A method for monitoring primary heating networks

Abstract

This invention relates to heating network monitoring technology and provides a method for monitoring primary heating networks. Based on a low-power wide-area network (LPWAN), the method facilitates data exchange between monitoring terminals and a cloud server. For each batch of reported pressure monitoring data, the method calculates a pipe pressure drop characterization sequence for node pressure sequences within the same cluster and aligned time window, based on the server's receiving time. Anomaly detection in the network is then based on this pipe pressure drop characterization sequence calculation. This monitoring method is applicable to primary heating network monitoring systems built on LPWANs, improving the sensitivity of such systems while reducing false alarm rates in low-cost deployment and operation scenarios.

Description

Technical Field

[0001] This disclosure belongs to the field of pipeline system technology, and relates to pipeline network engineering management methods, especially to heating pipeline network monitoring technology. Background Technology

[0002] The primary heating network refers to the partially or completely closed pressurized circulation pipeline in a centralized heating system, connecting heat sources such as combined heat and power plants and regional boiler rooms to the primary side of each heating station. It mainly includes supply and return water pipelines. Its core function is to transport heat energy over long distances and in large volumes between heat sources and distributed heat exchange nodes, and it generally does not directly supply heat to end users. The internal circulation medium is liquid water, operating under high-temperature and high-pressure conditions with a supply water temperature of 110℃-150℃, a return water temperature of 60℃-70℃, and an operating pressure of 1.0MPa-2.5MPa. The physical properties of the medium are sensitive to temperature changes. The primary heating network is usually hydraulically isolated from the secondary network through heat exchangers, forming a hydraulically continuous closed system. Its internal pressure dynamics mainly originate from the operating conditions of the heat source circulation pumps, the pipeline's own operating conditions, and the operation of limited regulating valves. It is largely unaffected by the frequent and random adjustments made by users on the secondary network, possessing independent system characteristics.

[0003] Current pipeline monitoring technologies for leak detection primarily rely on synchronous absolute pressure model analysis. This approach requires deploying pressure and flow sensors with high-precision clock synchronization capabilities (e.g., via the IEEE 1588 PTP protocol) at each node of the pipeline network. This enables millisecond-level time synchronization data acquisition, and the complete pressure time-series data is uploaded to the cloud. The cloud server then uses the synchronized pressure data from all nodes to construct a global model, solving an inverse problem to locate and quantify leaks. Theoretically, this approach offers the highest diagnostic accuracy and can precisely locate leaks at multiple points. For example, Chinese Patent Publication CN111706785A provides a method for identifying leaking pipe sections in a natural gas branched pipeline network. Based on initial and boundary conditions, it constructs a two-fluid model of the pipeline network using mass conservation equations, momentum conservation equations, and energy conservation equations. It uses multiphase flow calculations to obtain real-time changes in parameters such as pressure, temperature, and gas velocity along the pipeline before and after a leak. The pressure change pattern is used as a preliminary basis for judging whether a leak has occurred in the pipeline network, and a curve showing the pressure change amplitude of the leaked pipe section compared to the non-leaking pipe section is plotted. The leaking pipe section is accurately identified based on the rate of pressure amplitude change. Chinese Patent Publication CN112413414A provides a comprehensive detection method for leaks in a heating pipeline network. It determines the leaking pipe section by measuring flow and pressure data from flow meters and pressure transmitters installed on each pipe section. An acoustic sensor is installed on each pipe section, and the leak event is confirmed and the leak point is located based on the acoustic signals detected by the acoustic sensor. The high-precision time synchronization hardware, continuous data transmission bandwidth, and continuous power supply required for this type of solution make it applicable only to newly built IoT pipeline projects, and difficult to apply in the IoT transformation of old pipelines.

[0004] In the IoT transformation of some old heating primary pipelines, low-cost low-power wide-area networks such as NB-IoT are used for status monitoring. This type of solution abandons the strict time synchronization requirement. Each monitoring node independently and asynchronously reports monitoring data such as pressure. The server analyzes and processes the data using timestamps. Although this significantly reduces terminal costs and communication overhead and is feasible in engineering, its technical effectiveness is severely limited: 1) Due to the lack of precise time alignment, it is impossible to reliably calculate the real-time pressure difference between adjacent nodes. Only the delayed and error-prone statistical difference can be used. Affected by normal fluctuations in the pipeline, the monitoring effect is poor, and missed and false alarms cannot be avoided at the same time; 2) Asynchronous reporting causes data to be discrete and misaligned on the time axis, making it impossible to build a relatively accurate global pressure distribution snapshot. This completely loses the ability to perform multi-point leak analysis and cannot distinguish whether an anomaly is caused by a single leak or the result of multiple leaks; 3) It is impossible to distinguish between frequent sensor anomalies caused by cost control and physical leaks in the pipeline. Chinese patent application CN120160087A discloses a method for detecting pipeline leakage pressure. This method uses a ring array of differential pressure sensors to capture pressure gradient changes and combines this with an adaptive adjustment mechanism to achieve accurate leak detection and location. However, in large-scale deployments, this solution suffers from high monitoring terminal costs and high energy consumption, failing to leverage the low-cost advantages of low-power wide-area networks (LPWANs). Summary of the Invention

[0005] The purpose of this disclosure is to provide a monitoring method applicable to heating primary pipeline monitoring systems built on low-power wide area networks, which improves the sensitivity of such systems and reduces the false alarm rate in low-cost deployment and operation scenarios.

[0006] This disclosure, through various demonstrations, primarily provides a method for monitoring primary heating pipe networks. It utilizes a low-power wide-area network (LPWAN) to facilitate data exchange between monitoring terminals and a cloud server. The method clusters the node pressure sequences within each reported batch of pressure monitoring data using one or more aligned time windows based on the server's receiving time. Only node pressure sequences within the same aligned time window are used to calculate a pipe pressure drop characterization sequence. Anomaly detection in the pipe network is based on this pipe pressure drop characterization sequence. The aligned time window is less than a preset threshold. This method senses pipe segment pressure drop through time alignment, particularly using the change and direction of pressure drop per meter calibrated by static pressure difference, to identify currently leaking pipe segments. It also facilitates comparison of average pressure drop across multiple node segments to ensure consistent pressure distribution in continuous water bodies, thus eliminating abnormally reported data.

[0007] In some specific implementations, these methods for monitoring primary heating networks include the following steps:

[0008] Step 10: Obtain first pressure monitoring data asynchronously reported by multiple monitoring terminals in the primary heating network; the first pressure monitoring data includes node pressure sequence pairs reported by different monitoring terminals located on the same unbranched pipe section, with a receiving time difference within a first alignment time window; each node pressure sequence pair corresponds to a sub-pipe section with its monitoring terminal deployment location as the starting and ending point;

[0009] Step 20: For each unbranched pipe segment, select at least one sub-pipe segment's node pressure sequence pair, calculate the corresponding pipe pressure drop difference sequence, and generate a pipe pressure drop characterization sequence to characterize the hydraulic state of the corresponding sub-pipe segment; at the same time, use the pipe pressure drop characterization sequence of the sub-pipe segment to characterize the hydraulic state of the unbranched pipe segment.

