Pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection.

Abstract

The present invention discloses a method for detecting and evaluating the performance of a water supply network pipe rupture based on single-point and time-series anomaly detection, which includes the steps of (1) performing single-point anomaly detection on real-time monitoring data of the water supply network to distinguish between normal and abnormal working conditions of the water supply network, (2) performing time-series anomaly detection on the real-time monitoring data of the water supply network to distinguish between normal and abnormal working conditions of the water supply network, and (3) integrating the results of the single-point and time-series anomaly detection of the water supply network to accurately detect and distinguish between water supply network pipe ruptures and fault conditions in the monitoring system. The method of the present invention can accurately identify pipe ruptures and accurately distinguish between situations in which anomalies occur in various monitoring point data.

Description

[Technical Field]

[0001] The present invention relates to the field of urban water supply pipe burst detection technology, and more specifically to pipe burst detection and performance evaluation methods based on single-point and time-series anomaly detection. [Background technology]

[0002] Pipe ruptures are a major form of water loss in water supply systems, and despite their short duration, they result in significant water losses. Not only do pipe ruptures cause a great deal of waste of water resources, but the resulting pressure drop in the pipeline network can also affect the normal water supply. Furthermore, the intrusion of contaminants is more likely after a pipe rupture, potentially impacting the quality of drinking water. Pipe rupture detection is a crucial means of ensuring urban water safety by enabling timely responses to pipe rupture incidents and reducing the damage caused by such incidents.

[0003] The data-driven method is a new approach in the field of pipe rupture detection that treats pipe rupture detection as an anomaly detection problem, finding pipe ruptures by detecting anomalous features in real-time monitoring data. When detecting pipe ruptures, this method does not require the use of hardware equipment or pipe network models; it only requires anomaly detection using software on the pipe network monitoring data, offering advantages such as low cost, short time, and low labor intensity. At the same time, due to undesirable factors such as insufficient pipe rupture monitoring data and interference from dirty data, the false alarm rate for pipe rupture detection in this data-driven method is high at this stage.

[0004] Data-driven methods can be further divided into single-point anomaly detection-based methods and time-series anomaly detection-based methods. Single-point anomaly detection methods perform anomaly detection on a single value in the monitoring data, and if an anomaly is detected at multiple time points (time windows), a pipe rupture warning is issued. Such methods are widely used in the field of pipe rupture detection because they can find anomalies in the monitoring data in a timely manner and distinguish between normal and abnormal operating conditions in the piping network. However, such methods cannot accurately identify the various triggers for anomalies, and appropriate data preprocessing methods are necessary to remove anomalies from the monitoring data before pipe rupture detection. Under normal operating conditions, the monitoring data of the piping network follows a specific cycle pattern, and all monitoring data are normal values. When a pipe rupture occurs in the piping network, dirty data may appear at some monitoring points, or dirty data may appear at all monitoring points, altering the original cycle pattern of the monitoring data and leading to the occurrence of anomalies in the monitoring data. If the triggers for these anomalies cannot be accurately identified after they are detected, the false alarm rate for pipe rupture detection increases. Conversely, if no warning is issued after these anomalies are detected, the pipe rupture detection rate decreases.

[0005] Methods based on time-series anomaly detection can identify anomalous time series by considering the correlation between monitored data and analyzing the changes in differences between time series. Such methods can detect these changes. Such methods eliminate interference from unstable conditions on the results of pipe rupture detection. However, such methods are susceptible to noisy data, and even small amounts of noise can dominate the measure of similarity, potentially increasing the false alarm rate of such methods. [Overview of the project] [Problems that the invention aims to solve]

[0006] The present invention aims to provide a pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection, in order to compensate for the above shortcomings and solve the problems raised in the background art. [Means for solving the problem]

[0007] The present invention employs the following technical solutions to solve the above technical problems. A water pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection, comprising: (1) performing single-point anomaly detection on real-time monitoring data of a water pipe network to identify normal and abnormal operating conditions of the water pipe network; (2) performing time-series anomaly detection on real-time monitoring data of a water pipe network to identify normal and abnormal operating conditions of the water pipe network; and (3) integrating the results of single-point and time-series anomaly detection of the water pipe network to accurately detect and identify water pipe network ruptures and failure conditions in the monitoring system.

[0008] Preferably, step (1) includes, specifically, step (1.1) preparing the history of each monitoring point in the water supply network and real-time single-point monitoring data so as to perform single-point anomaly detection on the real-time monitoring data; step (1.2) using an unsupervised stacking integration algorithm to perform single-point anomaly detection on the real-time monitoring data of the water supply network and identify a single point anomaly in the real-time monitoring data; and step (1.3) using statistical theory to qualitatively detect a single value in the monitoring data of the water supply network and identify a single point anomaly in the real-time monitoring data.

[0009] Preferably, step (2) includes, specifically, the steps of: (2.1) preparing the history of each monitoring point in the water supply network and the time series of real-time monitoring data in order to detect time series anomalies in real-time monitoring data; (2.2) analyzing the differences between the time series of monitoring data from different monitoring points to determine anomalies between monitoring points; and (2.3) analyzing the differences between the time series of monitoring data from the same monitoring point to determine anomalies in the monitoring point itself.

[0010] Preferably, step (3) specifically includes: step (3.1) using a single-point anomaly detection method to perform single-point anomaly detection for various abnormal situations in the water supply network; step (3.2) using a time-series anomaly detection method to perform time-series anomaly detection for various abnormal situations in the water supply network; step (3.3) integrating the results of single-point and time-series anomaly detection to detect and identify various abnormal situations in the water supply network; and step (3.4) using a pipe burst detection method to detect and identify various abnormal situations in the water supply network and to evaluate the pipe burst detection performance of the provided method.

[0011] Preferably, step (3.1) includes specifically the following: When performing single-point anomaly detection for various abnormal situations in the water supply network using a single-point anomaly detection method, the following four situations are considered: (1) a pipe rupture occurs in the piping network; (2) dirty data appears at a single monitoring point; (3) dirty data appears at some monitoring points; (4) dirty data appears at all monitoring points. Each situation includes 25 sets of anomaly monitoring data, single-point anomaly detection is performed from the time the anomaly begins, and a decision is made on whether to issue a warning based on the anomaly detection results at five consecutive times.

[0012] Preferably, step (3.2) includes specifically the following: When performing time-series anomaly detection for various abnormal situations in the water supply network using a time-series anomaly detection method, the following four situations are considered: (1) a pipe rupture occurs in the piping network; (2) dirty data appears at a single monitoring point; (3) dirty data appears at some monitoring points; (4) dirty data appears at all monitoring points. Each situation includes 25 sets of anomaly monitoring data, and time-series anomaly detection is performed between monitoring points from the time the anomaly value begins, and a decision is made on whether to issue a warning based on the time-series anomaly detection results at five consecutive times.

[0013] Preferably, step (3.3) includes, specifically, a step of detecting whether there is a single point or time-series anomaly in the monitoring data (3.3.1), a step of identifying a situation in which anomalies appear in the monitoring data of some monitoring points (3.3.2), a step of identifying a pipe rupture and a situation in which dirty data appears at all monitoring points (3.3.3), and a step of evaluating or analyzing the pipe rupture detection performance of the method (3.3.4).

[0014] Preferably, step (3.3.1) is as follows: First, the normal operating conditions of the piping network and the situation in which abnormalities appear in the monitoring data are identified. If the presence of a single point outlier is detected in the single point monitoring data based on an unsupervised stacking integration algorithm or statistical theory, the time in which the single point outlier appears is marked, and single point outlier detection is continued for the monitoring data at the next time. If an abnormal time series is found in the time series monitoring data based on time series anomaly detection between monitoring points or time series anomaly detection within the monitoring point itself, the time in which the abnormal time series appeared is marked, and time series anomaly detection is continued for the monitoring data at the next time. Step (3.3.2) is as follows: If abnormalities appear in the single point and time series of monitoring data for some monitoring points, but not in the single point and time series of monitoring data for the remaining monitoring points, it is determined that dirty data appears at some monitoring points. Step (3.3.3) is as follows: If an anomaly is detected in the single-point and time-series monitoring data for all monitoring points, the increase and decrease in anomaly values ​​will be used to identify the situation in which a pipe rupture has occurred in the piping network and the situation in which dirty data appears at all monitoring points. If the single-point qualitative detection result of the monitoring data for all monitoring points is "decreasing", it indicates that a pipe rupture has occurred in the piping network. Conversely, if the single-point qualitative detection result of the monitoring data is "increasing", it indicates that dirty data has appeared at all monitoring points in the piping network. In step (3.3.4), the following several indicators are considered when evaluating the pipe rupture detection performance of the method: (1) detection accuracy (σ1); (2) anomaly identification rate (σ2); (3) anomaly detection rate (σ3).

[0015] Preferably, the detection accuracy rate is shown as follows. σ1 = N dn / N n In the formula, N n represents the total number of abnormal scenes, and N dn represents the quantity of detected abnormal scenes.

[0016] The abnormality discrimination rate is shown as follows. σ2 = N in / N n In the formula, N in represents the quantity of abnormally scenes accurately identified, and N n [[ID=2二十七]]represents the total number of abnormal scenes.

[0017] The abnormal detection rate is shown as follows. σ3 = t dn / t t In the formula, t dn and t t respectively represent the duration due to abnormal event detection and the actual duration.

[0018] Preferably, the step (3.4) specifically includes: when detecting and identifying various abnormal scenes of the water supply pipe network, removing the single-point abnormal value detection results to obtain the abnormal detection rate, detection accuracy rate, and abnormality discrimination rate of various abnormal scenes in step (3.4.1); when detecting and identifying various abnormal scenes of the water supply pipe network, removing the single-point value qualitative detection results to obtain the abnormal detection rate, detection accuracy rate, and abnormality discrimination rate of various abnormal scenes in step (3.4.2); when detecting and identifying various abnormal scenes of the water supply pipe network, removing the time-series abnormal detection results between monitoring points to obtain the abnormal detection rate, detection accuracy rate, and abnormality discrimination rate of various abnormal scenes in step (3.4.3); and when detecting and identifying various abnormal scenes of the water supply pipe network, removing the time-series abnormal detection results of the monitoring point itself to obtain the abnormal detection rate, detection accuracy rate, and abnormality discrimination rate of various abnormal scenes in step (3.4.4).

Advantages of the Invention

[0019] The beneficial effects of the present invention are as follows. The method of the present invention utilizes the history and real-time monitoring data of the water supply network to effectively detect and identify pipe ruptures and fault conditions at monitoring points in the water supply network. By performing single-point and time-series anomaly detection on the real-time monitoring data of the water supply network, it is possible to timely discover abnormal operating conditions in the water supply network and accurately distinguish between various abnormal situations based on the various warning conditions of the single-point and time-series anomaly detection results. In other words, the method of the present invention can accurately identify pipe ruptures and distinguish between abnormal conditions in various monitoring point data. [Brief explanation of the drawing]

