and the weight value normal_ω of each stable monitoring point every day, the daily pressure baseline day_line is calculated by the weighted mean of the weight value, and the monthly pressure baseline month_line and the annual pressure baseline year_line are summarized in turn;

[0135] Step 4: Perform water pressure anomaly analysis based on the pressure baseline;

[0136] For each monitoring point, first calculate the corresponding pressure deviation p_d based on the measured pressure value p, the annual baseline year_line and the maximum half-range value H of a single monitoring point; then calculate the continuous duration of the pressure deviation p_d continuously greater than the pressure deviation limit p_d_limit, recorded as the continuous deviation duration t_d; and calculate the average pressure deviation p_d_avg within the continuous deviation duration t_d; finally, perform water pressure anomaly analysis on a certain monitoring point based on the pressure deviation limit p_d, the continuous deviation duration t_d and the average pressure deviation p_d_avg;

[0137] Step 5: Real-time analysis of the leakage status of the region;

[0138] Furthermore, the specific method of discarding the error message in the data cleaning process in step 2.1 is:

[0139] The error message identification rule is defined as rule, and the message discarding method is defined as dis(); after discarding the error message, the original data set {origin_data} is obtained. The original data set is to extract the monitoring point pressure p and time t within T hours from the processed perception data message, and form a message pressure set {p} and a message time set {t} containing k data in chronological order, and store them in the data dictionary {origin_data} in the form of {key: value};

[0140] {p}=dis({report_p}, rule);

[0141] {t}=dis({report_t}, rule);

[0142]

[0143] The specific method for removing invalid data during the data cleaning process in step 2.1 is:

[0144] First, calculate the upper quantile Q3, median Q2, and lower quantile Q1 of the water pressure data {p} in the original data set {origin_data}; and calculate the interquartile range IQR as an indicator to measure the width of data distribution;

[0145] IQR = Q3-Q1;

[0146] Then, the upper limit and lower limit of the data are calculated based on IQR. Data exceeding the upper limit / lower limit are regarded as invalid data and are eliminated.

[0147] Upperlimit = Q3 + 15 * IQR;

[0148] Lowerlimit = Q1 + 15 * IQR;

[0149] The specific method of supplementing missing data in the data cleaning process of step 2.1 is:

[0150] The time interval [t 0 , t u ](year-month-day-hour-minute-second) is divided into u subintervals, each subinterval is [t i-1 , t i ], where i = 1, 2, ..., u; in each subinterval [t i-1 , t i ], construct a cubic polynomial S i (t), which has the form:

[0151] S i (t) = a i +b i (tt i )+c i (tt i ) 2 +d i (tt i ) 3

[0152] Among them, a i 、b i 、c i and d i is the coefficient to be determined;

[0153] At any time t∈[t i-1 , t i ], using the obtained cubic polynomial S i (t) to calculate the corresponding pressure value p′, and the calculated results are divided by day and stored in the result data set {result_data}, for a total of r days.

[0154] The detailed process of feature value extraction in step 2 above is:

[0155] Step 1: Stabilize the maximum half-range amplitude of water supply H:

[0156] The stable water supply state (ss) refers to the state in which the water pressure value at the monitoring point fluctuates within a certain range under normal working conditions of the water supply network, and the fluctuation amplitude is small, and there is no significant abnormal change. The fluctuation amplitude limit is defined as ss_limit. The maximum pressure fluctuation amplitude {p_dif} of each monitoring point is compared with the fluctuation amplitude limit. If it is less than the limit, it is in a stable water supply state.