[0010] Step 30: Based on the pressure drop characterization sequence of at least one pipe segment without branches, extract a feature vector characterizing the statistical stability of its hydraulic state; the feature values ​​of the feature vector include the mean.

[0011] Step 40: When the receiving time of the node pressure sequence reported by at least one monitoring terminal of the branchless pipe segment is not in the same first alignment time window, based on the degree of deviation between the feature vector and the historical normal state characteristics of the corresponding branchless pipe segment, or the degree of deviation between multiple feature vectors belonging to the same first alignment time window, identify whether it is a preliminary abnormal pipe segment or an abnormal pipe segment; when it is determined to be a preliminary abnormal pipe segment, proceed to step 50; when it is determined to be an abnormal pipe segment, proceed to step 60.

[0012] Step 50: Send a synchronous monitoring instruction to a number of monitoring terminals associated with the initial abnormal pipe section in order to obtain the node pressure sequence of the receiving time difference of all monitoring terminals associated with the initial abnormal pipe section in the second alignment time window in the second pressure monitoring data subsequently reported by the monitoring terminals.

[0013] Step 60: Extract feature vectors from the node pressure sequence pairs of all sub-pipe segments associated with the preliminary abnormal pipe segment and perform a consistency analysis based on spatial correlation to distinguish and locate physical leakage anomalies and monitoring equipment anomalies.

[0014] In one improved implementation, in step 10, each node pressure sequence pair is also used to generate a snapshot of the network status of its corresponding unbranched pipe segment in the first aligned time window.

[0015] A further improvement is that, within the same alignment time window, if different network status snapshots correspond to unbranched pipe segments or interconnected sub-segments, they are merged into a network status snapshot of the connected portion of the network.

[0016] A further improvement is that when the network status snapshots of any connected part of the network in multiple reported batches do not cover a branchless pipe segment, a synchronous monitoring instruction is issued to several monitoring terminals of that branchless pipe segment.

[0017] In another improved implementation, step 10 includes multiple first alignment time windows, and the constraint condition for each first alignment time window further includes: the reception time of the node pressure sequence reported by at least two monitoring terminals located in the same unbranched pipeline network is within the first alignment time window.

[0018] A further improvement is that steps 20 to 40 are executed in parallel for each first alignment time window, and the preliminary abnormal pipe segment selected before issuing the synchronization command in step 50 is integrated and determined according to the preliminary abnormal pipe segment identified in each step 40.

[0019] A further improvement is that the integrated determination method includes: if there is one pipeline status snapshot that does not indicate its abnormality based on all pipeline status snapshots containing the initially abnormal pipeline segment, then its abnormal status is canceled.

[0020] In another improved implementation, in step 50, if one of the node pressure sequence pairs used to determine the preliminary abnormal pipe segment originates from a monitoring terminal located at a branch point, then a synchronization monitoring command is also issued to a monitoring terminal adjacent to that monitoring terminal that does not belong to the preliminary abnormal pipe segment; in step 60, the consistency analysis based on spatial correlation also includes the node pressure sequence pairs of that monitoring terminal and adjacent monitoring terminals that do not belong to the preliminary abnormal pipe segment.

[0021] Further improvements in the above aspects include the following: In step 60, the method for distinguishing and locating physical leakage anomalies and monitoring equipment anomalies includes the following steps: comparing the feature vector obtained by the pipe pressure drop characterization sequence of all sub-pipe segments associated with the initially abnormal pipe segment with the historical normal state characteristics of the corresponding sub-pipe segment; when the anomaly score of a sub-pipe segment and any sub-pipe segment covering or adjacent to that sub-pipe segment is greater than a threshold, marking that sub-pipe segment as having an abnormal leakage; when the anomaly score of all sub-pipe segments covering or adjacent to that sub-pipe segment is less than a threshold, marking that the monitoring terminal sensor of that sub-pipe segment as having an anomaly.

[0022] In some implementations of the technical solutions disclosed herein, the significant increase in cost and power consumption of individual terminals due to high-precision time synchronization hardware is avoided as much as possible. In other implementations, targeted model design avoids the high communication bandwidth requirements of existing synchronization models for the transmission of continuous pressure data, keeping communication costs low. For some IoT applications with high-precision deployment and high maintenance complexity, the main cost of the technical solutions disclosed herein lies in equipment configuration, especially the server side. It is suitable for the asynchronous communication and low data stability hardware systems used in the current renovation of old pipeline networks. Some improved implementations can approach the diagnostic accuracy of existing synchronization solutions. Attached Figure Description

[0023] For those skilled in the art, the clear description of various embodiments in this disclosure is sufficient for them to understand the scope of the technical solutions claimed in this disclosure. The accompanying drawings described below are merely some exemplary specific embodiments and technical aspects of this disclosure. Other drawings can be obtained based on these drawings without any creative effort. The following is a brief introduction to the drawings used in the description of the specific embodiments. Obviously, since each drawing only describes one technical aspect, when its description is used in conjunction with other drawings or technical aspects to explain multiple aspects of the implementation in different specific embodiments, the content shown in the drawings has a distinguishing scope of reference when understood in context.

[0024] Figure 1 This is a schematic diagram of the deployment of monitoring terminals in a primary heating network monitoring system in one embodiment;

[0025] Figure 2 This is a schematic diagram of the deployment of monitoring terminals in a primary heating network monitoring system, as shown in another embodiment.

[0026] Figure 3 This is a schematic diagram of the time distribution of the first pressure monitoring data in one embodiment;

[0027] Figure 4 This is a schematic diagram of a method for monitoring a primary heating network in one embodiment;

[0028] Figure 5 This is a schematic diagram of the initial abnormal pipe section pressure gradient distribution in one embodiment;

[0029] Figure 6 This is a schematic diagram of the time distribution of the second pressure monitoring data in one embodiment;

[0030] Figure 7 This is a schematic diagram of a node pressure sequence time alignment method in one embodiment;

[0031] Figure 8 This is a schematic diagram of the initial abnormal pipe section pressure gradient distribution in another embodiment. Detailed Implementation