[0020] [Figure 1] This is a flowchart of a method for detecting and evaluating the performance of water supply pipe network ruptures based on single-point and time-series anomaly detection. [Figure 2] This is a schematic diagram of the single-point value anomaly detection provided by the present invention. [Figure 3] This is a schematic diagram of single-point anomaly detection based on unsupervised stacking integration provided by the present invention: (ac) time series of monitored data; (de) comparison of real-time data and historical data; (fg) single-point anomaly detection. [Figure 4] This is a schematic diagram of single-point outlier detection based on statistical theory provided by the present invention: (ac) time series of monitoring data; (de) comparison of real-time data and historical data; (fg) single-point outlier detection. [Figure 5] This is a schematic diagram of time-series anomaly detection between monitoring points provided by the present invention: (a) monitoring data time series; (bc) comparison of real-time monitoring data time series and historical data time series; (d) monitoring data time series anomaly detection. [Figure 6] This is a flowchart for detecting time-series anomalies between monitoring points provided by the present invention. [Figure 7]This is a schematic diagram of the time-series anomaly detection of the monitoring point itself provided by the present invention: (a) time series of monitoring data; (bc) comparison of real-time monitoring data time series and historical monitoring data time series; (c) detection of anomaly in the monitoring data time series. [Figure 8] This is a flowchart for detecting time-series anomalies at the monitoring point itself, as provided by the present invention. [Figure 9] This is a schematic diagram of single-point anomaly detection after a pipe rupture occurs in a piping network in an embodiment of the present invention: (a) time-series curve of monitoring data; (b) single-point value of monitoring data. [Figure 10] This is the single-point anomaly detection result after dirty data appears at some monitoring points in an embodiment of the present invention: (a) time-series curve of monitoring data; (b) single-point value of monitoring data. [Figure 11] The single-point anomaly detection result after dirty data appears at all monitoring points in an embodiment of the present invention: (a) time-series curve of monitoring data; (b) single-point value of monitoring data. [Figure 12] The following are the time-series anomaly detection results after a pipe rupture occurred in the piping network in an embodiment of the present invention: (a) time-series curve of monitoring data; (b) time series between monitoring points; (c) time series of the monitoring point itself. [Figure 13] The following are the time-series anomaly detection results after dirty data appeared at some of the monitoring points in an embodiment of the present invention: (a) time-series curve of the monitoring data; (b) time series between monitoring points; (c) time series of the monitoring point itself. [Figure 14] The time-series anomaly detection results after dirty data appears at all monitoring points in the embodiment of the present invention are as follows: (a) time-series curve of the monitoring data; (b) time series between monitoring points; (c) time series of the monitoring point itself. [Figure 15] This is a schematic diagram of the pressure monitoring point distribution in the Net 3 piping network model of the present invention. [Figure 16] This describes the experimental conditions and data preparation techniques used in embodiments of the present invention. [Figure 17] This is a schematic diagram of the monitoring data.csv file in an embodiment of the present invention. [Figure 18]The results of single-point anomaly detection based on unsupervised stacking integration in an embodiment of the present invention are as follows: (a) a pipe rupture occurs in the piping network; (b) dirty data appears at a single monitoring point; (c) dirty data appears at some monitoring points; (d) dirty data appears at all monitoring points. [Figure 19] The results of detecting a single-point anomaly based on statistical theory in an embodiment of the present invention are as follows: (a) A pipe rupture occurs in the piping network; (b) Dirty data appears at a single monitoring point; (c) Dirty data appears at some monitoring points; (d) Dirty data appears at all monitoring points. [Figure 20] The following are the time-series anomaly detection results between monitoring points in an embodiment of the present invention: (a) a pipe rupture occurs in the piping network; (b) dirty data appears at a single monitoring point; (c) dirty data appears at some monitoring points; (d) dirty data appears at all monitoring points. [Figure 21] The time-series anomaly detection results for the monitoring point itself in an embodiment of the present invention are as follows: (a) a pipe rupture occurs in the piping network; (b) dirty data appears at a single monitoring point; (c) dirty data appears at some monitoring points; (d) dirty data appears at all monitoring points. [Figure 22] These are the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations in the embodiments of the present invention. [Figure 23] This figure shows the removal of the single-point anomaly detection result in an embodiment of the present invention. [Figure 24] This figure shows the removal of the single-point qualitative detection result in an embodiment of the present invention. [Figure 25] This figure shows the removal of time-series anomaly detection results between monitoring points in an embodiment of the present invention. [Figure 26] This figure shows the removal of the time-series anomaly detection results for the monitoring point itself in an embodiment of the present invention. [Modes for carrying out the invention]

[0021] Next, the present invention will be described in more detail with reference to the drawings and specific embodiments.

[0022] (Example 1) As shown in Figure 1, a water pipe burst detection and performance evaluation method based on single-point and time-series anomaly detection includes the steps of: (1) performing single-point anomaly detection on real-time monitoring data of the water pipe network to identify normal and abnormal operating conditions of the water pipe network; (2) performing time-series anomaly detection on real-time monitoring data of the water pipe network to identify normal and abnormal operating conditions of the water pipe network; and (3) integrating the results of single-point and time-series anomaly detection of the water pipe network to accurately detect and identify water pipe network bursts and failure conditions in the monitoring system.

[0023] Furthermore, step (1) specifically includes the steps of: (1.1) preparing the history of each monitoring point in the water supply network and real-time single-point monitoring data so as to perform single-point anomaly detection on the real-time monitoring data; (1.2) using an unsupervised stacking integration algorithm to perform single-point anomaly detection on the real-time monitoring data of the water supply network and to identify a single point anomaly in the real-time monitoring data; and (1.3) using statistical theory to qualitatively detect a single value in the monitoring data of the water supply network and to identify a single point anomaly in the real-time monitoring data.

[0024] Furthermore, step (1.1) specifically includes the following: x indicates monitoring data for a monitoring point, x1 indicates monitoring data for monitoring point 1, and x1 indicates real-time monitoring data. 0 As shown, the historical monitoring data is x1 i (i=1,2,···,n d ) is shown, and the monitoring data for monitoring point 1 at time m is shown as x1(m). Single-point anomaly detection is performed using real-time monitoring data of the monitoring point and past n d The real-time monitoring data is compared with the historical monitoring data for the day to determine if it is an abnormal value. For example, if it is necessary to determine whether the real-time monitoring data at time m of monitoring point 1 is an abnormal value, then the real-time monitoring data x1 0 (m) and past n dDaily time m historical monitoring data x1 i (m)(i=1,2,···,n d ) is compared with and various anomaly detection methods are used to x1 0 Determine whether (m) is an outlier.

[0025] Before performing single-point anomaly detection on the monitoring data x for each monitoring point, it is necessary to prepare real-time monitoring data and historical monitoring data for each monitoring point, as shown below.

[0026]

number

[0027] In the formula, x 0 (i)(i=1,2,···,m) shows real-time monitoring data for the monitoring point, x 0 (m) represents the mth real-time monitoring data, where m is the total number of daily monitoring data at each monitoring point. j (i)(i=1,2,···,m;j=1,2,···,d) shows the historical monitoring data for the monitoring point, x j (m) is the mth historical monitoring data from the past j days at the monitoring point, and d is the total number of days for the historical monitoring data. Each row of matrix X shows the real-time and historical monitoring data at a given time at the monitoring point, for example, X(i) = [x 0 (i), x 1 (i), ···,x j (i), ···,x d (i)) shows the real-time and historical monitoring data for the i-th monitoring point. Each row of matrix X shows the monitoring data for each time of day at the monitoring point, for example, X(j)=[x j (1), x j (2), ···,x j (i), ···,x j (m)] T This shows the historical monitoring data for each time point over the past j days at the monitoring point.

[0028] After collecting real-time monitoring data from each monitoring point, it is compared with historical monitoring data for the corresponding time over the past d days to check if the real-time monitoring data is an anomaly. When an anomaly detection is performed on the mth real-time monitoring data, the mth real-time monitoring data and the mth historical monitoring data from the past d days are prepared for all monitoring points.

number

[0029] In the formula, x k 0 (m)(k=1,2,···,n) represents the m-th real-time monitoring data for monitoring point k, and n is the number of monitoring points distributed within the piping network. k j (m)(k=1,2,···,n;j=1,2,···,d) represents the m-th historical monitoring data for monitoring point k over the past j days. Each row of matrix X(m) represents the m-th historical and real-time monitoring data for a given monitoring point, for example, x k (m) = [x k 0 (m), x k 1 (m),···,x k j (m),···,x k d (m)](j=1,2,···,d) shows the m-th history and real-time monitoring data for monitoring point k. Each row of matrix X(m) shows the m-th monitoring data for a day at all monitoring points in the piping network, for example, x j (m) = [x1 j (m), x2 j (m),···,x k j (m),···,x n j (m)] T (k=1,2,···,n) represents the m-th historical monitoring data for the past j days at the monitoring point.

[0030] Furthermore, step (1.2) specifically includes the following: For single-point anomaly detection in monitoring data, the present invention employs a machine learning algorithm based on unsupervised stacking integration. This algorithm was first used for fault diagnosis in power systems and can effectively detect and accurately identify power system faults and anomalies in monitoring data using real-time monitoring data from SCADA systems. The algorithm is divided into three layers: (1) Layer 1 is an isolation forest algorithm; (2) Layer 2 is a K-means clustering algorithm and a local outlier factor algorithm; (3) Layer 3 is an integration of the K-means clustering algorithm and the local outlier factor algorithm.

[0031] When performing single-point anomaly detection on monitoring data, each row of X(m) is input into an unsupervised stacking integration algorithm as data to be detected. For monitoring data from n monitoring points, it is necessary to perform n single-point anomaly detections at each time step. For example, taking the monitoring data for monitoring point 1 as an example, x1(m) = [x1 0 (m), x1 1 (m),···,x1 d The total d+1 data to be detected (m) is input into an unsupervised stacking integration algorithm to obtain the probability that each monitored data is an anomaly. x1 0If (m) has the highest probability of being an outlier, the real-time monitoring data is marked as an outlier. Otherwise, the real-time monitoring data is normal and is not marked. Step (1.2) mainly includes the steps of inputting the data to be detected into a separation forest algorithm to obtain an anomaly score for each data (1.2.1), using each anomaly score obtained by the separation forest algorithm in Layer 1 as input to a K-means clustering algorithm in Layer 2, performing clustering on each anomaly score in the K-means clustering algorithm to obtain binary data 0 (normal) and 1 (abnormal) (1.2.2), using each anomaly score obtained by the separation forest algorithm in Layer 1 as input to a local outlier factor algorithm in Layer 2, and obtaining outlier factors for each anomaly score using the local outlier factor algorithm (1.2.3), and obtaining the probability that each anomaly score is an outlier, i.e., the probability that each data to be detected is an outlier, based on the output results of the K-means clustering algorithm and the local outlier factor algorithm (1.2.4).

[0032] Furthermore, step (1.3) specifically includes the following: To avoid identifying instances where dirty data appears at a monitoring point as a pipe network rupture, the present invention provides a single-point anomaly detection method based on statistical theory. Since pipe rupture incidents usually lead to a rapid decrease in monitoring data, the abnormal monitoring value after a pipe rupture occurs must be smaller than the normal monitoring value. If the abnormal monitoring value is larger than the normal monitoring value, it can be determined that no pipe rupture incident has occurred in the pipe network and that dirty data has appeared at the monitoring point. In addition, to avoid identifying normal monitoring values ​​due to fluctuations in water demand as abnormal monitoring values, real-time monitoring data should only be marked as an anomaly after it exceeds a certain range.

[0033] For monitoring data x of a single monitoring point, historical monitoring data x i The qualitative threshold [ζ] is determined by (i=1,2,···,d) - (x), ζ +Get (x). ζ - (x) = μ(x) i )-m k σ(x i ) ζ + (x) = μ(x) i )+m k σ(x i )

[0034] x 0 ζ - (x) and ζ + When compared with (x), the situation can be divided into three cases: (1) x 0 < ζ - (x) indicates that the real-time monitoring data is an anomaly and shows a "decrease" trend compared to the historical monitoring data, and the anomaly detection result is -1; (2)ζ - (x)≦x 0 ≦ζ + (x) indicates that the real-time monitoring data is normal and the anomaly detection result is 0; (3)ζ + (x) <x 0 In this case, the real-time monitoring data is an anomaly and shows an increasing trend compared to the historical monitoring data, and the anomaly detection result is indicated as 1.

[0035] Furthermore, step (2) specifically includes the steps of: (2.1) preparing the history of each monitoring point in the water supply network and the time series of real-time monitoring data in order to detect time series anomalies in real-time monitoring data; (2.2) analyzing the differences between the time series of monitoring data from different monitoring points to determine anomalies between monitoring points; and (2.3) analyzing the differences between the time series of monitoring data from the same monitoring point to determine anomalies in the monitoring point itself.

[0036] Furthermore, step (2.1) specifically includes the following: The monitoring data time series refers to a vector containing multiple monitoring values, for example, the monitoring data time series for monitoring point 1 is S1 = [x1 m-l+1 ,x1 m-l+2 ,···,x1 m This can be shown as ]. x1m represents the monitoring data at time m at monitoring point 1, l is the length of the monitoring data time series. That is, the monitoring data time series S1 contains l monitoring data, and x1 m-l+1 represents the monitoring data at time m - l + 1 at monitoring point 1. If each monitoring data in S1 is a normal value, that is, if the change in the shape of the monitoring data time series S1 is not large, then S1 is regarded as the normal time series of the monitoring data. Otherwise, if an abnormality appears in the monitoring data in S1, that is, if the shape of the monitoring data time series S1 changes (compared with the normal time series of the monitoring data), then S1 is regarded as the abnormal time series of the monitoring data. To perform the anomaly detection of the monitoring data time series, the present invention uses a method based on the detection of the difference between monitoring data time series. This method uses the distance between two monitoring data time series to detect the abnormal time series of the monitoring data. For two monitoring data time series S1 and S2, the distance of the monitoring data time series is shown as follows.