[0157] First, compare the daily pressure concentration p′ of each monitoring station, calculate the maximum pressure in the pressure concentration, form the maximum pressure set {p_max} and the maximum pressure set {p_min}, and store them in the pressure feature set {p_feature} in the form of {key:value}. The calculation formula is as follows:

[0158] {p_max}={p_max 1 , ..., p_max r}

[0159] ={max(p′ 1 , p′ 2 , ..., p′ k′ ) 1 , ..., max(p′ 1 , p′ 2 , ..., p′ k′ ) r}

[0160] {p_min}={p_min 1 , ..., p_min r}

[0161] ={min(p′ 1 , p′ 2 , ..., p′ k′ ) 1 , ..., min(p′ 1 , p′ 2 , ..., p′ k′ ) r}

[0162]

[0163] Then calculate the maximum pressure variation p_dif, where p_dif is the difference between the maximum pressure and the minimum pressure, and store all p_dif in the pressure feature set {p_feature};

[0164] {p_max}={p_dif 1 , ..., p_dif r}={p_max 1 -p_min 1 ,,..,p_max r -p_min r}

[0165]

[0166] Next, the stable water supply state is judged, and the rules (as shown in Table 1) are as follows:

[0167] Table 1 Rules for judging stable water supply status

[0168] Judgment rules Stable state judgment Code p_dif≤ss_limit Stablize 0 p_dif>ss_limit fluctuation 1

[0169] 1) When the maximum pressure variation value p_dif is less than or equal to ss_limit, it is judged to be in a stable water supply state.

[0170] 2) When the maximum pressure variation value p_dif is greater than ss_limit, it is judged to be in a fluctuating water supply state.

[0171] Eliminate the monitoring points with fluctuating water supply status ss=1, select the monitoring points with stable water supply status ss=0, and save them as the stable monitoring point set {stable_monitor}, the number is n′;

[0172] Finally, the maximum half-range fluctuation value H of stable water supply is calculated based on the maximum fluctuation value of pressure.

[0173] The maximum half-range fluctuation of daily stable water supply at all stable monitoring points is integrated into the maximum half-range fluctuation set {H}, which is expressed as:

[0174] After obtaining the maximum half-range amplitude set {H}, the boundary value of the water pressure data of each monitoring point is calculated to be divided into two parts, that is, the 1 / 2 quantile M. The 1 / 2 quantile M reflects the average level of pressure at each monitoring point. The specific calculation method is:

[0175] First, the daily pressure set {p′} of each monitoring point is sorted from small to large using the quick sort method to ensure that the sorted data is in ascending order;

[0176] If ′ If is an odd number, the 1 / 2 quantile M is the data point in the middle after sorting;

[0177] If k ′ If is an even number, the 1 / 2 quantile M is the average of the two middle data points after sorting;

[0178] Then, the daily 1 / 2 quantiles M of all stable monitoring points are integrated into the 1 / 2 quantile set {M}; the calculation formula is as follows:

[0179] k ′ When it is an odd number:

[0180] k ′ When it is an even number:

[0181]

[0182] Furthermore, the specific method for calculating the weight of the monitoring point in step 3 is:

[0183] First, the 1 / 2 quantile M of each monitoring point is standardized: the mean μ and standard deviation σ of the 1 / 2 quantile M of all monitoring points are calculated to obtain the standardized 1 / 2 quantile of each station. The 1 / 2 quantile Transformed to standard normal distribution data with mean 0 and standard deviation 1, so that the 1 / 2 quantiles of different sites It can be compared and divided on the same scale, and the calculation formula is as follows:

[0184]

[0185]

[0186] Due to the standardized is a dimensionless value, indicating the degree of deviation of the original data M from the mean value μ. When , it means that the M of the site is higher than ω; when When , it means that the M of the site is lower than ω; when When , it means that the M of the site is equal to ω;

[0187] Then, according to each monitoring point The weight ω of each monitoring point is calculated based on the stability of the monitoring point. The calculation formula is as follows:

[0188]

[0189] in, The constant C is an adjustment factor used to adjust the distribution range of the weight to avoid the denominator being 0 or a negative number; when the monitoring point The smaller the standardized score is, the more stable it is and the larger the weight ω is; conversely, The larger the standardized score of , the more unstable the identification, and the smaller the weight ω;

[0190] Next, to ensure that the sum of the weights of all sites is 1, the calculated weights are normalized to obtain the final weight value normal_ω, which is calculated as follows:

[0191]

[0192] The daily pressure baseline day_line is calculated as follows:

[0193]

[0194] Form a daily pressure baseline set {day_line} for each stable monitoring point every day, {day_line} = {day_line 1 , day_line 2 , ..., day_line r};

[0195] Summarize the daily pressure baseline values ​​day_line, calculate the monthly average pressure baseline value as the monthly pressure baseline month_line, and form the monthly pressure baseline set {month_line}.