[0032] First, it should be noted that phrases such as "in one embodiment" or "in an embodiment" in this specification do not necessarily refer to the same embodiment, but rather provide specific technical aspects for combining with a particular embodiment, wherein specific features, structures, or characteristics can be combined in any suitable manner consistent with this disclosure. The terms "comprising" and "including" are open-ended, as used in the claims, and do not exclude additional structures or steps. Consider the following cited claim: "A method comprising one or more of the following steps..." Such claims do not exclude the method from including additional steps (e.g., parameter initialization and other execution steps at the transport layer, network layer, user layer, physical layer, etc.). A carrier "configured to" perform one or more tasks or task steps can be various terminals, servers, or other devices. In such a context, "configured to" imply a structure (e.g., circuitry) by indicating that the terminal / server / computer device includes structures (e.g., circuitry) that perform the one or more tasks and task steps by its processing unit during operation. Thus, the terminal / server / computer device is allegedly configured to perform the task or specific steps within the task even when the specified terminal / server / computer device is currently inoperable or not running (e.g., not connected). Terminal / server / computer devices used with the language “configured as” include hardware—such as circuits, memory storing executable program instructions to perform operations, etc. Furthermore, “configured as” can include general-purpose structures (e.g., general-purpose circuits) manipulated by software or firmware (e.g., FPGAs or general-purpose processors executing software) to operate in a manner capable of performing one or more tasks to be solved. “Configured as” can also include adjusting manufacturing processes (e.g., semiconductor fabrication facilities) to manufacture devices (e.g., integrated circuits) suitable for implementing or performing one or more tasks. As used herein, indicative terms such as “first” and “second” act as labels for the nouns preceding them and do not imply any type of ordering (e.g., spatial, temporal, logical, etc.). For example, a terminal / server or the task and task-specific characteristics of a terminal / server configured to perform may be described herein as the execution of a “first” task / step / algorithm and a “second” task / step / algorithm. The terms “first” and “second” do not necessarily imply that the second algorithm must be executed before the first algorithm. As used herein, the terms “based on,” “according to,” or “depending on” are used to describe one or more factors influencing a determination, and these terms do not exclude additional factors that may influence the determination. That is, the determination may be based solely on these factors or at least in part on these factors. Consider the phrase “A is determined based on B,” in which case B is a factor influencing the determination of A. Such phrases do not exclude that the determination of A may also be based on C, and in other instances, A may be determined solely on B. When used in the claims, the term “or” is used as an inclusive or, not an exclusive, or.For example, the phrase "at least one of x, y, or z" means any one of x, y, and z, and any combination thereof.

[0033] It should also be noted that low-power wide-area network (LPWAN) terminal devices typically reduce total power consumption by maintaining a long-term low-power deep sleep state, such as the PSM operation mode of NB-IoT and the state after closing the receive window in LoRa's Class A mode. This is characterized by shutting down the power supply to functional modules such as the radio frequency (RF) module and relying on the terminal's local on-chip real-time clock (RTC) for timing. Maintaining this long-term low-power deep sleep state results in a significant cumulative clock drift between the on-chip clock and the standard clock provided by satellite timing or NTP services. Since the RF module's power consumption accounts for 50% to 70% of the overall power consumption of the monitoring equipment, in general engineering projects, monitoring equipment typically operates by avoiding additional time calibration after waking from the low-power deep sleep state. For example, in existing NB-IoT-based heating network monitoring systems, the terminal typically does not perform time calibration during the sampling, transmission, and then sleep process after a timed interrupt wake-up. Instead, it performs individual actual calibrations at a uniform cycle, such as once a day or once a week. This approach can meet the needs of single-point acquisition scenarios with low real-time requirements, but it is difficult to meet the needs of scenarios requiring multi-point pressure difference monitoring of long-distance pipeline sections. One technical solution proposes a communication method for monitoring long-distance heating pipelines. This method uses a low-power wide-area network (LPWAN) to achieve data exchange between monitoring terminals and a cloud server. Specifically, the cloud server defines and stores spatial pairing relationships between monitoring terminals based on the pipeline topology. For each pair of spatially paired monitoring terminals, the cloud server calculates the time difference between the received monitoring data. If the time difference exceeds a second tolerance threshold, the cloud server generates a control command and sends it to at least one of the terminals in the pair, causing it to adjust its stored transmission time offset. This maintains the time difference between the received monitoring data reported by the spatially paired terminals within a predetermined first tolerance threshold. This method uses a quasi-synchronous approach to keep the overall power consumption of the monitoring terminals from radio frequency (RF) at a low level while satisfying the allowable error during the differential alignment of some monitoring terminal sampling data. However, due to cost control, on the one hand, with the reduction of time synchronization control commands, the allowable window for time difference may be preset to a large value, the distribution of received time is scattered, and it is difficult to cluster data from multiple pipe sections in the same time window. When comprehensively analyzing problems such as leakage in upstream and downstream pipe sections, existing synchronous real-time data analysis methods cannot be directly applied. On the other hand, there is also the problem of data failure when the sensor works for a long time or is subjected to interference, which affects the analysis and judgment conclusions.

[0034] Since the method disclosed herein is applied to a type of pipeline monitoring system, the technical problems faced by those skilled in the art can be specifically illustrated through a general embodiment. Figure 1The diagram illustrates a deployment scheme for monitoring terminals in a primary heating network monitoring system. Monitoring nodes P1, P2…P21… are located near the intersection of the supply and return water pipelines, upstream of the pipeline, at the supply inlet / outlet, or in the middle of long straight pipe sections—locations with suitable installation conditions. These nodes possess stable geographical location information, including elevation. The monitoring nodes are connected by both supply and return water pipelines. Hollow arrows in the diagram indicate the heating direction, while solid arrows indicate the actual pipelines and the flow direction of the supply and return water media. Monitoring terminals E1, E2…E40… are deployed at each monitoring node. Each monitoring terminal is independent, with its own power supply, on-chip clock, and RF feeder.