Equation

[0037] If each monitoring data in monitoring data time series S1 and S2 is constant, then d(S1,S2) is also constant, i.e., d(S1,S2) remains constant and unchanging. If the monitoring data in monitoring data time series S1 and S2 changes, it causes a change in d(S1,S2). Conversely, if d(S1,S2) changes, it is understood that the monitoring data in monitoring data time series S1 or S2 has changed, i.e., an anomaly exists in the monitoring data. Therefore, the present invention finds anomalies in monitoring data time series by detecting distance changes in the monitoring data time series. Depending on the source of the monitoring data in the data time series to be detected, it can be divided into (1) detection of time series anomalies between monitoring points and (2) detection of time series anomalies within the monitoring point itself. Before detecting anomalies in the monitoring data time series, it is necessary to prepare the monitoring data time series for each monitoring point. For a single monitoring point 1, it is shown that the m-th monitoring data time series for the past j days is as follows. S1 j (m) = [x1 j (m-l+1),x1 j (m-l+2),···x1 j (m-1),x1 j (m)] In the formula, l represents the length of the monitoring data time series, i.e., the monitoring data time series contains l monitoring data, the starting monitoring data for the mth monitoring data time series at monitoring point 1 is x1(m-l+1), and the last monitoring data is x1(m).

[0038] The time series of monitoring data is shown as follows:

number

[0039] In the formula, S 0 (i)(i=1,2,···,m) shows the time series of real-time monitoring data for the monitoring point, S 0 (m) represents the time series of the mth real-time monitoring data at the monitoring point, and m is the total number of daily monitoring data at each monitoring point. j (i)(i=1,2,···,m;i=1,2,···,d) shows the time series of historical monitoring data for the monitoring point, S j(m) represents the m-th historical monitoring data time series for the past j days at the monitoring point, and d is the total number of days for the historical monitoring data. Each row of matrix S represents the real-time and historical monitoring data time series at a given time at the monitoring point, for example, S(i) = [S 0 (i), S 1 (i), ···,S j (i), ···,S d (i)) represents the real-time and historical monitoring data time series for the i-th monitoring point. Each column of matrix S represents the monitoring data time series for each time on a given day at the monitoring point, for example, S(j) = [S j (1), S j (2), ..., S j (i), ···,S j (m)] T This shows the time series of historical monitoring data for each time point over the past j days at the monitoring point.

[0040] When performing anomaly detection on the mth real-time monitoring data time series, the mth real-time monitoring data time series and the mth historical monitoring data time series for the past d days are prepared and shown as follows.

number

[0041] In the formula, S k 0 (m)(k=1,2,···,n) represents the time series of the mth real-time monitoring data at monitoring point k, and n is the number of monitoring points distributed within the piping network. k j (m)(k=1,2,···,n;j=1,2,···,d) represents the m-th historical monitoring data for the past j days of monitoring point k. Each row of matrix S(m) represents the m-th historical and real-time monitoring data for a given monitoring point, for example, S k (m) = [S k 0 (m), S k 1 (m),···,S k j (m),···,S kd (m)](j=1,2,···,d) shows the m-th history and real-time monitoring data for monitoring point k. Each column of matrix S(m) shows the time series of the m-th monitoring data for the day at the monitoring point, for example, S j (m) = [S1 j (m), S2 j (m),···,S k j (m),···,S n j (m)] T (k=1,2,···,n) represents the time series of the mth monitoring data for the past j days at the monitoring point.

[0042] Furthermore, step (2.2) specifically includes the following:

[0043] When the time series of monitoring data to be detected originates from different monitoring points, this is called inter-point time series anomaly detection. Under normal operating conditions, monitoring data from different monitoring points in a piping network follows the same rules; that is, the distance between the time series of monitoring data from different monitoring points does not change significantly. If an anomaly appears in the monitoring data from a certain monitoring point, it causes a change in the shape of the time series of monitoring data from that point, and therefore causes a change in the distance between the time series of monitoring data from different monitoring points. Through inter-point time series anomaly detection, such changes can be detected, and thus anomalies in the monitoring data can be detected.

[0044] To detect whether the time-series distance between two monitoring points has changed, row vectors corresponding to the monitoring data time series of the two monitoring points are extracted, and then the distance between the real-time and historical monitoring data time series at the two monitoring points is calculated and obtained. Time-series anomaly detection between monitoring points is divided into three steps: (1) the distance d of the historical monitoring data time series m i (1) Calculate and obtain (S1,S2)(i=1,2,···,d); (2) Distance d of the real-time monitoring data time series m 0 (S1, S2) is calculated and obtained; (3)d m 0 (S1,S2) and dm i Comparing with (S1,S2), d m 0 Check if (S1,S2) are outliers.

[0045] Distance d of historical monitoring data time series m i (S1,S2) is the distance between the time series of monitoring data at monitoring points 1 and 2 at time m over the past d days, and the distance d of the previous real-time monitoring data time series. m 0 (S1,S2) can be exported directly. Therefore, when performing anomaly detection on the time series between monitoring points, the distance d of the real-time monitoring data time series is used. m 0 All that is needed is to calculate and obtain (S1,S2). The distance between monitoring points 1 and 2 in the time series of monitoring data at time m is as follows: d m (S1,S2)=[d m 0 (S1,S2),d m 2 (S1,S2),···,d m d (S1, S2)] In the formula, d m 0 (S1,S2) indicates the distance between the time series of real-time monitoring data, and d m i (S1,S2)(i=1,2,···,d) represents the distance between the time series of the historical monitoring data.

[0046] The determination threshold ξ1(S1,S2) is calculated and obtained by the distance of the historical monitoring data time series, and ξ1(S1,S2) is d m i It is calculated and obtained using the mean and variance of (S1,S2). ξ1(S1,S2)=μ(d m i (S1,S2))+m S1,S2 σ(d m i (S1, S2)) In the formula, μ(d m i(S1,S2)) and σ(d m i (S1, S2)) are the distance d from the time series of the monitored data, respectively. m i The mean and variance of (S1,S2)(i=1,2,···,d) are shown.

[0047] After obtaining the decision threshold ξ1(S1,S2), it is used to determine the distance d of the monitoring data time series. m 0 Compare (S1,S2) and determine the distance d between the time series of monitoring data at monitoring points 1 and 2. m 0 Determine whether (S1,S2) is an outlier. m 0 If (S1,S2)≦ξ1(S1,S2), it indicates that there are no abnormalities in the monitoring data for monitoring points 1 and 2 at the current time, and the detection result "Normal" is output and shown as 0. Otherwise, the detection result "Abnormal" is output and shown as 1.

[0048] Furthermore, step (2.3) specifically includes the following: When the time series of monitoring data to be detected originates from the same monitoring point, this is called time series anomaly detection at the monitoring point itself. Under normal operating conditions, the monitoring data at each monitoring point in the piping network follows the same rules, i.e., the distance of the time series of monitoring data at the same monitoring point does not change much. When an anomaly appears in the monitoring data at a certain monitoring point, it causes a change in the shape of the time series of monitoring data at that point, and therefore the distance of the time series of monitoring data at the monitoring point itself also changes.

[0049] When detecting anomalies in the time series of the monitoring point itself, it is necessary to calculate, obtain, and compare the distance between the real-time and historical monitoring data time series, and the distance between the historical and historical monitoring data time series. Anomaly detection in the time series of the monitoring point itself can be mainly divided into three steps: (1) the distance between the historical and historical monitoring data time series d m i,j (i≠j; i,j=0,1,···,d) is calculated and obtained; (2) the distance d between the real-time and historical monitoring data time series. m 0,j Calculate and obtain (j=1,2,···,d); (3)d m0,j and d m i,j Comparing with, d m 0,j Check if there are any abnormal values.

[0050] In fact, the distance d between the history and the historical monitoring data time series m i,j This was already calculated in the previous time-series anomaly detection of the monitoring point itself. Each time anomaly detection is performed, d of d m 0,j We only need to calculate the minimum value min{d m 0,j}:min{d m 0,j}=min{d m 0,1 d m 0,2 ,···,d m 0,d Obtain the distance d of the mth monitoring data time series over the past d days. m i,j The threshold ξ2(d) is determined by m i,j ):ξ2(d m i,j ) = μ(d m i,j )+mσ(d m i,j ) obtain, and in the formula, μ(d m i,j ) and σ(d m i,j ) are the distance d of the time series of the monitoring data, respectively. m i,j These are the mean and variance of (i≠j; i,j=0,1,···,d).

[0051] Decision threshold ξ²(d m i,j After obtaining the value of the time series distance of the monitoring point itself, min{d m 0,j Compared to}, min{d m 0,j Determine whether} is an outlier. min{d m 0,j}≦ξ2(d mi,j In the case of ( ), it indicates that there are no abnormalities in the time series of monitoring data at the current time of the monitoring point, and outputs a detection result of 0; otherwise, outputs a detection result of 1.

[0052] Furthermore, step (3) specifically includes the following: Step (3.1), single-point anomaly detection is performed for various abnormal situations in the water supply network using the provided single-point anomaly detection method. The provided method is tested using various abnormal situations to determine the detection effectiveness of the unsupervised classification-based single-point anomaly detection method for various abnormal situations, mainly considering the following four situations: (1) a pipe rupture occurs in the piping network; (2) dirty data appears at a single monitoring point; (3) dirty data appears at some monitoring points; (4) dirty data appears at all monitoring points. Each situation includes 25 sets of abnormal monitoring data, single-point anomaly detection is performed from the time the abnormal value begins, and it is determined whether to issue a warning based on the abnormal value detection results at five consecutive times.

[0053] Step (3.2) is to perform time-series anomaly detection on various abnormal situations in the water supply network using the provided time-series anomaly detection method. To determine the detection effectiveness of the time-series anomaly detection method on various abnormal situations, the provided method is tested using various abnormal situations, mainly considering the following four situations: (1) a pipe rupture occurs in the piping network; (2) dirty data appears at a single monitoring point; (3) dirty data appears at some monitoring points; (4) dirty data appears at all monitoring points. Each situation includes 25 sets of anomaly monitoring data, and single-point anomaly detection is performed from the time the anomaly value begins, and it is determined whether to issue a warning based on the anomaly detection results at five consecutive times.

[0054] Step (3.3) integrates single-point and time-series anomaly detection results to detect and identify various abnormal situations in the water supply network. The integration of single-point and time-series anomaly detection results aims to timely detect abnormal values ​​in real-time monitoring data and accurately identify situations where pipe ruptures occur in the piping network or dirty data appears at monitoring points. Anomaly detection is performed on the monitoring data for each monitoring point using a sliding time window, completing single-point and time-series anomaly detection for one time point at a time. If a single-point or time-series anomaly is detected at a monitoring point, the time corresponding to the single-point or time-series anomaly is marked, and the abnormal value is replaced using a data cleansing method. When the marked time window reaches a certain length, an anomaly warning is issued, and various abnormal situations are identified. This is mainly divided into the following steps.

[0055] Step (3.3.1) detects whether a single point or time-series anomaly exists in the monitoring data. First, the normal operating conditions of the piping network and the conditions under which anomalies appear in the monitoring data are identified. For single-point monitoring data, if the presence of a single-point anomaly is detected based on an unsupervised stacking integration algorithm or statistical theory, the time at which the single-point anomaly appears is marked, and single-point anomaly detection is subsequently performed on the monitoring data for the next time point. For time-series monitoring data, if an abnormal time series is found based on time-series anomaly detection between monitoring points or time-series anomaly detection within the monitoring point itself, the time at which the abnormal time series appeared is marked, and time-series anomaly detection is subsequently performed on the monitoring data for the next time point.