[0196] {month_line}={month_line 1 , month_line 2 , ..., day_line 12}

[0197] The monthly pressure baseline values ​​month_line obtained each month are summarized and the average pressure baseline value for the whole year is calculated as the annual pressure baseline year_line; this annual baseline will be used as the benchmark value for pipeline pressure monitoring and leakage analysis in the next year, such as Figure 2 shown.

[0198] Furthermore, the pressure deviation p_d of a certain monitoring point in step 4 is i The calculation formula is:

[0199]

[0200] p i is the measured pressure value of the monitoring point, H and year_line are the maximum half-range value and monthly pressure baseline corresponding to the monitoring point;

[0201] The calculation formula for the average pressure deviation p_d_avg within the continuous deviation duration is:

[0202]

[0203] p_d_count is the number of pressure values ​​within the continuous deviation duration;

[0204] Then, the abnormal pressure situation is judged, and the judgment rules (Table 2) are:

[0205] 1) If the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is less than or equal to the continuous deviation duration limit t_d_limit, it is judged as a first-level abnormality and the pressure abnormality state p_state is set to 1;

[0206] 2) When the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is greater than the continuous deviation duration limit t_d_limit and the pressure average deviation p_d_avg is less than or equal to the pressure deviation limit p_d_limit, it is judged as a secondary abnormality and the pressure abnormality state p_state is set to 2;

[0207] 3) When the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is greater than the continuous deviation duration limit t_d_limit and the pressure average deviation p_d_avg is greater than the pressure deviation limit p_d_limit, it is judged as a level 3 abnormality and the pressure abnormality state p_state is set to 3;

[0208] p_d_limit is the pressure deviation limit and t_d_limit is the continuous deviation time limit;

[0209] Table 2 Pressure abnormality judgment rules

[0210]

[0211]

[0212] The detailed method of performing real-time analysis on the newly added leakage status of the region in step 5 of this embodiment is as follows:

[0213] If the pressure abnormal state p_state is 1, the newly added leakage point is judged as a newly added first-level leakage;

[0214] If the pressure abnormal state p_state is 2, the newly added leakage point is judged as a newly added secondary leakage;

[0215] If the pressure abnormal state p_state is 3, the newly added leakage point is determined as a newly added third-level leakage.

[0216] Specific rules such as Figure 3 And as shown in Table 3.

[0217] Table 3 Status of newly added leakage points

[0218] Judgment rules Leakage status Code p_state=1 Primary leakage 0 p_state=2 Secondary leakage 1 p_state=3 Level 3 leakage 2

[0219] This embodiment is selected for actual application in a typical community (3,000 households, DN200 water inlet pipe) in a southern city. A total of 129,555 pieces of equipment monitoring data are obtained during the period from 2024-09-01 00:00:00 to 2023-09-30 23:59:59. The monitoring data is recorded every 15 minutes and stored in the original data set {origin_data} after message processing. After a series of calculations, the final result is: a total of 256 new first-level leakages, 31 second-level leakages, and 2 third-level leakages. This embodiment specifically includes the following five steps.

[0220] Step 1. Hydraulic equipment layout and message analysis

[0221] A total of 45 pressure monitoring point information {monitor} was collected and stored in the pressure station set. 582919 message pressure report_p and message time report_t were obtained from the original message to form the message data set {report_data}.

[0222] {monitor}={"1","2",...,"45"}

[0223]

[0224] Step 2. Water pressure data preprocessing

[0225] Perform data preprocessing operations on the report data set {report_data}, including data cleaning and feature value extraction, to obtain the result data set {result_data}.

[0226] (1) Data cleaning

[0227] After discarding the error messages in the message data set, the monitoring data in the period of 2024 / 09 / 01 00:00:00-2024 / 09 / 30 23:59:59 are extracted, totaling 45 records, forming the original data set {origin_data}.