[0035] In this embodiment, P1 is the monitoring node connected to the heat source interface, and P13, P29, and P32 are monitoring nodes located at the heat exchange station. Each of the unbranched pipe sections P1-P8, P8-P13, P8-P25, P25-P29, and P25-P32 contains more than three monitoring nodes, and at least one monitoring terminal is deployed on both the supply and return water pipes of each monitoring node. For monitoring nodes P8 and P25 at the pipe network intersection, the monitoring terminals can be deployed on the upstream main pipe in the heating direction of that monitoring node. Due to pipe connectivity, the pressure near each flow channel of the tee is relatively close; in some other embodiments, the terminal can be deployed on one of the branch pipes depending on the site installation conditions. In other embodiments, where the branch node experiences significant pressure drop due to structural issues or has a large amount of water hammer, a monitoring terminal can be deployed on each of the three pairs of pipes. In this disclosure, the specific meaning of unbranched pipe sections does not particularly depend on the physical pipe section division; it can also be divided based on the average pressure drop caused by pipe resistance. (Example) Figure 1 In the P1-P8 pipe section, due to differences in construction plans or inherent structures, during the shutdown and maintenance period, the pressure drop per meter of the P1-P2, P2-P3, and P3-P4 pipe sections, after deducting the elevation static pressure difference and calculated based on the planned pipe length, is basically the same under the same flow velocity conditions. At the same time, the pressure drop per meter of the P4-P5, P5-P6, P6-P7, and P7-P8 pipe sections, after deducting the elevation static pressure difference and calculated based on the planned pipe length, is also basically the same. Therefore, P1-P4 can be divided into a non-branching pipe section, and P4-P8 is a non-branching pipe section. Obviously, a non-branching pipe section is either a water supply pipe or a return pipe. When a monitoring node is used to represent a non-branching pipe section, it refers to one of the water supply pipe or the return pipe. Figure 1 The diagram only shows the case when the primary pipeline network is a continuously pressure-distributed water body. In actual operation, the primary pipeline network is affected by physical factors such as the presence of relay pump stations or load regulation of heating stations, etc. Figure 2Nodes P4 and P26 are deployed at relay booster pump stations, while node P8, due to its physical structure being significantly affected by water hammer reflection, exhibits poor continuity in upstream and downstream water pressure distribution. Therefore, more terminals are deployed on the supply or return water pipelines, resulting in inconsistent pipeline segment divisions for supply and return water. For example, on the server, the supply pipeline between nodes P1-P4 and P4-P8 is allocated as a single unbranched segment, as is the return water pipeline between nodes P1-P8. The proposed pipe length is a server-preset parameter. It can be directly derived from the physical pipeline length obtained through actual engineering surveying or a geographic information system, or it can be a converted unbranched pipe length obtained by converting the average resistance and total resistance under typical operating conditions within the measured flow velocity range under normal maintenance conditions, or through other conversion methods. Clearly... Figure 1 , 2 In this invention, odd-numbered monitoring terminals are deployed on the water supply pipeline, and even-numbered monitoring terminals are deployed on the return pipeline. Monitoring terminals are deployed near the starting point and end of each unbranched pipe section. For ease of understanding, unless otherwise specified, in the descriptions of various parts of this disclosure, "system" refers to the heating primary pipeline network monitoring system applicable to this method, which uses a low-power wide area network to achieve data exchange between monitoring terminals and a cloud server. "Terminal" and "server" refer to the monitoring terminal and cloud server in this system, respectively. "Node" refers to the monitoring node where the terminal is deployed. "Pipeline" refers to a unidirectional physical pipeline for water supply or return that passes through one or more nodes. "Pipe section" refers to a connected pipeline with definite endpoints, obtained by logically dividing the physical pipeline according to resistance parameters and configured in the server. Since pipeline branching generally involves significant changes in resistance characteristics, these pipe sections are usually unbranched.

[0036] Based on the above or similar deployment schemes, in some instances, monitoring terminals are configured to use a time-window sampling strategy to report pressure detection data. Examples include... Figure 2 As shown, Figure 1 Monitoring terminals E1, E2...Ei are deployed at different monitoring nodes or on different pipelines. According to the preset configuration, a specific sampling window should be selected during the terminal sampling phase. Each process completes one sampling during its own runtime, followed by a fixed time interval. Process and The network unit is constantly woken up to notify the core network to prepare for uplink transmission. For example, it can be configured to sample at equal intervals from 2026-07-15T13:41:30Z (UTC) to 2026-07-15T13:44:30Z (UTC), compressing data in real time during sampling. After interception and communication packetization, transmission begins at 2026-07-15T13:45:00Z (UTC) using a communication interrupt. However, in actual operation, the actual sampling window varies depending on whether it's measured by the network clock or the cloud system's standard clock. , ... And their respective actual sampling window end time and their respective pre-configured transmission time. , ... The basically fixed on-chip clock interval between them, or the time lead (TA) used to ensure that the monitoring terminal completes one sampling, compression and transmission preparation. , ... None of them will be the same as the system preset. It's easy to understand that the asynchronicity of a low-power wide-area network (LPWAN) system is mainly determined by the following factors: First, it is affected by the local clock drift of the monitoring terminal; although their respective pre-configured transmission times are all set to the preset values... The time is as follows: 2026-07-15T13:45:30Z (UTC). However, due to cumulative errors such as the on-chip clock reference clock and clock drift, the actual sampling start time of each monitoring terminal varies. Secondly, in half-duplex mode, the uplink must wait for a network idle signal, and the actual start time of reporting after the network unit wakes up has an uncertain delay. Finally, due to low-bandwidth relay forwarding, the actual successful reception time of the server is also random due to the influence of the network itself. This results in a wide time distribution range for the reception time of the node pressure sequence reported by all terminals in the same reporting period on the server, even if all terminals have the same sampling and reporting time during configuration. Although these node pressure sequences reported in the same batch are not synchronized, they still constitute a batch of pressure monitoring data for all terminals. In this disclosure, the different distributions of reception time in a batch of pressure monitoring data are processed accordingly. To show the difference in distribution characteristics, "first / second" is used for distinction.

[0037] Some leak detection theories, based on the Darcy-Weisbach equation and fluid pressure gradient distribution theory, involve continuously deploying multiple monitoring nodes along long-distance pipeline segments. Then, by calculating the pressure gradient distribution from a static pressure starting point, the direction of pressure gradient change in each segment is observed to identify relevant leak points. Some real-time systems obtain a snapshot of the pipeline pressure distribution by monitoring the water pressure values ​​collected by the monitoring nodes at the same time, thus implementing related theoretical solutions. However, due to the asynchronous nature of low-power wide-area network transmission and common sensor interference problems in engineering, existing analysis techniques struggle to utilize the asynchronously reported pressure monitoring data from these monitoring terminals.

[0038] Figure 3As shown, this disclosure provides a novel method for monitoring heating pipe networks through multiple embodiments, applicable to hydraulically continuous closed systems such as primary heating pipe networks. Based on existing low-power wide-area network monitoring technology, this method clusters the node pressure sequences contained in each reported batch of pressure monitoring data using one or more aligned time windows based on their server reception time. Only node pressure sequences within the same aligned time window are used to calculate a pipe pressure drop characterization sequence. Anomaly judgments for the pipe network are based on this pipe pressure drop characterization sequence. The aligned time window is less than a preset threshold. After appropriately adjusting the sampling time window and the allowable aligned time window, due to the continuous fluid flow in most primary pipe networks, data within the same aligned time window exhibit the same fluctuation trend, possessing time alignment capability. By aligning and subtracting the node pressure sequences of unbranched pipe sections and sub-pipe sections with hydraulic state continuity, a difference sequence characterizing related pipe sections can be obtained, thus enabling partial synchronous analysis. Some specific implementations include steps 10 to 60.