[0056] Step (3.3.2) identifies a situation in which anomalies appear in the monitoring data of some monitoring points. If anomalies appear in the single point and time series of monitoring data of some monitoring points, but no anomalies appear in the single point and time series of monitoring data of the remaining monitoring points, it is determined that dirty data appears at some monitoring points. For example, if an anomaly appears in the monitoring data of monitoring point 1, but there are no anomalies in the monitoring data of monitoring points 2 and 3, a single point anomaly can be detected in the monitoring data of monitoring point 1, and a continuous anomaly warning situation occurs in the time series of monitoring point 1 itself, while there are no warning situations for monitoring points 2 and 3. The time series distance d(S1,S2) of the monitoring data of monitoring points 1 and 2 and the time series distance d(S1,S3) of the monitoring data of monitoring points 1 and 3 are both abnormal values, while the time series distance d(S2,S3) of the monitoring data of monitoring points 2 and 3 is a normal value.

[0057] Step (3.3.3) identifies the conditions for a pipe rupture and the appearance of dirty data at all monitoring points. If an anomaly is detected in the single point and time series of monitoring data at all monitoring points, the conditions for a pipe rupture in the piping network and the appearance of dirty data at all monitoring points are identified through the increase and decrease of the anomaly values. If the single point qualitative detection result of the monitoring data at all monitoring points is "decreasing", it indicates that a pipe rupture has occurred in the piping network. Conversely, if the single point qualitative detection result of a certain monitoring data is "increasing", it indicates that dirty data has appeared at all monitoring points in the piping network.

[0058] Step (3.3.4) evaluates and analyzes the pipe rupture detection performance of the method. The provided method aims to detect single-point and time-series anomalies in real-time monitoring data, and by combining the single-point and time-series anomaly detection results, it identifies situations in which pipe ruptures occur in the piping network and situations in which dirty data appears at monitoring points. Therefore, when evaluating the method, the following several indicators are mainly considered: (1) detection accuracy (σ1); (2) anomaly identification rate (σ2); (3) anomaly detection rate (σ3). Detection accuracy refers to the probability that various anomaly situations are detected, that is, identifying and warning about single-point or time-series anomalies in an anomaly situation. Detection accuracy is shown as follows. σ1=N dn / N n In the formula, N n This indicates the total number of abnormal situations, N dn This indicates the number of anomalous scenes detected.

[0059] The anomaly identification rate refers to the probability that various anomaly scenarios are accurately identified. For example, in 10 types of anomaly scenarios, the number of pipe ruptures occurring in the piping network is 8, and the number of detected pipe rupture incidents is 4. Therefore, the anomaly identification rate for pipe ruptures in the piping network is 4 / 8 = 50%. The anomaly identification rate is shown below. σ² = N in / N n In the formula, N in This indicates the number of accurately identified anomalous scenes, N n This indicates the total number of abnormal situations.

[0060] The anomaly detection rate is the ratio of the duration of various detected anomalies to the actual duration of the anomalies. For example, if the duration of dirty data appearing at a certain monitoring point is 1 hour, and the detected duration of dirty data appearing at that monitoring point is 0.5 hours, the anomaly detection rate is 0.5 / 1 = 50%. The anomaly detection rate is expressed as follows: σ³=t dn / t t In the formula, t dn and t t These represent the duration of the detected abnormal event and the actual duration, respectively. Clearly, the higher σ1, σ2, and σ3 are, the better.

[0061] Step (3.4) Detect and identify various abnormal situations in the water supply network using the pipe burst detection method and evaluate the pipe burst detection performance of the provided method. Step (3.4.1) When detecting and identifying various abnormal situations in the water supply network, remove single-point anomaly detection results and obtain the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations. Step (3.4.2) When detecting and identifying various abnormal situations in the water supply network, remove single-point qualitative detection results and obtain the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations. Step (3.4.3) When detecting and identifying various abnormal situations in the water supply network, remove time-series anomaly detection results between monitoring points and obtain the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations. Step (3.4.4) When detecting and identifying various abnormal situations in the water supply network, remove time-series anomaly detection results for the monitoring point itself and obtain the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations.

[0062] In the process described above, various anomalous situations are detected using single-point anomaly detection methods and time-series anomaly detection methods, respectively. Based on the single-point and time-series anomaly detection results, the final anomaly detection results for various situations are obtained by integrating them and using four integration policies: i. Remove single-point anomaly detection results. ii. Remove single-point qualitative detection results. iii. Remove time-series anomaly detection results between monitoring points. iv. Remove time-series anomaly detection results for the monitoring point itself. The final anomaly detection results under various situations are obtained and saved to anomaly detection results.csv. Then, performance indicators of the provided method under various anomalous situations, namely detection accuracy, anomaly identification rate, and anomaly detection rate, are calculated and obtained and saved to anomaly evaluation results.csv.

[0063] Specific examples are as follows:

[0064] Embodiments of the present invention provide a method for detecting pipe ruptures in a water supply network and evaluating its performance based on single-point and time-series anomaly detection. This method utilizes the history and real-time monitoring data of the water supply network to effectively detect and identify pipe ruptures and failure conditions at monitoring points in the water supply network. By performing single-point and time-series anomaly detection on the real-time monitoring data of the water supply network, it is possible to discover abnormal operating conditions in the water supply network in a timely manner and to accurately distinguish between various abnormal situations based on the various warning conditions of the single-point and time-series anomaly detection results. The method specifically includes the following steps.

[0065] Step (1) performs single-point anomaly detection on real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. Step (1) specifically includes the following:

[0066] Step (1.1) A hydraulic simulation is performed on the water supply network to perform single-point anomaly detection on abnormal monitoring data, and pressure monitoring data is obtained under various abnormal conditions of the water supply network. Figure 2 shows a schematic diagram of single-point anomaly detection. The present invention mainly considers three types of abnormal conditions: (a) a pipe rupture occurs in the water supply network; (b) a failure occurs at some monitoring points in the water supply network; (c) a failure occurs at all monitoring points in the water supply network. Step (1.2) Single-point anomaly detection is performed on pressure monitoring data under various abnormal operating conditions of the water supply network based on the provided unsupervised stacking integration algorithm.

[0067] Figure 3 shows a schematic diagram of single-point anomaly detection based on an unsupervised stacking integration algorithm. The figure includes real-time and historical monitoring data for two monitoring points, labeled Monitoring Point 1 and Monitoring Point 2, respectively. Figures 3(a) and 3(b) show the historical monitoring data for Monitoring Points 1 and 2, and Figure 3(c) shows the real-time monitoring data for Monitoring Points 1 and 2, both containing 15 days of historical monitoring data, i.e., d=15. The monitoring data is collected every 15 minutes, with the first monitoring data collected at 0:15 and the second at 0:30, and the total number of daily monitoring data at each monitoring point is 96, i.e., m=96.

[0068] As shown in Figure 3(a), the top 95 real-time monitoring data for monitoring points 1 and 2 are all normal values, and there is no significant difference compared to the historical monitoring data in Figures 3(b) and 3(c). As shown in Figure 3(a), the top 96 real-time monitoring data (x1 0 (96) and x2 0 (96)) is history monitoring data (x1 i (96) and x2 i This is clearly an increase compared to (96)(i=1,2,···,15). As shown in Figure 3(d), monitoring point 1 is the 96th historical monitoring data x1 over the past 15 days. i (96) is distributed entirely within the yellow area (i.e., the normal range), and is the 96th real-time monitoring data x1 of monitoring point 1. 0 (96) is distributed above the yellow area and clearly deviates from the normal range. As shown in Figure 3(e), monitoring point 2 is the 96th historical monitoring data x2 over the past 15 days. i (96) is distributed entirely within the yellow area (i.e., the normal range), and is the 96th real-time monitoring data x2 of monitoring point 2. 0 (96) is distributed above the yellow area and clearly deviates from the normal range.

[0069] As shown in Figure 3(f), the mth real-time monitoring data x of monitoring point n n 0 If you need to detect whether (m) is an outlier, then xn 0 (m) and the mth historical monitoring data x from the past d days. n i (m)(i=1,2,···,d) A total of d+1 values ​​are input into the unsupervised stacking integration algorithm, and each x n Get the probability that (m) is an outlier. n 0 If (m) has the highest probability of being an outlier, mark it as an outlier; otherwise, x n 0 (m) is indicated as a normal value and is not marked.

[0070] Step (1.3) uses statistical theory to perform qualitative detection on single values ​​in the water supply network monitoring data and identify single-point anomalies in the real-time monitoring data. Figure 4 shows a schematic diagram of single-point anomaly detection based on statistical theory. The figure includes real-time and historical monitoring data for two monitoring points, indicated as monitoring point 1 and monitoring point 2, respectively. Figures 4(a) and 4(b) show the historical monitoring data for monitoring points 1 and 2, and Figure 4(c) shows the real-time monitoring data for monitoring points 1 and 2, both including 15 days of historical monitoring data, i.e., d=15. The monitoring data is collected every 15 minutes, with the first monitoring data collected at 0:15 and the second monitoring data collected at 0:30, and the total number of daily monitoring data at each monitoring point is 96, i.e., m=96.

[0071] As shown in Figure 4(a), the top 95 real-time monitoring data for monitoring points 1 and 2 are all normal values, and the historical monitoring data (x1 i (96) and x2 i There is no significant difference compared to (96)(i=1,2,···,15). The 96th monitoring data point x1 of monitoring point 1. 0 (96) is historical monitoring data x1 i (96) (i=1,2,···,15) is clearly lower, and the 96th monitoring data x2 of monitoring point 2 0 (96) is history monitoring data x2 i This is clearly an increase compared to (96)(i=1,2,···,15).

[0072] As shown in Figures 4(d) and 4(e), the light gray area indicates the normal range of the monitored data, and the dark gray line indicates the average value of the 15 monitored data points, μ+m k σ and μ-m k σ represents the upper and lower limits of the range, respectively. As shown in Figure 4(d), the 96th historical monitoring data for monitoring point 1 is distributed entirely within the light gray area, and the 96th real-time monitoring data is located below the light gray area, meaning that the real-time monitoring data is smaller than the historical monitoring data. As shown in Figure 4(e), the 96th historical monitoring data for monitoring point 2 is distributed entirely within the light gray area, and the 96th real-time monitoring data is located above the light gray area, meaning that the monitoring data is increasing. Therefore, based on statistical theory, a single value of the monitoring data can be detected and divided into three types: (1) an outlier (increase); (2) a normal value; (3) an outlier (decrease).

[0073] Figure 5 shows a schematic diagram of single-point anomaly detection after a pipe rupture occurred in the piping network. Figure 5(a) is the time-series curve of pressure monitoring data for monitoring points 1 and 2. The monitoring data is collected every 15 minutes, and each monitoring point contains 96 monitoring data points daily. Normal values ​​are distributed within the light gray area, and abnormal values ​​are distributed within the green area. From November 12th to November 15th, the data represents historical monitoring data under normal operating conditions for the piping network. On November 16th, the data represents real-time monitoring data, in which case a pipe rupture occurred in the piping network at 10:30, and monitoring points 1 and 2 show a continuous decline starting from the 42nd monitoring data point.

[0074] As shown in Figure 5(b), a schematic diagram is provided comparing five sets of real-time and historical monitoring data after a pipe rupture occurred in the piping network at monitoring points 1 and 2. As shown in the figure, the five sets of real-time monitoring data (42nd, 43rd, 44th, 45th, and 46th) at monitoring points 1 and 2 all show a clear decrease compared to the historical monitoring data corresponding to the past four days. In the figure, all five sets of historical monitoring data are distributed within the light gray area (normal), and all five sets of real-time monitoring data are distributed within the dark gray area (abnormal). When single-point anomaly detection is performed on the five sets of real-time and historical monitoring data using an unsupervised stacking integration algorithm, the real-time monitoring data is marked as "abnormal". At the same time, all real-time monitoring data decreases compared to the historical data; that is, when single-point qualitative detection is performed on real-time and historical monitoring data using statistical theory, the detection result is output as "decrease". Therefore, after a pipe rupture occurs in the piping network, single-point anomaly detection usually yields the following result: (1) Abnormalities are detected in the monitoring data for each monitoring point. (2) Abnormal values ​​tend to decrease compared to normal values.