[0228]

[0229] Use a box plot to view the data distribution of the original data set, obtain the upper limit and lower limit of each monitoring point, and remove the data that exceeds the upper and lower limits. Take monitoring point "1" as an example, its upper limit is 0.253 and the lower limit is 0.331, and 50 water pressure data are removed. A total of 7668 water pressure data are removed from all monitoring points.

[0230] Cubic spline interpolation is used to fill in the missing data and the removed data. The processed pressure set {p′} and the processed time set {t′} are calculated according to the cubic polynomial, and the interpolation results are divided by day and stored in the result data set {result_data}. The calculation results are shown by taking monitoring point "1" as an example:

[0231]

[0232] (2) Feature value extraction

[0233] The maximum pressure {p_max}, minimum pressure {p_min}, and maximum amplitude {p_dif} of the pressure concentration are calculated and stored in the pressure feature set {p_feature} in the form of {key:value}. The calculation results are shown using monitoring point "1" as an example:

[0234]

[0235] In this case, the fluctuation amplitude limit ss_limit is set to 0.156. By comparing the maximum pressure fluctuation amplitude with the fluctuation amplitude limit, 23 monitoring points in a stable water supply state are selected and stored as a stable monitoring point set {stable_monitor}, and the maximum half-range fluctuation amplitude of stable water supply {H} of the stable monitoring point is calculated.

[0236] {stable_monitor}={"4","7",...,"43"}

[0237]

[0238] Calculate the 1 / 2 quantile {M} of each monitoring point in the stable monitoring point set.

[0239]

[0240] Step 3. Pressure baseline calculation based on long series data

[0241] The 1 / 2 quantile M of each monitoring point is standardized to obtain According to each monitoring point The weight ω of each monitoring point is calculated based on its stability. The calculated weight is normalized, and the sum of the weights is 100 to obtain the final weight value normal_ω.

[0242]

[0243] The daily stress baseline day_line is calculated by weighted mean to form the daily stress baseline set {day_line}.

[0244] {day_line}={0.183, 0.183,..., 0.183}

[0245] Calculate the mean of the daily stress baseline set as the monthly stress baseline month_line to form the monthly stress baseline set {month_line}.

[0246] {month_line}={0.183, 0.183,..., 0.183}

[0247] Calculate the annual pressure baseline year_line.

[0248] year_line=0.183

[0249] Step 4. Water pressure anomaly analysis based on pressure baseline

[0250] In this case, the pressure deviation limit p_d_limit is set to 4.5, and the continuous deviation duration limit t_d_limit is set to 3 hours. Based on the measured pressure value p, the annual baseline year_line, and the maximum half-range value H of a single monitoring point, the pressure deviation p_d, the continuous deviation duration t_d, and the average pressure deviation p_d_avg are calculated. By comparing the size of p_d, p_d_avg and p_d_limit, and the size of t_d and t_d_limit, it is calculated that there are 256 first-level anomalies, 31 second-level anomalies, and 2 third-level anomalies.

[0251] Judgment rules Abnormal judgment Code p_d>4.5 and t_d≤3 hours Level 1 abnormality 1 p_d>4.5 and t_d>3 hours and p_d_avg≤4.5 Secondary abnormality 2 p_d>4.5 and t_d>3 hours and p_d_avg>4.5 Level 3 abnormality 3

[0252] Step 5. Analysis of new regional leakage status

[0253] According to the pressure anomaly judgment result p_state, a real-time analysis of the newly added leakage status was conducted, with a total of 256 new first-level leakages, 31 second-level leakages, and 2 third-level leakages. After site surveys at selected locations, some locations did show signs of leakage, and many locations were confirmed to be potential risk points for leakage.

[0254] Finally, an on-site survey was conducted on the selected monitoring points, and it was found that some of the points did show signs of leakage, and many points were confirmed to be potential risk points for leakage, which further verified the accuracy of the scheme of the present invention.

[0255] The present invention uses advanced data analysis technology to accurately identify abnormal pressure fluctuation patterns, thereby quickly locating possible new leakage points. Compared with the prior art, the present invention can adapt to large-scale, multi-level pipe network systems, effectively analyze leakage problems in complex pipe networks, reduce manual steps, and reduce labor costs.