[0039] Step 10: Obtain the first pressure monitoring data asynchronously reported by multiple monitoring terminals in the primary heating network. The first pressure monitoring data includes node pressure sequence pairs reported by different monitoring terminals located on the same branchless pipe segment, with a reception time difference within a first alignment time window. Each node pressure sequence pair corresponds to a sub-pipe segment with its monitoring terminal deployment location as the starting and ending point. In the exemplary example, a sufficient number of monitoring terminals are deployed at predefined intervals in each branchless pipe segment. At least for a period of time after all monitoring terminals in the system are synchronized, the reception time distribution of the node pressure sequences reported by each monitoring terminal in each reporting cycle is within a small time window. This distribution is mainly caused by network random delay. Subsequently, as the local real-time clock of each terminal shifts, the time window of the reception time distribution gradually increases. Because a sufficient number of monitoring terminals are deployed on each branchless pipe segment—for example, when segmenting according to monitoring nodes such as P1-P4, P4-P8, P8-P13, P8-P25, P25-P29, and P25-P32—and at least two pressure monitoring terminals are deployed on the return water pipe segments between each node, an aligned time window can inevitably be found in each reporting cycle based on the reported node water pressure sequence and the receiving time. This ensures that the receiving times of at least two monitoring terminals on each branchless pipe section of the return water are within this window. For example, at least two terminals in E2, E4, E6, and E8 must have their receiving times within this window, and at least two terminals in E8, E10, E12, E14, and E16 must also have their receiving times within this window, and so on. This window is the first alignment time window, and the pressure monitoring data containing the pressure sequences of all nodes is the first pressure monitoring data. This alignment time window is used to create a snapshot of the pipe network status that can be correlated and analyzed. The midpoint of this alignment time window can be used as the timestamp of its pipe network status snapshot. In some dynamic time window schemes, when the alignment time window gradually increases, making it impossible to achieve meaningful time series alignment of the node pressure sequences to obtain a sufficiently long pipe pressure drop difference sequence, a time synchronization command should be issued to the relevant monitoring terminals. In some static time window schemes, when the number of monitoring terminals meeting the conditions within the alignment time window gradually decreases to a certain extent, for example, only two monitoring terminals on the branchless pipe section with the fewest monitoring terminals meet the conditions, a time synchronization command should be issued to the relevant monitoring terminals. Time synchronization commands can be delayed, and unbranched pipe segments that do not meet the conditions will not affect the subsequent processing of other unbranched pipe segments that do meet the conditions. For unbranched pipe segments with two or more terminals, the pipe segments between adjacent terminals can be regarded as a sub-segment of the unbranched pipe segment.

[0040] Step 20: For each unbranched pipe segment, select at least one sub-segment's node pressure sequence pair, calculate the corresponding pipe pressure drop difference sequence, and generate a pipe pressure drop characterization sequence to represent the hydraulic state of the corresponding sub-segment; simultaneously, use the pipe pressure drop characterization sequence of the sub-segment to characterize the hydraulic state of the unbranched pipe segment. In Step 10, this is actually a clustering and classification of the first pressure monitoring data. A first alignment time window is used as the classification radius. For a set of node pressure sequences reported by each terminal included in the first pressure monitoring data, those received within the window are marked as the first sequence, and those outside the window are marked as the second sequence. Appropriate selection of the classification radius ensures that each unbranched pipe segment of a connected local pipe network in a given network has node pressure sequences reported by two or more terminals. When the window is smaller than the maximum time threshold required for time alignment of two node pressure sequences, time alignment of the two node pressure sequences can be performed to obtain a difference sequence representing the time change of pipe pressure drop between the two nodes over a period of time. The time-aligned sequence difference method can refer to any existing technology and be improved. The method for obtaining some difference sequences, particularly applicable to the technical solutions of this disclosure, can be demonstrated in the sequence alignment section of this disclosure. For the obtained pipe pressure drop sequence of an unbranched pipe segment, based on the node elevation difference, temperature difference, and the proposed pipe segment length, an average pipe pressure drop sequence per meter representing the water state of the unbranched pipe segment is obtained, i.e., the pipe pressure drop characterization sequence. When an unbranched pipe segment has multiple average pipe pressure drop sequences per meter, the one with the longest alignment length can be used as the optimal pipe pressure drop characterization sequence representing the hydraulic state of the unbranched pipe segment. Since the pipe pressure drop characterization sequence generally takes a preset medium flow direction, i.e., according to the server's preset direction, the upstream node pressure minus the downstream node pressure, it is generally positive and can be directly normalized to obtain characteristic values. During some shutdown maintenance periods, when checking for leaks in sections, negative values ​​may occur when calculating according to the preset direction because the pressurized water has no clear flow direction. In step 20, a branchless pipe segment may have multiple pipe pressure drop characterization sequences of different sub-pipe segments in the same first alignment window or different first alignment windows. Due to the asynchronous transmission characteristics, most of the time, it is difficult for the reporting and receiving times of all terminals of the branchless pipe segment to be in the same alignment time window. Due to the continuity of hydraulic state, the pipe pressure drop characterization sequence of any of its sub-pipe segments can be used as the pipe pressure drop characterization sequence of the branchless pipe segment in subsequent analysis for calculation.

[0041] Step 30: Based on at least one pipe pressure drop characterization sequence of an unbranched pipe section, extract a feature vector representing the statistical stability of its hydraulic state; the feature value of the feature vector includes the mean. It is easy to understand that each pipe pressure drop characterization sequence carries information about the operating characteristics of its unbranched pipe section, which can be used to extract feature values ​​reflecting the statistical stability of the relevant hydraulic state. These include sequence features independent of the sampling method, such as the sequence mean reflecting the average hydraulic gradient or resistance condition, the sequence standard deviation reflecting pressure drop fluctuations, and the sequence coefficient of variation reflecting pressure drop stability under low pressure differential conditions. When the sampling window is long, it also includes long-range features indicating conditions such as poor insulation and air accumulation in the pipe, such as the first-order autocorrelation coefficient and the detrended fluctuation analysis index. And when the sampling intervals are dense, it also includes statistical features that can capture transient disturbance anomalies, such as kurtosis, skewness, and entropy.

[0042] Step 40: When the receiving time of the node pressure sequence reported by at least one monitoring terminal of an unbranched pipe segment is not within the same first alignment time window, based on the degree of deviation of the feature vector from the historical normal state features of the corresponding unbranched pipe segment, or the degree of deviation between multiple feature vectors belonging to the same first alignment time window, identify whether it is a preliminary abnormal pipe segment or an abnormal pipe segment; if it is determined to be a preliminary abnormal pipe segment, proceed to step 50; if it is determined to be an abnormal pipe segment, proceed to step 60. It is easy to understand that the anomaly judgment in this step is mainly based on the terminal sampling scheme. After obtaining at least one pipe pressure drop characterization sequence for each unbranched pipe segment, the feature vector of its hydraulic state is calculated. After comparing it with a feature vector of the normal hydraulic state as a benchmark, it is determined whether an anomaly exists. The feature vector of the normal hydraulic state as a benchmark can be obtained from the feature vectors of multiple operating conditions under the historical normal state of the corresponding pipe segment. Typical methods for judging deviation include calculating the Mahalanobis distance using the mean vector of these historical normal state feature vectors and the covariance matrix of the current feature vector obtained from the monitoring data. When the Mahalanobis distance exceeds a threshold, an anomaly is output.