[0075] Figure 6 shows a schematic diagram of single-point anomaly detection after a pipe rupture occurs in the piping network. Figure 6(a) is the time-series curve of pressure monitoring data for monitoring points 1 and 2. The monitoring data is collected every 15 minutes, and each monitoring point contains 96 monitoring data points daily. Normal values ​​are distributed within the yellow area, and abnormal values ​​are distributed within the green area. From November 12th to November 15th, the data represents historical monitoring data under normal operating conditions for the piping network. On November 16th, the data represents real-time monitoring data, in which case a pipe rupture occurred in the piping network at 10:30, and monitoring points 1 and 2 show a continuous decline starting from the 42nd monitoring data point.

[0076] As shown in Figure 6(b), a schematic diagram is provided comparing five sets of real-time and historical monitoring data after a pipe rupture occurred in the piping network at monitoring points 1 and 2. As shown in the figure, the five sets of real-time monitoring data (numbers 42, 43, 44, 45, and 46) at monitoring points 1 and 2 all show a clear decrease compared to the historical monitoring data corresponding to the past four days. In the figure, all five sets of historical monitoring data are distributed within the yellow area (normal), and all five sets of real-time monitoring data are distributed within the green area (abnormal). When single-point anomaly detection is performed on the five sets of real-time and historical monitoring data using an unsupervised stacking integration algorithm, the real-time monitoring data is marked as "abnormal". At the same time, all real-time monitoring data decreases compared to the historical data; that is, when single-point qualitative detection is performed on the real-time and historical monitoring data using statistical theory, the detection result is output as "decrease". Therefore, after a pipe rupture occurs in the piping network, single-point anomaly detection usually yields the following results: (1) Anomalies are detected in the monitoring data at each monitoring point; (2) The anomaly values ​​tend to decrease compared to normal values.

[0077] Figure 7 shows a schematic diagram of single-point anomaly detection after dirty data appeared at some monitoring points. Figure 7(a) shows the time-series curves of pressure monitoring data for monitoring points 1 and 2. The monitoring data is collected every 15 minutes, and each monitoring point contains 96 monitoring data points daily. From November 12 to November 15, the data is historical monitoring data, and on November 16, the data is real-time monitoring data. At monitoring point 1, multiple (42nd, 43rd, 44th, 45th, 46th, and 47th) dirty data points appear consecutively from 10:30, while all real-time monitoring data at monitoring point 2 are normal values.

[0078] As shown in Figure 7(b), the five sets of real-time monitoring data (42nd, 43rd, 44th, 45th, and 46th) at monitoring point 1 differ significantly from the historical monitoring data corresponding to the past four days. Here, the 42nd, 43rd, 44th, 45th, and 46th real-time monitoring data show an increase, while the 44th and 45th real-time monitoring data show a decrease. In contrast, the five sets of real-time monitoring data at monitoring point 2 do not differ significantly from the historical monitoring data and are all distributed within the yellow area. When single-point anomaly detection is performed on the five real-time monitoring data at monitoring points 1 and 2 using an unsupervised stacking integration algorithm, the anomaly detection result for monitoring point 1 is "abnormal," and the anomaly detection result for monitoring point 2 is "normal." At the same time, when qualitative detection is performed on the real-time monitoring data at monitoring points 1 and 2 using statistical theory, the anomaly detection result for monitoring point 1 is "increase" or "decrease," and the anomaly detection result for monitoring point 2 is "normal." Therefore, after dirty data appears at some monitoring points, single-point anomaly detection usually yields the following results: (1) Anomalies are detected at some monitoring points; (2) Anomalies tend to decrease or increase compared to normal values.

[0079] Figure 8 shows a schematic diagram of single-point anomaly detection after dirty data has appeared at all monitoring points. Figure 8(a) is the time-series curve of pressure monitoring data for monitoring points 1 and 2, with a data collection frequency of 15 minutes / time, and each monitoring point containing 96 monitoring data points daily. From November 12th to November 15th, the data is historical monitoring data, all of which are normal values. On November 16th, the data is real-time monitoring data, and from 10:30 AM, multiple (42nd, 43rd, 44th, 45th, 46th, and 47th) dirty data points appear consecutively at monitoring points 1 and 2.

[0080] As shown in Figure 8(b), the five sets of real-time monitoring data (numbers 42, 43, 44, 45, and 46) at monitoring points 1 and 2 differ significantly from the historical monitoring data corresponding to the past four days. Here, the real-time monitoring data for numbers 42, 43, and 46 increases, while the real-time monitoring data for numbers 44 and 45 decreases. When single-point anomaly detection is performed on the five sets of real-time monitoring data at monitoring points 1 and 2 using an unsupervised stacking integration algorithm, the following results are obtained: x1 0 (42), x1 0 (43), x1 0 (44), x1 0 (45), x1 0 (46), x2 0 (42), x2 0 (43), x2 0 (44), x2 0 (45), and x2 0 (46) are all outliers, meaning that the single-point outlier detection results for monitoring points 1 and 2 are "abnormal". When performing qualitative detection on real-time monitoring data for monitoring points 1 and 2 using statistical theory, the following results are obtained. x1 0 (42), x1 0 (43), x1 0 (46), x2 0 (42), x2 0 (43), and x2 0 (46) increases, x1 0 (44), x1 0 (45), x2 0 (44), and x2 0 (45) decreases, meaning that the single-point anomaly qualitative detection results for monitoring points 1 and 2 are either "increase" or "decrease". Therefore, after dirty data appears at all monitoring points, the single-point anomaly detection results are as follows: (1) Anomalies were detected at all monitoring points; (2) The anomalies tend to increase or decrease compared to the accurate values.

[0081] In summary, both situations involving pipe ruptures in the piping network and the appearance of dirty data at monitoring points result in the detection of a single-point anomaly. However, the results of detecting single-point anomalies differ for different anomaly events, as shown below: (1) After a pipe rupture occurs in the piping network, a single-point anomaly is detected at each monitoring point, and all anomalies are smaller than normal values; (2) After dirty data appears at some monitoring points, a single-point anomaly is detected at the monitoring points where dirty data appeared, and the anomalies tend to increase or decrease; (3) After dirty data appears at all monitoring points, a single-point anomaly is detected at all monitoring points, and the anomalies tend to increase or decrease.

[0082] Step (2) performs time-series anomaly detection on real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. Step (2) specifically includes the following: Step (2.1) performs a hydraulic simulation on the water supply network to perform time-series anomaly detection on abnormal monitoring data and obtains pressure monitoring data under various abnormal situations of the water supply network. The present invention mainly considers three types of abnormal situations: (a) a pipe burst occurs in the water supply network. (b) a failure occurs at some monitoring points of the water supply network. (c) a failure occurs at all monitoring points of the water supply network. Step (2.2) performs anomaly detection on time-series monitoring data from different monitoring points, i.e., identifies time-series anomalies between monitoring points, as shown in Figure 10.

[0083] Figure 9 shows a schematic diagram of time-series anomaly detection between monitoring points. Figure 9(a) shows the time-series curves of real-time and historical monitoring data at monitoring points 1 and 2. December 1st to December 15th is historical monitoring data, and December 16th is real-time monitoring data. The historical monitoring data are normal values, the 95th real-time monitoring data at monitoring point 1 is an anomaly, and all other real-time monitoring data are normal values. d m j If (S1,S2) shows the distance between the mth monitoring data time series on day j at monitoring points 1 and 2, then the distance between the mth monitoring data time series at monitoring points 1 and 2 is d m (S1,S2)=[dm 0 (S1,S2),d m 1 (S1,S2),···,d m j (S1,S2),···,d m 15 (S1,S2)] is shown, d m 0 (S1,S2) is the distance between the time series of real-time monitoring data from monitoring points 1 and 2, and d m j (S1,S2)(j=1,2,···,15) is the distance between the time series of historical monitoring data at monitoring points 1 and 2. Time series anomaly detection between monitoring points is performed by d m 0 (S1,S2) and d m j Comparing with (S1,S2), d m 0 The objective is to check whether (S1,S2) is an outlier. m 0 If (S1,S2) is an abnormal value, it indicates that the real-time monitoring data time series S1(m) or S2(m) of monitoring point 1 or 2 has changed, and the time series between monitoring points is marked as "abnormal". m 0 If (S1,S2) are within normal limits, the time series between monitoring points is marked as "normal".

[0084] Figure 9(b) shows the time series of the 94th monitoring data at monitoring points 1 and 2. The monitoring data at monitoring points 1 and 2 are both normal values, and the time series of the 94th monitoring data at monitoring point 1 is S1(94)=[S1 0 (94), S1 1 (94),···,S1 15 The shape of (94)) is similar, and the time series of the 94th monitoring data at monitoring point 2 is S2(94)=[S2 0 (94), S2 1 (94),..., S2 15 (94)) The change is small. Therefore, the distance d of the real-time monitoring data time series of monitoring points 1 and 2 94 0 (S1,S2) is the distance d of the historical monitoring data time series. 94j The change is not as large as (S1,S2)(j=1,2,···,15), that is, d 94 0 (S1,S2) are within the normal range, and the time series between monitoring points 1 and 2 is "normal".

[0085] Figure 9(c) shows the time series of the 95th monitoring data at monitoring points 1 and 2. The 95th real-time monitoring data at monitoring point 1 is an abnormal value, while all other monitoring data are normal values. As shown in the figure, the time series S1 of the 95th monitoring data at monitoring point 1 0 (95) is the historical monitoring data time series S1 i Compared to (95)(j=1,2,···,15), the shape has changed significantly, and the time series of the 95th monitoring data at monitoring point 2, S2(95)=[S2 0 (95), S2 1 (95),..., S2 15 (95)) The change is small. Therefore, the distance d of the real-time monitoring data time series of monitoring points 1 and 2 95 0 (S1,S2) is the historical monitoring data time series d 95 j This is a significant change compared to (S1,S2)(j=1,2,···,15), that is, d 95 0 (S1,S2) are outliers, and the time series between monitoring points 1 and 2 is "abnormal".

[0086] Step (2.3) Anomaly detection is performed on the time series of monitoring data at the same monitoring point, that is, the time series anomalies of the monitoring point itself are identified, as shown in Figure 11.

[0087] Figure 10 shows a schematic diagram of the time-series anomaly detection at the monitoring point itself. Figure 10(a) shows the time-series curves of real-time and historical monitoring data for monitoring points 1 and 2. The data from December 1st to December 15th is historical monitoring data, and December 16th is real-time monitoring data. All historical monitoring data are normal values, the 95th real-time monitoring data at monitoring point 1 is an anomaly, and all other real-time monitoring data are normal values.

[0088] d m i,j (S1)(i≠j;i,j=0,1,···,15) represents the distance between the mth monitoring data time series on day i and day j at monitoring point 1. Then the distance between the mth real-time monitoring data time series at monitoring point 1 and the mth historical monitoring data time series from the past 15 days is d m 0,j (S1) = [d m 0,1 (S1),d m 0,2 (S1),···,d m 0,15 (S1)] and the distance of the m-th historical monitoring data time series over the past 15 days is d m i,j (S1) = [d m 1,2 (S1),d m 1,3 (S1),···,d m 14,15 (S1)]. Time-series anomaly detection of the monitoring point itself is d m 0,j (S1)(j=1,2,···,15) and d m i,j Compare with (S1)(i≠j;i,j=0,1,···,15), d m 0,j The task is to check whether the distance of one of the monitoring data time series in (S1) is an outlier. m 0,j If (S1) is an outlier, it is found that the mth real-time monitoring data time series at monitoring point 1 changes, and the time series of monitoring point 1 itself is marked as "outlier", and all d m 0,j If (S1) is within the normal range, the time-series anomaly detection result for monitoring point 1 itself is marked as "normal".

[0089] Figure 10(b) shows the time series of the 94th monitoring data at monitoring point 1. All monitoring data at monitoring point 1 are normal values, and the shape of the time series of the 94th monitoring data at monitoring point 1 is similar. Therefore, the distance d of the time series of monitoring point 1 itself is... m i,jThe change in (S1)(i≠j;i,j=0,1,···,15) is small, that is, all d m 0,j (S1) is a normal value, and the time-series anomaly detection result for monitoring point 1 itself is marked as "normal".