Claims

1. A water supply network leakage analysis method based on water pressure baseline, characterized in that: The following steps are involved: Step 1: Hydraulic equipment layout and message analysis; The water pressure monitoring equipment in the analysis and study area is defined as monitoring points, and the original messages of each monitoring point are obtained and parsed in real time. The water pressure data in the original message is defined as message pressure, and the time data is defined as message time; n monitoring points are encoded to generate the corresponding monitoring point set {monitor}, the original message data of all monitoring points are obtained and recorded as the message data set {report_data}, and then the message pressure set {report_p} and the message time set {report_t} are stored in the data dictionary {report_data} in the form of {key:value}; {monitor}={monitor1,monitor2,...,monitor n }; {report_p}={report_p1,report_p2,...,report_p l }; {report_t}={report_t1,report_t2,...,report_t l }; In the above formula, monitor n Indicates the nth monitoring point, report_p l and report_t l Respectively represent the message pressure and message time of the th message data; Step 2: pre-process the water pressure data in the data dictionary obtained in step 1 to obtain the corresponding maximum half-range amplitude H of the stable water supply pressure and the water pressure 1 / 2 quantile M, which specifically includes the following contents: Step 2.1, data cleaning, including discarding error messages, removing invalid data and supplementing missing data, to obtain the cleaned data set {result_data}; Among them, the message pressure set after data cleaning is: {p′}={p′1,p′2,...,p′ k′ }; The message time set after data cleaning is: {t′} = {t′1, t′2, ..., t′ k′ }; k′ is the amount of data in the daily pressure set of each monitoring station, a total of r days, Step 2.2, extract the feature values ​​of the cleaned data set {result_data} to obtain the maximum half-range amplitude set {H} and 1 / 2 quantile set {M} of the daily stable water supply of n′ stable monitoring points; Among them, the maximum half-range fluctuation value H of the stable water supply of a certain stable monitoring point on a certain day is equal to half of its maximum pressure fluctuation value p_dif, and the maximum pressure fluctuation value p_dif is the difference between the maximum pressure and the minimum pressure of the stable monitoring point on that day; Among them, the 1 / 2 quantile M of a certain stable monitoring point on a certain day is the boundary value that divides the water pressure data of all monitoring points into two parts; Step 3: Calculate the pressure baseline based on the long series data, that is, calculate the weighted mean of the water pressure data of each stable monitoring point, which specifically includes the following contents: Calculate the weight value normal_ω of each stable monitoring point. First, standardize the 1 / 2 quantile M of each stable monitoring point to obtain the corresponding standardized Then standardize the obtained The value is calculated based on the stability weight ω, and then each weight ω is normalized to obtain the corresponding weight value normal_ω, forming the weight value set {normal_ω} of all stable monitoring points; For the n′th stable monitoring point monitor n′ The weight value on day r is normal_ω r ; Next, for the 1 / 2 quantile M and the weight value normal_ω of each stable monitoring point every day, the daily pressure baseline day_line is calculated by the weighted mean of the weight value, and the monthly pressure baseline month_line and the annual pressure baseline year_line are summarized in turn; Step 4: Perform water pressure anomaly analysis based on the pressure baseline; For each monitoring point, first calculate the corresponding pressure deviation p_d based on the measured pressure value p, the annual baseline year_line and the maximum half-range value H of a single monitoring point; then calculate the continuous duration of the pressure deviation p_d continuously greater than the pressure deviation limit p_d_limit, recorded as the continuous deviation duration t_d; and calculate the average pressure deviation p_d_avg within the continuous deviation duration t_d; finally, perform water pressure anomaly analysis on a certain monitoring point based on the pressure deviation limit p_d, the continuous deviation duration t_d and the average pressure deviation p_d_avg; Step 5: Add leakage status analysis to the area in real time.