[0043]

[0044] in, This is a feature vector extracted from a specified unbranched pipe segment pressure drop characterization sequence used to identify anomalies. This is the mean vector of multiple feature vectors extracted from historical normal state data of the corresponding historical pipe section under different operating conditions. To and covariance matrix Regularization bias processing.

[0045] Other commonly used methods based on feature space deviation judgment include automatic relation decoupling combined with T. 2Principal component analysis (PCA) of statistics and SPE (or Q) statistics, cosine similarity focusing on shape analysis, and single-class support vector machines or isolated forests based on machine learning for anomaly boundary delineation.

[0046] The method for determining the degree of deviation also applies to situations where the same unbranched pipe segment has multiple feature vectors. By comparing each feature vector, the consistency of hydraulic state can be used to determine if there are significant deviations. When a significant deviation is found when comparing multiple feature vectors within the same alignment time window for the same unbranched pipe segment, especially when the terminal corresponding to a feature vector covers most of the sub-segments of the unbranched pipe segment, the unbranched pipe segment can be directly marked as an abnormal segment. Otherwise, due to poor system data quality, further temporary synchronization is required.

[0047] Step 50 involves sending synchronization monitoring commands to several monitoring terminals associated with the initially abnormal pipe section. This allows the acquisition of the node pressure sequence within the second alignment time window, representing the reception time difference of all monitoring terminals associated with the initially abnormal pipe section in the second pressure monitoring data subsequently reported by the monitoring terminals. Each terminal in the system is configured not to actively synchronize its time within the radio frequency communication window, but only when the server issues a time synchronization control command, in order to reduce the power consumption of the IoT sensing layer terminals. For the initially abnormal pipe section, due to measurement errors in the terminals deployed in its upstream and downstream pipe sections, the anomaly may not be detected in time in step 40, or it may be a false alarm, actually reflecting anomalies in other upstream and downstream pipe sections. Furthermore, the initially abnormal pipe section refers to the entire pre-divided, branchless pipe section; even if a leak occurs within this section, the more specific location of the anomaly cannot be determined. Therefore, after identifying the initial abnormal pipe segment, it is necessary to re-perform quasi-synchronous verification. This involves sending time synchronization instructions to more terminals in the unbranched pipe segment and adjacent upstream and downstream pipe segments. This ensures that in the second pressure monitoring data uploaded by these terminals in the next reporting cycle, the pressure sequences of each node are within an allowed second alignment time window at the time of reception by each server. This means that, referring to steps 10 and 20, these node pressure sequences can be aligned and subtracted to obtain the pipe pressure drop difference sequence. In a demonstrative implementation, such as... Figure 5 As shown, Figure 1The normal pressure gradient distribution curve of the P8-P25 return water pipe section where E42 and E46 are located is L0. At a certain moment during the periodic execution of steps 10 to 40, a leak occurs at point "X" in the pipe between E44 and E46. Based on a first alignment time window, the pressure sequences reported by nodes E42 and E46 are filtered to obtain the pipe pressure drop characterization sequence of the P8-P25 return water pipe section. Based on the pressure sequences reported by nodes E42 and E46 and the pipe pressure drop characterization sequence, the inferred pressure gradient distribution curve can be obtained as L1. After executing the current step 40, the P8-P25 return water pipe section is marked as a preliminary abnormal pipe section. Based on the deviation between the data reception time of E16, E44, E48, and E50 and the point in the first alignment time window during this reporting cycle, a relative time synchronization signal is sent to E16, E44, E48, and E50 to reduce the time deviation between the second pressure data reception time sent by these terminals in the next reporting cycle and the point in the second alignment time window obtained in the next clustering. Figure 6 As shown, after time synchronization of local terminals in the network, the reception times of E16, E42, E44, E46, E48, and E50 are all within the second aligned time window. Within this window, the pressure sequences of each node can be aligned and subtracted. Some methods can also expand the scope of time synchronization commands, even traversing all terminals in the perception layer to send network protocol time synchronization commands.

[0048] In some embodiments, during step 10, the asynchronous node pressure sequence is clustered into multiple first alignment windows. Different first alignment windows have different feature vectors for the same branchless pipe segment. Steps 20 to 40 are executed in parallel for each first alignment time window. The initially abnormal pipe segment selected before issuing the synchronization command in step 50 is integrated and determined based on the initially abnormal pipe segments identified in each step 40. In some integration determinations, if there is one pipe network status snapshot containing the initially abnormal pipe segment that has not been flagged as abnormal, its abnormal status is canceled. Others consider the time continuity of non-recoverable abnormalities; when the first alignment time windows are relatively dispersed, the data from the last first alignment time window for the branchless pipe segment is used.

[0049] In some embodiments, when the pipe pressure drop characterization sequence of the unbranched pipe segment used to determine the initial abnormal pipe segment comes from a nearby pipe branch point, the terminal issuing the synchronous detection command also includes a sub-pipe segment connected to the initial abnormal pipe segment in the adjacent unbranched pipe segment. For example, Figure 1In the P4-P8 return water pipe section where E14 and E16 are located, at a certain moment during the periodic execution of steps 10 to 40, the pressure sequences of the nodes reported by E14 and E16 are selected according to a first alignment time window to obtain the pipe pressure drop characterization sequence of the P4-P8 return water pipe section. When an anomaly is inferred from the pressure sequences of the nodes reported by E14 and E16 and their pipe pressure drop characterization sequences, due to the proximity of the branch point, the terminal issuing the synchronous monitoring command should include E18 and E42, so that the pressure sequences of each node of other terminals on the P4-P8 return water pipe section are simultaneously in the second alignment time window in the next reporting cycle. By performing consistency analysis on the network status snapshots of the P9-P8 return water sub-pipe section, the P21-P8 return water sub-pipe section, and the P4-P8 pipe section in the second alignment time window, specific analytical conclusions can be obtained.

[0050] Step 60: Extract feature vectors from the node pressure sequence pairs of all sub-pipe segments associated with the initial abnormal pipe segment and perform a consistency analysis based on spatial correlation to distinguish and locate physical leakage anomalies and monitoring equipment anomalies.

[0051] This disclosure does not require that terminals be deployed at both ends of a branchless pipe segment, such as Figure 2 P8 nodes E16 and E16 ’ It has a large voltage drop, but for systems with this deployment, Figure 1 In the case of node P8, where branch point pressure data is shared, another specific example considering this situation involves the following steps: In step 50, if one of the node pressure sequence pairs used to determine the initial abnormal pipe segment originates from a monitoring terminal located at a branch point, a synchronization monitoring command is also issued to a monitoring terminal adjacent to that terminal but not belonging to the initial abnormal pipe segment. In step 60, the consistency analysis based on spatial correlation also includes the node pressure sequence pairs between that monitoring terminal and adjacent monitoring terminals not belonging to the initial abnormal pipe segment. Based on this second pressure monitoring data, feature vectors are extracted from the node pressure sequence pairs of the initial abnormal pipe segment and its adjacent continuous sub-segments, and then spatial correlation analysis is performed to distinguish and locate physical leakage anomalies and monitoring equipment malfunction anomalies.