[0090] Figure 10(c) shows the time series of the 95th monitoring data at monitoring point 1. The 95th real-time monitoring data at monitoring point 1 is an abnormal value, while all other monitoring data are normal values. As shown in the figure, the time series S1 of the 95th monitoring data at monitoring point 1 0 (95) is the historical monitoring data time series S1 i The shape changes significantly compared to (95) (i=0,1,···,15). Therefore, the time series distance d between real-time and historical monitoring data at monitoring point 1 is m 0,j (S1)(j=1,2,···,15) is the distance d of the historical monitoring data time series at monitoring point 1. m i,j Compared to (S1)(i≠j;i,j=0,1,···,15), it changes similarly, that is, d m 0,j An anomaly exists in (S1), and the time-series anomaly detection result for monitoring point 1 itself is marked as "anomaly".

[0091] Figure 12 shows a schematic diagram of time-series anomaly detection after a pipe rupture occurs in the piping network. Figure 10(a) is the time-series curve of pressure monitoring data for monitoring points 1 and 2, with each monitoring point containing 96 monitoring data points daily. November 14th and 15th are historical monitoring data, both of which are normal values. November 16th is real-time monitoring data, showing that a pipe rupture occurred in the piping network at 10:30, and monitoring points 1 and 2 show a continuous decline starting from the 42nd monitoring data point.

[0092] Figure 12(b) shows a schematic diagram of time-series anomaly detection between monitoring points. At monitoring points 1 and 2, the shape of the time series of the 42nd, 43rd, 44th, and 45th monitoring data on the 16th changes significantly compared to the time series of the monitoring data on the 15th. However, because the time series of the monitoring data at monitoring points 1 and 2 change simultaneously and the laws of change are similar, the change in the distance between the time series at monitoring points 1 and 2 is not large. For example, the distance between monitoring points 1 and 2 for the 42nd time series on the 16th is d 42 0 (S1,S2) is the distance d of the 42nd time series from the previous day (i.e., the 15th). 42 1 The change is smaller compared to (S1, S2), meaning that the 42nd time-series anomaly detection result for monitoring points 1 and 2 on the 16th is "normal". As shown in the figure, if anomaly detection is performed on the distance of the 43rd, 44th, and 45th monitoring data time series for monitoring points 1 and 2 on the 16th, all detection results may be "normal".

[0093] Figures 12(c) and 12(d) show schematic diagrams of time-series anomaly detection at monitoring points 1 and 2, respectively. As shown in Figure 12(c), at monitoring point 1, the shape of the three sets of monitoring data time series (42nd, 43rd, and 44th) over 16 days changes significantly compared to the monitoring data time series over 14 days and 15 days, resulting in a large change in the distance between the real-time and historical monitoring data time series. For example, the distance between the 42nd real-time (16th) and historical (15th) monitoring data time series at monitoring point 1 is d 45 16,15 (S1) The distance between the 42nd history (15 days) and the history (14 days) of the monitoring data time series is d 45 15,14 (S1) and d 45 16,15 (S1) is clearly d 45 15,14 (S1) is greater than the result, meaning that the time-series anomaly detection result for monitoring point 1 itself is "abnormal". Similarly, the 42nd, 43rd, and 44th time-series anomaly detection results for monitoring point 2 are all "abnormal".

[0094] In summary, after a pipe rupture occurs in the piping network, time-series anomaly detection typically yields the following results: (1) Time-series detection results between monitoring points may be abnormal or normal; (2) Time-series detection for each monitoring point itself will always show anomalies.

[0095] Figure 13 shows a schematic diagram of time-series anomaly detection after dirty data appears at some monitoring points. Figure 13(a) shows the time-series curves of pressure monitoring data for monitoring points 1 and 2, with each monitoring point containing 96 monitoring data points daily. From November 14th to 15th, the data is historical monitoring data, all of which are normal values. On November 16th, the data is real-time monitoring data, and at monitoring point 1, multiple (44th, 45th, 46th, and 45th) dirty data points appear consecutively from 10:30, while all monitoring data at monitoring point 2 are normal values.

[0096] Figure 13(b) shows a schematic diagram of time-series anomaly detection between monitoring points. At monitoring point 1, the time series of monitoring data for the 4th set (42nd, 43rd, 44th, and 45th) on the 16th shows a significant change compared to the time series of monitoring data on the 15th. At the same time, the change in the time series of monitoring data for the 4th set at monitoring point 2 is small. Therefore, the change in the real-time monitoring data time series at monitoring points 1 and 2 is larger than that of the historical monitoring data time series. For example, the distance between monitoring points 1 and 2 for the 42nd monitoring data time series on the 16th is d 42 0 (S1,S2) is the distance d of the monitoring data time series from the previous day (15th). 42 1 The change is larger compared to (S1, S2), meaning that the anomaly detection result for the 42nd monitoring data time series on the 16th at monitoring points 1 and 2 is "abnormal". As shown in the figure, when anomaly detection is performed on the distance of the 43rd, 44th, and 45th monitoring data time series on the 16th at monitoring points 1 and 2, the detection result is "abnormal" in all cases.

[0097] Figures 13(c) and 13(d) show schematic diagrams of time-series anomaly detection for monitoring points 1 and 2, respectively. As shown in Figure 11(c), at monitoring point 1, the shape of the time series of the 16th day (3 sets: 42nd, 43rd, and 44th) of monitoring data changes significantly compared to the time series of the 14th and 15th day of monitoring data, resulting in a large change in the distance between the real-time and historical monitoring data time series. For example, the distance between the 42nd real-time (16th) and historical (15th day) monitoring data time series at monitoring point 1 is d 42 16,15 (S1) The distance between the 42nd history (15 days) and the history (14 days) of the monitoring data time series is d 42 15,14 (S1) and d 42 16,15 (S1) is clearly d 42 15,14 (S1) is greater than the result, meaning the time-series anomaly detection result for monitoring point 1 itself is "abnormal".

[0098] As shown in Figure 13(d), at monitoring point 2, the time series of monitoring data for the three sets of 16 days (42nd, 43rd, and 44th) shows less change compared to the time series of monitoring data for 14 days and 15 days, and the change in the distance between the real-time and historical monitoring data time series is not large. For example, the distance between the 42nd real-time (16th) and historical (15th) monitoring data time series at monitoring point 1 is d 42 16,15 (S2) The distance between the 42nd history (15 days) and the history (14 days) monitoring data time series is d 42 15,14 (S2) and d 42 16,15 (S2) is d 42 15,14 There is no significant difference compared to (S2), meaning that the time-series S anomaly detection result for monitoring point 2 itself is "normal".

[0099] In summary, after dirty data is generated at some monitoring points, time-series anomaly detection typically yields the following results: (1) Anomalies are detected in the time series between monitoring points; (2) Anomalies are detected in the time series of some monitoring points themselves.

[0100] Figure 14 shows a schematic diagram of time-series anomaly detection after dirty data appears at all monitoring points. Figure 4.4.6(a) shows the time-series curves of pressure monitoring data for monitoring points 1 and 2, with each monitoring point containing 96 monitoring data points daily. November 14th and 15th are historical monitoring data, and all historical monitoring data are normal values. November 16th is real-time monitoring data, and monitoring points 1 and 2 show multiple (42nd, 43rd, 44th, 45th, and 46th) anomaly values ​​appearing consecutively from 10:30.

[0101] Figure 14(b) shows a schematic diagram of time-series anomaly detection between monitoring points. At monitoring points 1 and 2, the time series of monitoring data for 4 sets (43rd, 44th, 45th, and 46th) on the 16th differ significantly from the time series of monitoring data on the 15th. If the change trends of the real-time monitoring data time series at monitoring points 1 and 2 are the same, the time-series anomaly detection result between monitoring points is "normal". If the real-time monitoring data time series at monitoring points 1 and 2 change significantly, the time-series anomaly detection result between monitoring points is "abnormal". For example, the distance between monitoring points 1 and 2 for the 43rd monitoring time series on the 16th is d 43 0 (S1,S2) is the distance d of the monitoring data time series from the previous day (i.e., the 15th). 43 1 If the change is small compared to (S1,S2), then monitoring points 1 and 2 have a "normal" time-series anomaly detection result for the 43rd time series on the 16th. The distance between monitoring points 1 and 2 for the 46th monitoring data time series on the 16th is d 46 0 (S1,S2) is the distance d of the monitoring data time series from the previous day (i.e., the 15th). 46 1 Compared to (S1,S2), there is a significant change, meaning that monitoring points 1 and 2 show that the 46th time-series anomaly detection result on the 16th is "abnormal".

[0102] Figures 14(c) and 14(d) show schematic diagrams of time-series anomaly detection at monitoring points 1 and 2, respectively. As shown in Figure 14(c), at monitoring point 1, the shape of the three sets of monitoring data time series (42nd, 43rd, and 44th) over 16 days changes significantly compared to the monitoring data time series over 14 days and 15 days, resulting in a large change in the distance between the real-time and historical monitoring data time series. For example, the distance between the 42nd real-time (16th) and historical (15th) monitoring data time series at monitoring point 1 is d 42 16,15 (S1) The distance between the 42nd history (15 days) and the history (14 days) of the monitoring data time series is d 42 15,14 (S1) and d 42 16,15 (S1) is clearly d 42 15,14 (S1) is greater than the result shown, meaning that the time-series anomaly detection result for monitoring point 1 itself is "abnormal". Similarly, as shown in Figure 14(d), all of the time-series anomaly detection results for monitoring point 2 itself are "abnormal".

[0103] In summary, after dirty data appears at all monitoring points, time-series anomaly detection typically yields the following results: (1) Time-series detection results between monitoring points may be anomalous or normal; (2) Anomalies are detected in the time series of all monitoring points themselves.

[0104] The occurrence of a pipe rupture in the piping network and the appearance of dirty data at some (or all) monitoring points both result in anomalies in the time series of monitoring data, but various time series anomaly detection results appear, as shown below: (1) A pipe rupture occurs in the piping network. The time series anomaly detection results between monitoring points may be normal or abnormal. Anomalies are detected in the time series of all monitoring points themselves; (2) Anomalies are detected in the time series of some monitoring points themselves; (3) Dirty data appears at all detection points. The time series anomaly detection results between monitoring points may be normal or abnormal. Anomalies are detected in the time series of all monitoring points themselves.

[0105] Step (3) Experiments are conducted using the standard test piping network Net 3 piping network model to verify and evaluate the pipe burst detection performance of the provided method.

[0106] As shown in Figure 15, assuming that three pressure monitoring points are distributed within the piping network, the distributed nodes are 169, 204, and 275 in order, and are indicated as pressure monitoring points 1, 2, and 3, respectively.

[0107] Using monitoring data from three pressure monitoring points, single-point and time-series anomaly detection is performed. The experimental conditions and data preparation techniques are shown in Figure 16 and include the following steps: (a) Obtain monitoring data through simulation regarding pipe ruptures in the piping network or the appearance of dirty data at monitoring points; (b) Perform single-point anomaly detection on the monitoring data based on a single-point unsupervised classification method; (c) Perform time-series anomaly detection on the monitoring data based on a time-series unsupervised classification method; (d) Integrate the single-point and time-series anomaly detection results to obtain detection results for various anomaly scenarios.

[0108] Figure 16 shows a schematic diagram of the monitoring data.csv file, which includes the collection date, time, time step, status, and tab information for monitoring data at each monitoring point. The status of the monitoring data is divided into two types: normal (0) and abnormal (1). The tabs of the monitoring data are divided into monitoring data under normal operating conditions of the piping network (0) and monitoring data under abnormal operating conditions of the piping network (1-9). Table 1 shows the situations in which abnormalities appear in various monitoring data.

[0109] Table 1: Situations in which anomalies appear in various monitoring data [Table 1]

[0110] Figure 17 shows the results of single-point anomaly detection based on an unsupervised stacking integration algorithm, considering four scenarios: (1) As shown in Figure 17(a), a pipe rupture occurs in the piping network, and the monitoring data from monitoring points 1, 2, and 3 are all abnormal values; (2) As shown in Figure 17(b), all the monitoring data from monitoring point 1 are abnormal values, and all the monitoring data from monitoring points 2 and 3 are normal values; (3) As shown in Figure 17(c), all the monitoring data from monitoring points 1 and 2 are abnormal values, and the monitoring data from monitoring point 3 is normal values; (4) As shown in Figure 17(d), all the monitoring data from monitoring points 1, 2, and 3 are abnormal values. If all five consecutive time points of monitoring data are detected as abnormal values, an anomaly warning is issued, and the anomaly detection result "abnormal" is output, indicated by a dot in the yellow area in the figure. Conversely, if there is no abnormal event, the anomaly detection result "normal" is output, indicated by a dot in the green area in the figure. As shown in Figure 17, the unsupervised stacking integration algorithm can detect anomalies in most anomaly detection data, while simultaneously failing to issue anomaly warnings for some anomaly monitoring data. Some anomalies in the monitoring data may have small changes, making it difficult to meet the requirement of detecting anomalies at all five consecutive time points, and therefore, some anomalies fail to trigger anomaly warnings. In addition, the unsupervised stacking integration algorithm can detect situations where pipe ruptures occur in the piping network and situations where dirty data appears at all monitoring points, but it cannot distinguish between these two types of anomaly scenarios.