2. The water supply network leakage analysis method based on water pressure data according to claim 1 is characterized in that: The specific method of discarding error messages during the data cleaning process in step 2.1 is: The error message identification rule is defined as rule, and the message discarding method is defined as dis(); after discarding the error message, the original data set {origin_data} is obtained. The original data set is to extract the monitoring point pressure p and time t within T hours from the processed perception data message, and form a message pressure set {p} and a message time set {t} containing k data in chronological order, and store them in the data dictionary {origin_data} in the form of {key: value}; {p}=dis({report_p}, rule); {t}=dis({report_t}, rule); The specific method for removing invalid data during the data cleaning process in step 2.1 is: First, calculate the upper quantile Q3, median Q2 and lower quantile Q1 of the water pressure data {p} in the original data set {origin_data}; and calculate the interquartile range IQR, IQR = Q3-Q1; Then, the upper limit and lower limit of the data are calculated based on IQR. Data exceeding the upper limit / lower limit are regarded as invalid data and are eliminated. Upperlimit = Q3 + 15 * IQR; Lowerlimit = Q1 + 15 * IQR; The specific method of supplementing missing data in the data cleaning process of step 2.1 is: The time interval [t0, t u ] is divided into u subintervals, each subinterval is [t i-1 , t i ], where i = 1, 2, ..., u; in each subinterval [t i-1 , t i ], construct a cubic polynomial S i (t), which has the form: S i (t)=a i +b i (t-t i )+c i (t-t i ) 2 +d i (t-t i ) 3 Among them, a i , b i 、c i and d i is the coefficient to be determined; At any time t∈[t i-1 , t i ], using the obtained cubic polynomial S i (t) is used to calculate the corresponding pressure value p′, and the calculated results are divided by day and stored in the result data set {result_data}, for a total of r days, where k′ is the amount of data in the daily pressure set of each monitoring station.

3. The water supply network leakage analysis method based on water pressure data according to claim 1 or 2, characterized in that: The detailed process of feature value extraction in step 2 is as follows: Step 1: Stabilize the maximum half-range amplitude of water supply H: The stable water supply state (ss) refers to the state in which the water pressure value at the monitoring point fluctuates within a certain range under normal working conditions of the water supply network, and the fluctuation amplitude is small, and there is no significant abnormal change. The fluctuation amplitude limit is defined as ss_limit. The maximum pressure fluctuation amplitude {p_dif} of each monitoring point is compared with the fluctuation amplitude limit. If it is less than the limit, it is in a stable water supply state. First, compare the daily pressure concentration p′ of each monitoring station, calculate the maximum pressure in the pressure concentration, form the maximum pressure set {p_max} and the maximum pressure set {p_min}, and store them in the pressure feature set {p_feature} in the form of {key:value}. The calculation formula is as follows: {p_max}={p_max1,...,p_max r } ={max(p′1,p′2,...,p′ k′ )1,...,max(p′1,p′2,...,p′ k′ ) r } {p_main}={p_main1,...,p_main r } ={min(p′1,p′2,...,p′ k′ )1,...,min(p′1,p′2,...,p′ k′ ) r } Then calculate the maximum pressure variation p_dif, where p_dif is the difference between the maximum pressure and the minimum pressure, and store all p_dif in the pressure feature set {p_feature}; {p_max}={p_dif1,...,p_dif r }={p_max1-p_min1,...,p_max r -p_min r } Next, the stable water supply status is judged, and the judgment rules are as follows: 1) When the maximum pressure variation value p_dif is less than or equal to ss_limit, it is judged to be in a stable water supply state. 2) When the maximum pressure variation value p_dif is greater than ss_limit, it is judged to be in a fluctuating water supply state. Eliminate the monitoring points with fluctuating water supply status ss=1, select the monitoring points with stable water supply status ss=0, and save them as the stable monitoring point set {stable_monitor}, the number is n′; Finally, the maximum half-range fluctuation value H of stable water supply is calculated based on the maximum fluctuation value of pressure. The maximum half-range fluctuation of daily stable water supply at all stable monitoring points is integrated into the maximum half-range fluctuation set {H}, which is expressed as:

4. The water supply network leakage analysis method based on water pressure data according to claim 3 is characterized in that: After obtaining the maximum half-range amplitude set {H}, the boundary value of the water pressure data of each monitoring point is calculated to be divided into two parts, that is, the 1 / 2 quantile M. The 1 / 2 quantile M reflects the average level of pressure at each monitoring point. The specific calculation method is: First, the daily pressure set {p′} of each monitoring point is sorted from small to large using the quick sort method to ensure that the sorted data is in ascending order; If k′ is an odd number, the 1 / 2 quantile M is the data point located in the middle after sorting; If k′ is an even number, the 1 / 2 quantile M is the average of the two middle data points after sorting; Then, the daily 1 / 2 quantiles M of all stable monitoring points are integrated into the 1 / 2 quantile set {M}; the calculation formula is as follows: When k′ is an odd number: When k′ is an even number:

5. The water supply network leakage analysis method based on water pressure data according to claim 1 is characterized in that: The specific method for calculating the weight of the monitoring point in step 3 is: First, the 1 / 2 quantile M of each monitoring point is standardized: the mean μ and standard deviation σ of the 1 / 2 quantile M of all monitoring points are calculated to obtain the standardized 1 / 2 quantile of each station. The 1 / 2 quantile Transformed to standard normal distribution data with mean 0 and standard deviation 1, so that the 1 / 2 quantiles of different sites It can be compared and divided on the same scale, and the calculation formula is as follows: Due to the standardized is a dimensionless value, indicating the degree of deviation of the original data M from the mean value μ. When , it means that the M of the site is higher than μ; when When , it means that the M of the site is lower than μ; when When , it means that the M of the site is equal to μ; Then, according to each monitoring point The weight ω of each monitoring point is calculated based on the stability of the monitoring point. The calculation formula is as follows: in, The constant C is an adjustment factor used to adjust the distribution range of the weight to avoid the denominator being 0 or a negative number; Next, to ensure that the sum of the weights of all sites is 1, the calculated weights are normalized to obtain the final weight value normal_ω, which is calculated as follows: The daily pressure baseline day_line is calculated as follows: Form a daily pressure baseline set {day_line} for each stable monitoring point every day, {day_line} = {day_line1, day_line2, ..., day_line r }; Summarize the daily pressure baseline values ​​day_line, calculate the monthly average pressure baseline value as the monthly pressure baseline month_line, and form the monthly pressure baseline set {month_line}. {month_line}={month_line1,month_line2,...,day_line 12 } The monthly pressure baseline values ​​month_line obtained each month are summarized and the average pressure baseline value for the whole year is calculated as the annual pressure baseline year_line.

6. The water supply network leakage analysis method based on water pressure data according to claim 1 is characterized in that: The pressure deviation p_d of a certain monitoring point in step 4 i The calculation formula is: p i is the measured pressure value of the monitoring point, H and year_line are the maximum half-range value and monthly pressure baseline corresponding to the monitoring point; The calculation formula for the average pressure deviation p_d_avg within the continuous deviation duration is: p_d_count is the number of pressure values ​​within the continuous deviation duration; Then, the abnormal pressure situation is judged, and the judgment rules are: 1) If the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is less than or equal to the continuous deviation duration limit t_d_limit, it is judged as a first-level abnormality and the pressure abnormality state p_state is set to 1; 2) When the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is greater than the continuous deviation duration limit t_d_limit and the pressure average deviation p_d_avg is less than or equal to the pressure deviation limit p_d_limit, it is judged as a secondary abnormality and the pressure abnormality state p_state is set to 2; 3) When the pressure deviation p_d is greater than the pressure deviation limit p_d_limit and the continuous deviation duration t_d is greater than the continuous deviation duration limit t_d_limit and the pressure average deviation p_d_avg is greater than the pressure deviation limit p_d_limit, it is judged as a level 3 abnormality and the pressure abnormality state p_state is set to 3; p_d_limit is the pressure deviation limit and t_d_limit is the continuous deviation time limit.

7. The water supply network leakage analysis method based on water pressure data according to claim 1 is characterized in that: The detailed method of performing real-time analysis on the newly added leakage status of the region in step 5 is as follows: If the pressure abnormal state p_state is 1, the newly added leakage point is judged as a newly added first-level leakage; If the pressure abnormal state p_state is 2, the newly added leakage point is judged as a newly added secondary leakage; If the pressure abnormal state p_state is 3, the newly added leakage point is determined to be a newly added third-level leakage.

Images