[0052] In some implementations, step 60, the method for distinguishing and locating physical leakage anomalies and monitoring equipment anomalies, includes the following steps: comparing the feature vector obtained by comparing the pipe pressure drop characterization sequence of all sub-pipe segments associated with the initially abnormal pipe segment with the historical normal state characteristics of the corresponding sub-pipe segment; when the anomaly score of a sub-pipe segment and any sub-pipe segment covering or adjacent to that sub-pipe segment is greater than a threshold, the sub-pipe segment is marked as having an abnormal leakage; when the anomaly score of all sub-pipe segments covering or adjacent to that sub-pipe segment is less than a threshold, the monitoring terminal sensor of that sub-pipe segment is marked as having an anomaly.

[0053] In a specific example, step 10 includes the following steps:

[0054] Step 11: The server continuously receives data packets asynchronously reported from each terminal. Each data packet contains the terminal ID, batch code, pressure compression summary sequence, and temperature. The server obtains the reception time of data packets within the same batch and decompresses and resamples the pressure compression summary sequence into node pressure sequences. A feasibility check is performed on the set of node pressure sequences. Only when the feasibility check is satisfied is the set of node pressure sequences considered as a first set of pressure monitoring data for subsequent processing. The formal expression of the feasibility check is as follows: For a set of m unbranched pipe segments in a primary heating network… Each pipe section It has Each monitoring terminal has its current batch data packet reception time, i.e., the set of node pressure sequence reception times. The preset upper limit of the first alignment time window width is Define the first alignment time window as an interval If and only if At least two reception times of this pipe section fall within the specified time range. Inside, referred to as pipe section window Coverage, for each pipe segment Check if a pair of times exists satisfy If it does not exist, then the pipe segment cannot be replaced by any pipe with a width less than [a certain value]. The alignment time window coverage, when at least one pipe segment in the primary heating network set can be covered by a width of When the alignment time window is covered, the set of pressure sequences of that node is a set of first pressure monitoring data.

[0055] Step 12: Generate the first aligned time window set for this set of first pressure monitoring data. Specifically, this is formally expressed as finding a set of first aligned time windows using the following constraints. :

[0056]

[0057] Clearly, a sliding search using the maximum-width window to traverse the pipe segments can obtain a set of first-aligned time windows. To improve the connected portion of the pipe network in subsequent pipe network status snapshots, preferably, the objective function is set as follows:

[0058]

[0059] This combinatorial optimization problem can be solved by using a greedy strategy to reduce complexity and obtain approximately the minimum number of first alignment time windows.

[0060] In a specific example, step 20 includes the following steps:

[0061] Step 21: Within each first alignment time window, find the node pressure sequence pairs of each unbranched pipe segment covered by it, perform time alignment, and calculate the pipe pressure drop difference sequence. Although the sampling window and sampling interval are the same for each terminal, due to the uncontrollable differences in the sampling start time and the different durations corresponding to the actual compressed pressure summary sequence obtained by the server after compression, these node pressure sequence pairs cannot be directly subtracted and require time alignment. An example alignment operation based on time dynamic programming includes the following steps:

[0062] For node pressure sequence pairs and Resample using the same fixed sampling interval to re-obtain the time series pair and , and The resampling interval is redefined. It is easy to understand that this resampling satisfies the maximum error condition during sampling compression.

[0063] For sequence and Alignment is performed using dynamic time planning to obtain the optimal warping path. .like Figure 7 As shown, Time series and The distance matrix is ​​measured in Euclidean distance, where:

[0064]

[0065] make Distance matrix The dynamic time-bending path on, where It is the first of the paths There are elements, and they satisfy boundary, continuity, and monotonicity constraints:

[0066]

[0067] It's easy to understand that there are multiple dynamic time-warped paths on the distance matrix that satisfy the conditions. Among all dynamic time-warped paths, the sum of DTW distances, for example, when using Euclidean distance as a metric, is... The path that recursively minimizes the curve is the optimal curved path, which this disclosure refers to as the aligned path. The sum of the DTW distances along this path can be denoted as:

[0068]

[0069] Generally, there is a relatively stable difference between the two first sequences. This difference includes the static pressure difference and the dynamic pressure difference related to the flow velocity in the long-distance pipeline. The distance in formula (4) can also be measured using the Manhattan distance metric after deducting the static pressure difference, for example:

[0070]

[0071] in, To provide The location of the monitoring terminal points to the provided The static pressure difference at the monitoring terminal location is such that the actual sampling duration corresponding to the first sequence received by the cloud server is basically close to the first alignment time window preset by the monitoring terminal. The duration of the signal is generally several times the actual maximum water hammer wave period in the field. Therefore, the actual alignment is based on the delay transmission of the low-frequency pressure wave in the pipeline between the matching monitoring terminals.

[0072] Trim away the "gap" portions at the beginning and end of the aligned path with a slope of 0 or 1 to obtain a continuous path. This continuous path is the node pressure sequence pair. and The aligned portion obtained by time series alignment corresponds to the duration of the alignment. and Based on the duration corresponding to the resampling time interval. This part has a defined duration and corresponding relationship, depending on the pipeline design flow direction, such as... The collection location is Upstream of the sampling location, the sequence of tube pressure drop differences for the aligned portions of the two sequences is represented as follows:

[0073]

[0074] When the static pressure difference needs to be subtracted from the pipe pressure drop difference sequence, it can be expressed as:

[0075]

[0076] It's easy to understand that time-based dynamic programming algorithms are existing technology. If the alignment effect is adjusted to reduce alignment error or lower the complexity of the alignment algorithm, the steps mentioned above, such as the distance matrix, can be improved. For example, the Needleman-Wunsch algorithm, which has better fitting results for multiple line segments, could be used. Another continuous path... The existence of multiple alignment examples with slopes of 0 or 1, representing local one-to-many or many-to-one "gaps," suggests that these gaps may contain resampled, unseparated high-frequency components. Since multiple consecutive high-frequency waves can create long gaps along a continuous path, a sliding window approach can also be used to obtain the aligned portion of the node pressure sequence pairs for the purpose of aligning low-frequency waveforms. An example of a time alignment operation using a sliding window includes the following steps:

[0077] Pairs of node pressure sequences and Resampling was performed separately to generate sequences. and Select the range of sliding window lengths. Traverse the sequence over its entire length. and The pursuit Previous child window and Previous child window Normalized cross-correlation coefficient:

[0078]

[0079] in, and These are the means of the two sub-windows, respectively. The closer the value is to 1, the more similar the two sub-windows are. This operation is equivalent to dividing a window of length L into separate sequences. and Swipe up to see more. Each starting position in ,and Each starting position in Perform pairing scoring. It's easy to understand that this example uses normalized cross-correlation coefficients as the similarity score. Other examples may use metrics such as cosine similarity as the alignment score. Some examples using deep learning models for alignment may also set conventional long-range or short-range cues to achieve time series alignment.