[0111] Figure 18 shows the results of single-point anomaly detection based on statistical theory, considering four scenarios: (1) As shown in Figure 18(a), a pipe rupture occurs in the piping network, and all monitoring data at monitoring points 12 and 3 are abnormal values; (2) As shown in Figure 4.5.5(b), all monitoring data at monitoring point 1 are abnormal values, but all monitoring data at monitoring points 2 and 3 are normal values; (3) As shown in Figure 18(c), all monitoring data at monitoring points 1 and 2 are abnormal values, and all monitoring data at monitoring point 3 are normal values; (4) As shown in Figure 18(d), all monitoring data at monitoring points 1, 2 and 3 are abnormal values. If all monitoring data for five consecutive time points are detected as abnormal values, an anomaly warning is issued: (1) If any abnormal value is greater than the normal value, the anomaly detection result "increase" is output and indicated by a dot in the yellow area; (2) If all abnormal values ​​are smaller than the normal value, the anomaly detection result "decrease" is output and indicated by a dot in the orange area. Conversely, if there are no abnormal events, the abnormality detection result is output as "normal" and shown as a dot within the green area. As shown in Figure 18, the statistical theory-based method can detect some abnormal monitoring data, and most of the abnormal monitoring data has not yet been detected, but the method can distinguish between a scenario in which a pipe rupture occurs in the piping network and a scenario in which dirty data appears at all monitoring points.

[0112] Figure 19 shows the time-series anomaly detection results between monitoring points, considering four types of situations: (1) As shown in Figure 19(a), a pipe rupture occurs in the piping network, and the monitoring data from monitoring points 1, 2, and 3 are all abnormal values; (2) As shown in Figure 19(b), the monitoring data from monitoring point 1 are all abnormal values, and the monitoring data from monitoring points 2 and 3 are all normal values; (3) As shown in Figure 19(c), the monitoring data from monitoring points 1 and 2 are all abnormal values, and the monitoring data from monitoring point 3 are all normal values; (4) As shown in Figure 19(d), the monitoring data from monitoring points 1, 2, and 3 are all abnormal values. An anomaly warning is issued if the time-series detection results for five consecutive time points are all abnormal: (1) the distance between the time series of monitoring data from two monitoring points (e.g., d m (S1, S2) and d mIf all of the distances (e.g., d m m (S1, S2) or d m m (S1, S3)) are abnormal, output the result "Two abnormalities" and indicate it with points within the orange area. (2) If only the distance of the monitoring data time series of one monitoring point (e.g., d m

[0113] Figure 20 shows the time series anomaly detection results of the monitoring points themselves, considering four situations: (1) As shown in Figure 20(a), a pipe rupture occurs in the pipe network, and the monitoring data of monitoring points 1, 2, and 3 are all abnormal values; (2) As shown in Figure 20(b), the monitoring data of monitoring point 1 are all abnormal values, and the monitoring data of monitoring points 2 and 3 are all normal values; (3) As shown in Figure 20(c), the monitoring data of monitoring points 1 and 2 are all abnormal values, and the monitoring data of monitoring point 3 are all normal values; (4) As shown in Figure 20(d), the monitoring data of monitoring points 1, 2, and 3 are all abnormal values. If the time series anomaly detection results for five consecutive time moments are all abnormal, issue an anomaly warning and indicate it with points within the yellow area. Conversely, output the result "Normal" and indicate it with points within the green area. As shown in Figure 20, based on the time series of the monitoring points themselves, various abnormal scenarios can be effectively detected. At the same time, scenarios where dirty data appears at a single monitoring point and some monitoring points can be identified, but scenarios where a pipe rupture occurs in the pipe network and scenarios where dirty data appears at all monitoring points cannot be accurately identified.

[0114] Detect the situation where dirty data appears at the monitoring points. Except for scenes 4 and 8, σ2 for all scenes is approximately 98%. As shown in Figure 21, the provided method can accurately detect outliers in the monitoring data and effectively identify various abnormal scenes. Here, scene 4 is identified as an abnormal scene where dirty data appears at three monitoring points, as a scene where a pipeline rupture occurs in the pipeline network. When identifying an abnormal scene where a pipeline rupture occurs in the pipeline network or dirty data appears at the monitoring points, it is usually easy to identify for abnormal scenes where dirty data appears at a single or some monitoring points. However, when abnormalities (all monitoring data clearly decreases) appear in the monitoring data of all monitoring points, it may be identified as a scene where a pipeline rupture occurs in the pipeline network. Considering the safety of water supply, it is necessary to issue a warning of pipeline rupture and conduct an investigation on the pipeline network suspected of pipeline rupture.

[0115] Figure 22 shows the abnormal detection situation of various scenes after removing the single-point outlier detection results. As shown in Figure 22, after removing the single-point outlier detection results, the impact on the detection results of scenes 2 - 9 is not significant, that is, the impact on abnormal scenes where dirty data appears at the monitoring points is small. However, the detection results of pipeline rupture in the pipeline network are greatly affected, mainly reflected in the abnormal detection rate and abnormal identification rate. After removing the single-point outlier detection results, some abnormal monitoring data caused by pipeline rupture in the pipeline network are not detected, resulting in a decrease in the abnormal detection rate. At the same time, when identifying an abnormal scene where a pipeline rupture occurs in the pipeline network or dirty data appears at the monitoring points, the pipeline rupture in the pipeline network at some times is identified as a situation where dirty data appears at the monitoring points, causing a decrease in the abnormal identification rate. This is because after a certain period after the pipeline rupture occurs, the monitoring data of each monitoring point all decreases, so no abnormalities are detected in the time series of the monitoring points themselves and the time series between the monitoring points. Single-point outlier detection can just detect the decrease in the pressure monitoring data after the pipeline rupture.

[0116] Figure 23 shows the detection results for various abnormal situations after removing the single-point qualitative detection results. As shown in Figure 23, the abnormal detection rate for all various situations reaches 100%, and the detection accuracy and abnormality identification rate all clearly decrease. This indicates that all abnormal situations were detected, although there are situations where false warnings or various abnormal situations are misdetected. Clearly, the single-point qualitative detection results are mainly used to exclude false warnings. When the single-point qualitative detection results are integrated, the detection accuracy and abnormality identification rate for various situations are both high, and at the same time, the abnormality detection rate is all below 100%. Under normal operating conditions, fluctuations in water demand due to weather or holidays cause changes in monitoring data at each monitoring point in the piping network, and these changes are detected by detecting time-series anomalies within the monitoring points themselves and time-series anomalies between monitoring points. Through the single-point qualitative detection results, abnormal warning situations in monitoring data due to fluctuations in water demand are excluded. When the single-point qualitative detection results are removed, the false warning rate for pipe burst detection increases.

[0117] As shown in Figure 24, after removing the time-series anomaly detection results between monitoring points, the detection results for pipe ruptures in the piping network (Scenario 1) are not significantly affected, but the anomaly scenario in which dirty data appears at a monitoring point is significantly affected, and the anomaly detection rate for all eight scenarios clearly decreases. This indicates that multiple anomalies are not detected in the anomaly scenario in which dirty data appears at a monitoring point. After dirty data appears at a monitoring point, the shape of the time series of the monitoring data for each or some of the monitoring points changes. If the shape of the time series of the monitoring data for a monitoring point matches the shape of the time series of its historical monitoring data, no anomaly is detected in the time series of the monitoring point itself. When the changes in these monitoring values ​​are small, it is difficult to detect anomalies through single-point anomaly detection. Clearly, these anomalies can be detected only through time-series anomaly detection between monitoring points.

[0118] Figure 25 shows the detection results for various anomalous scenarios after removing the time-series anomaly detection results for the monitoring point itself. As shown in Figure 25, the anomaly detection rate and anomaly identification rate for all various anomalous scenarios decrease. Clearly, after dirty data appears at a monitoring point, the shape of its own time series changes. When anomalies appear in some monitoring data, the magnitude of the monitoring value does not change significantly, so it cannot be detected by single-point anomalies or time-series anomaly detection methods between monitoring points, resulting in a decrease in the anomaly detection rate. In addition, the loss of time-series warning status for each monitoring point itself also leads to a decrease in the identification rate for various anomalous scenarios.

[0119] (Note) (Note 1) Step (1) involves detecting a single-point anomaly in real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. Step (2) involves performing time-series anomaly detection on real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. A method for detecting and evaluating the performance of a water supply pipe network based on single-point and time-series anomaly detection, comprising the step (3) of integrating single-point and time-series anomaly detection results of the water supply pipe network to accurately detect and identify the status of pipe ruptures in the water supply pipe network and failures in the monitoring system.

[0120] (Note 2) Step (1) above specifically means, Step (1.1) involves preparing the history and real-time single-point monitoring data for each monitoring point in the water supply network in order to perform single-point anomaly detection on real-time monitoring data, Step (1.2) involves using an unsupervised stacking integration algorithm to detect single-point anomalies in real-time monitoring data of the water supply network and identifying single-point anomalies in the real-time monitoring data. A method for detecting pipe ruptures in a water supply network and evaluating its performance based on single-point and time-series anomaly detection, as described in Appendix 1, comprising the step (1.3) of using statistical theory to perform qualitative detection on a single value of monitoring data of the water supply network and identifying a single point anomaly in real-time monitoring data.

[0121] (Note 3) Step (2) above specifically refers to, Step (2.1) involves preparing the history and real-time monitoring data time series at each monitoring point in the water supply network in order to perform time-series anomaly detection on real-time monitoring data, Step (2.2) involves performing a difference analysis on the time series of monitoring data at different monitoring points to identify anomalous time series between monitoring points, A method for detecting bursts in a water supply network pipes and evaluating performance based on the detection of anomalies at a single point and in a time series, as described in Appendix 1, characterized by including the step (2.3) of performing a difference analysis on the time series of monitoring data at the same monitoring point to determine the abnormal time series of the monitoring point itself.

[0122] (Note 4) Step (3) above specifically refers to, Step (3.1) involves using a single-point anomaly detection method to perform single-point anomaly detection for various abnormal situations in the water supply pipe network, Step (3.2) involves using a time-series anomaly detection method to perform time-series anomaly detection on various abnormal situations in the water supply network, Step (3.3) involves integrating single-point and time-series anomaly detection results to detect and identify various abnormal situations in the water supply network, A method for detecting and evaluating the performance of a water supply pipe network rupture based on single-point and time-series anomaly detection, as described in Appendix 1, comprising the step (3.4) of detecting and identifying various abnormal situations in the water supply pipe network using a pipe rupture detection method, and evaluating the pipe rupture detection performance of the provided method.

[0123] (Note 5) The aforementioned step (3.1) specifically means, The water supply pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection described in Appendix 4 is characterized in that, when performing single-point anomaly detection for various abnormal situations in a water supply pipe network using a single-point anomaly detection method, four scenarios are considered: (1) a pipe rupture occurs in the piping network, (2) dirty data appears at a single monitoring point, (3) dirty data appears at some monitoring points, and (4) dirty data appears at all monitoring points, each scenario includes 25 sets of anomaly monitoring data, single-point anomaly detection is performed from the time the anomaly begins, and a decision is made on whether or not to issue a warning based on the anomaly detection results at five consecutive times.

[0124] (Note 6) The aforementioned step (3.2) specifically means, The water supply pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection described in Appendix 4 is characterized in that, when performing time-series anomaly detection for various abnormal situations in the water supply pipe network using a time-series anomaly detection method, four scenarios are considered: (1) a pipe rupture occurs in the piping network, (2) dirty data appears at a single monitoring point, (3) dirty data appears at some monitoring points, and (4) dirty data appears at all monitoring points, and each scenario includes 25 sets of anomaly monitoring data, and time-series anomaly detection is performed between monitoring points from the time the anomaly value begins, and a decision is made on whether or not to issue a warning based on the time-series anomaly detection results at five consecutive times.

[0125] (Note 7) The aforementioned step (3.3) specifically refers to, The steps include detecting whether a single point or time-series anomaly exists in the monitoring data (3.3.1), Step 3.3.2: Identify the situation in which anomalies appear in the monitoring data at some monitoring points. Step 3.3.3 involves identifying a pipe rupture and the resulting appearance of dirty data at all monitoring points. The step (3.3.4) of evaluating and analyzing the pipe rupture detection performance in the method, and the water supply pipe network pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection described in Appendix 4, characterized by including this step.

[0126] (Appendix 8) Specifically, first, identify the normal operating conditions of the pipe network and the situations where anomalies appear in the monitoring data. When detecting the existence of single-point anomaly values for single-point monitoring data based on the teacherless stacking integration algorithm or statistical theory, mark the time when the single-point anomaly value appears, and continue to perform single-point anomaly value detection for the monitoring data at the next time. For time-series monitoring data, when finding an abnormal time series based on time-series anomaly detection between monitoring points or time-series anomaly detection of the monitoring points themselves, mark the time when the abnormal time series appears, and continue to perform time-series anomaly detection for the monitoring data at the next time (step 3.3.1). Specifically, when anomalies appear in the single-point and time-series of the monitoring data at some monitoring points and do not appear in the single-point and time-series of the monitoring data at the remaining monitoring points, determine it as a situation where dirty data appears at some monitoring points (step 3.3.2). Specifically, when anomalies are detected in both the single-point and time-series of the monitoring data at all monitoring points, identify the situation where a pipe rupture occurs in the pipe network through the increase and decrease of the anomaly values, and the situation where dirty data appears at all monitoring points. When the single-point qualitative detection results of the monitoring data at all monitoring points are all "decrease", it can be known that a pipe rupture has occurred in the pipe network. Conversely, when the single-point qualitative detection result of a certain monitoring data is "increase", it can be known that dirty data has appeared at all monitoring points of the pipe network (step 3.3.3). Evaluating the pipe rupture detection performance in the method includes the step (3.3.4) of considering several indicators such as (1) detection accuracy rate (σ1), (2) anomaly discrimination rate (σ2), and (3) anomaly detection rate (σ), and the water supply pipe network pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection described in Appendix 7, characterized by including this step.

[0127] (Appendix 9) The detection accuracy is shown as follows: σ1=N dn / N n In the formula, N n The total number of abnormal situations is shown, N dn This indicates the number of abnormal scenes detected. The anomaly detection rate is shown as follows: σ² = N in / N n In the formula, N in This indicates the number of accurately identified anomalous scenes, N n This indicates the total number of abnormal situations. The anomaly detection rate is shown as follows: σ³=t dn / t t In the formula, t dn and t t The method for detecting and evaluating the performance of a water supply pipe network rupture based on single-point and time-series anomaly detection, as described in Appendix 8, is characterized in that the duration of the anomaly detection and the actual duration are shown respectively.

[0128] (Note 10) The aforementioned step (3.4) specifically means, When detecting and identifying various abnormal situations in the water supply pipe network, the steps include (3.4.1) removing single-point anomaly detection results and obtaining the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations, When detecting and identifying various abnormal situations in the water supply pipe network, the steps (3.4.2) include removing single-point qualitative detection results and obtaining the abnormality detection rate, detection accuracy rate, and abnormality identification rate for various abnormal situations, When detecting and identifying various abnormal situations in the water supply network, the time-series abnormality detection results between monitoring points are removed to obtain the abnormality detection rate, detection accuracy rate, and abnormality identification rate for various abnormal situations (3.4.3). The water supply pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection described in Appendix 4, characterized in that when detecting and identifying various abnormal situations in the water supply pipe network, the method includes the step (3.4.4) of removing the time-series anomaly detection results of the monitoring point itself and obtaining the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations.

Claims

1. A step (1) of detecting a single point anomaly in real-time monitoring data of the water supply network and identifying normal operating conditions and abnormal operating conditions of the water supply network, Step (2) involves performing time-series anomaly detection on real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. The process includes (3) integrating single-point and time-series anomaly detection results in the water supply network to accurately detect and identify pipe bursts and monitoring system failures in the water supply network, Step (1) above specifically means, Step (1.1) involves preparing the history and real-time single-point monitoring data for each monitoring point in the water supply network in order to perform single-point anomaly detection on real-time monitoring data, Steps (1.2) include using an unsupervised stacking integration algorithm to detect single-point anomalies in real-time monitoring data of the water supply network and identifying single-point anomalies in the real-time monitoring data, A method for detecting pipe ruptures in a water supply network and evaluating its performance based on the detection of single-point and time-series anomalies, characterized by including the step (1.3) of using statistical theory to perform qualitative detection on a single value of monitoring data of the water supply network and identifying a single point anomaly in real-time monitoring data.

2. A step (1) of detecting a single point anomaly in real-time monitoring data of the water supply network and identifying normal operating conditions and abnormal operating conditions of the water supply network, Step (2) involves performing time-series anomaly detection on real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. The process includes (3) integrating single-point and time-series anomaly detection results in the water supply network to accurately detect and identify pipe bursts and monitoring system failures in the water supply network, Step (2) above specifically refers to, Step (2.1) involves preparing the history and real-time monitoring data time series at each monitoring point in the water supply network in order to perform time-series anomaly detection on real-time monitoring data, Step 2.2 involves performing a difference analysis on the time series of monitoring data at different monitoring points to identify anomalous time series between monitoring points, A method for detecting and evaluating the performance of a water supply pipe network based on the detection of anomalies at a single point and in a time series, characterized by including the step (2.3) of performing a difference analysis on the time series of monitoring data at the same monitoring point to determine the abnormal time series of the monitoring point itself.

3. A step (1) of detecting a single point anomaly in real-time monitoring data of the water supply network and identifying normal operating conditions and abnormal operating conditions of the water supply network, Step (2) involves performing time-series anomaly detection on real-time monitoring data of the water supply network to identify normal and abnormal operating conditions of the water supply network. The process includes (3) integrating single-point and time-series anomaly detection results in the water supply network to accurately detect and identify pipe bursts and monitoring system failures in the water supply network, Step (3) above specifically refers to, Step (3.1) involves using a single-point anomaly detection method to perform single-point anomaly detection for various abnormal situations in the water supply pipe network, Step (3.2) involves using a time-series anomaly detection method to perform time-series anomaly detection on various abnormal situations in the water supply network, Step (3.3) involves integrating single-point and time-series anomaly detection results to detect and identify various abnormal situations in the water supply network, A method for detecting and evaluating the performance of pipe ruptures in a water supply network based on single-point and time-series anomaly detection, comprising the steps of: (3.4) detecting and identifying various abnormal situations in the water supply network using a pipe rupture detection method, and evaluating the pipe rupture detection performance of the provided method.

4. The aforementioned step (3.1) specifically refers to, The water supply pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection, as described in claim 3, is characterized in that, when performing single-point anomaly detection for various abnormal situations in a water supply pipe network using a single-point anomaly detection method, four scenarios are considered: (1) a pipe rupture occurs in the piping network, (2) dirty data appears at a single monitoring point, (3) dirty data appears at some monitoring points, and (4) dirty data appears at all monitoring points, and each scenario includes 25 sets of anomaly monitoring data, and single-point anomaly detection is performed from the time the anomaly value begins, and a decision is made on whether or not to issue a warning based on the anomaly detection results at five consecutive times.

5. The aforementioned step (3.2) specifically refers to, The water supply pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection, as described in claim 3, is characterized in that, when performing time-series anomaly detection for various abnormal situations in a water supply pipe network using a time-series anomaly detection method, four scenarios are considered: (1) a pipe rupture occurs in the pipe network, (2) dirty data appears at a single monitoring point, (3) dirty data appears at some monitoring points, and (4) dirty data appears at all monitoring points, and each scenario includes 25 sets of anomaly monitoring data, and time-series anomaly detection is performed between monitoring points from the time the abnormal value begins, and a decision is made on whether or not to issue a warning based on the time-series anomaly detection results at five consecutive times.

6. The aforementioned step (3.3) specifically refers to, The steps include detecting whether a single point or time-series anomaly exists in the monitoring data (3.3.1), Step 3.3.2: Identify the situation in which anomalies appear in the monitoring data at some monitoring points. Step 3.3.3 involves identifying a pipe rupture and the appearance of dirty data at all monitoring points, A method for detecting and evaluating the performance of a water supply network pipe burst based on single-point and time-series anomaly detection, as described in 3, further comprising the step (3.3.4) of evaluating and analyzing the pipe burst detection performance in the method.

7. Specifically, the process involves first identifying the normal operating conditions of the piping network and the circumstances under which abnormalities appear in the monitoring data, and, if a single-point anomaly is detected in the single-point monitoring data based on an unsupervised stacking integration algorithm or statistical theory, marking the time at which the single-point anomaly appeared, and continuing to perform single-point anomaly detection on the monitoring data for the next time point. Furthermore, if an abnormal time series is found in the time-series monitoring data based on time-series anomaly detection between monitoring points or within the monitoring point itself, marking the time at which the abnormal time series appeared, and continuing to perform time-series anomaly detection on the monitoring data for the next time point (3.3.1). Specifically, if an anomaly appears in a single point and time series of monitoring data at some monitoring points, but no anomalies appear in a single point and time series of monitoring data at the remaining monitoring points, this is determined to be a situation where dirty data appears at some monitoring points (3.3.2), Specifically, if an anomaly is detected in the single-point and time-series monitoring data at all monitoring points, the system identifies the situation where a pipe rupture has occurred in the piping network and the situation where dirty data appears at all monitoring points by examining the increase and decrease in the anomaly values. If the single-point qualitative detection result of the monitoring data at all monitoring points is "decreasing," it indicates that a pipe rupture has occurred in the piping network. Conversely, if the single-point qualitative detection result of any monitoring data is "increasing," it indicates that dirty data has appeared at all monitoring points in the piping network (step 3.3.3). Evaluating the pipe rupture detection performance of the method involves (1) the detection accuracy rate (σ １ ), (2) Anomaly identification rate (σ ２ ), (3) Anomaly detection rate (σ ３ A method for detecting and evaluating the performance of a water supply pipe network rupture based on single-point and time-series anomaly detection, as described in 6, comprising the step (3.3.4) of considering several indicators such as ).

8. The detection accuracy is shown as follows: s １ =N ｄｎ / N ｎ In the formula, N ｎ The total number of abnormal situations is shown, N ｄｎ This indicates the number of abnormal scenes detected. The anomaly detection rate is shown as follows: s ２ =N ｉｎ / N ｎ In the formula, N ｉｎ This indicates the number of accurately identified anomalous scenes, N ｎ This indicates the total number of abnormal situations. The anomaly detection rate is shown as follows: σ ３ =t ｄｎ / t ｔ In the formula, t ｄｎ and t ｔ The method for detecting and evaluating the performance of a water supply pipe network rupture based on single-point and time-series anomaly detection, as described in claim 7, characterized in that the duration of the anomaly detection and the actual duration are shown respectively.

9. The aforementioned step (3.4) specifically refers to, When detecting and identifying various abnormal situations in the water supply network, the steps (3.4.1) include removing single-point anomaly detection results and obtaining the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations, When detecting and identifying various abnormal situations in the water supply network, the steps (3.4.2) include removing single-point qualitative detection results and obtaining the abnormality detection rate, detection accuracy rate, and abnormality identification rate for various abnormal situations, When detecting and identifying various abnormal situations in the water supply pipe network, the time-series abnormality detection results between monitoring points are removed to obtain the abnormality detection rate, detection accuracy rate, and abnormality identification rate for various abnormal situations (3.4.3). The water supply pipe rupture detection and performance evaluation method based on single-point and time-series anomaly detection, characterized in that when detecting and identifying various abnormal situations in the water supply pipe network, the method includes the step (3.4.4) of removing the time-series anomaly detection results of the monitoring point itself and obtaining the anomaly detection rate, detection accuracy rate, and anomaly identification rate for various abnormal situations.

Images