[0080] Find the one among all the calculated normalized cross-correlation coefficients that makes The largest Triples, at this time the window subsequence in and This refers to the aligned portion. The voltage drop difference sequence of the aligned portions of the two sequences is represented as follows:

[0081]

[0082] Step 22: Obtain the proposed pipe lengths of the two terminal deployment locations of the reported node pressure sequence pair. Based on the proposed pipe lengths and the pipe pressure drop difference sequence obtained in Step 21 after deducting the static pressure difference, calculate the pipe pressure drop characterization sequence used to characterize the hydraulic state of the unbranched pipe section where the two terminals are located.

[0083] Step 23, Create the pipe segment First alignment time window A snapshot of the pipeline network status, in one example, structured tuples The data storage format for a snapshot of the pipeline network status can be represented as:

[0084]

[0085] in, A unique identifier for an unbranched pipe segment. The start and end times of the first alignment time window. The pipe section obtained from the above calculation First alignment time window A sequence of pipe pressure drop characterizations, From The feature vectors extracted from it.

[0086] It's easy to understand that when two or more are in the same alignment time window... Snapshots corresponding pipe section When the pipeline networks are interconnected in terms of physical topology, they can be merged and stored as a composite pipeline state snapshot of a connected component C. :

[0087]

[0088] in, To connect the pipe sections The collection. In some embodiments, network status snapshots of individual pipe segments or merged connected parts of the network obtained from different batches of pressure monitoring data are used to refresh the network status. In some implementations, these network status snapshots are used for synchronization rate control; when the network status snapshots of connected parts of the network in multiple aligned time windows do not cover a branchless pipe segment, a synchronization monitoring command is issued to several monitoring terminals of that branchless pipe segment. In some implementations, network status snapshots of connected parts of the network carrying connectivity relationships can also be used for the analysis of a time slice of the local network it covers.

[0089] refer to Figure 5 , 8 In one implementation, Figure 1The normal pressure gradient distribution curve of the P8-P25 return water pipe section where E42 and E46 are located is L0. At a certain moment when performing steps 10 to 40 periodically, the pressure sequence of nodes reported by E42 and E46 is selected according to a first alignment time window to obtain the pipe pressure drop characterization sequence of the P8-P25 return water pipe section. The inferred pressure gradient distribution curve L1 can be obtained when the abnormality is inferred from the pressure sequence of nodes reported by E42 and E46 and the pipe pressure drop characterization sequence. At this point, one possibility is that a leak has occurred at point "X" in the pipeline between E44 and E46. After executing the current step 40, the P8-P25 return water pipeline section is marked as a preliminary abnormal pipeline section. Based on the deviation between the data reception time of E16, E44, E48, and E50 and the point in the first alignment time window in this reporting cycle, a relative time synchronization signal is sent to E16, E44, E48, and E50 to reduce the time deviation between the second pressure data reception time sent by these terminals in the next reporting cycle and the point in the second alignment time window obtained in the next clustering. In step 60, the pressure sequences of each node in this window can be aligned and subtracted to form a snapshot of the pipeline network status including this pipeline section. In step 60, if the eigenvectors of the pipe pressure drop characterization sequence from its sub-pipe segment lack consistency, it can be determined that a leak has occurred in this pipe segment. Furthermore, based on the simulated pressure distribution curve L3 formed by backtracking the average pressure drop of adjacent terminals, a more specific sub-pipe segment is identified. If the eigenvectors of the pipe pressure drop characterization sequence from its sub-pipe segment lack consistency but are concentrated at a single free point (as provided by E46), after deducting this free point, the consistency of the eigenvectors of the pipe pressure drop characterization sequences of all sub-pipe segments is restored, and they all align with L3. ’ If the actual pipeline pressure distribution is close, it may only be a terminal anomaly. In this case, another terminal anomaly handling procedure can be executed, and it is not necessary to report a pipeline segment anomaly. If the characteristic vector of the pipeline pressure drop characterization sequence of its sub-pipeline segments after time calibration is restored to consistency, it may be a short-term false alarm, and the anomaly can be eliminated.

[0090] In a specific example of the consistency analysis of this disclosure, for each sub-pipe segment, the pipe pressure drop characterization sequence is calculated using the nodal pressure sequences of its two endpoints, and the corresponding consistent feature vector is extracted as described in step 30. To simplify the example, this implementation uses the mean as the key feature (obviously, variance, skewness, etc., could be added to form a multi-dimensional feature vector). Then, from historical normal data, the mean and distribution (e.g., standard deviation σ distribution) of each sub-pipe segment's features are obtained. The degree of deviation of the current state is quantified by calculating the standardized deviation between the current feature value and the historical benchmark. A spatial consistency test is used to calculate the deviation consistency index of a sub-pipe segment and all its neighbors as an anomaly score. In some examples, when the deviation consistency index is abnormal, its deviation is checked to see if it is "continuous" with its neighbors. If not, the deviation gradient with each neighbor is calculated. When the deviation consistency index is abnormal, and at least one neighbor is also abnormal, and the deviation gradient is small, it indicates a smooth spatial transition of the deviation, and is judged as a leakage anomaly. When the deviation consistency index is abnormal, but all its neighbors are normal, i.e., the deviation gradient changes abruptly at this point, it is judged as a leakage anomaly. In some examples, a consistency score is directly defined, and the eigenvector weights are set according to the pipe segment length or hydraulic impedance. Then, the average distance between this eigenvector and the standardized deviation of all associated sub-pipe segments is calculated. The smaller the distance, the more consistent it is with its neighbors. In this case, if the standardized deviation is large but the consistency score is small, the anomaly is likely part of a continuous physical phenomenon (leakage); if the standardized deviation is large but the consistency score is large, the anomaly is isolated and more likely points to anomalies in the monitoring equipment data at its endpoint. In some anomaly type localization and output leak localization processes, one or more connected subgraphs with large deviation values ​​and low consistency scores are identified. The physical area covered by this connected subgraph is the suspected leak impact range, and the node with the largest deviation value is likely near the leak point. For locating equipment failures, isolated sub-pipe segments with large deviation values ​​but high consistency scores are identified, and the faulty equipment is located at one of the two endpoints of this sub-pipe segment. Further, the relationship between the absolute pressure values ​​at the two endpoints of this sub-pipe segment and neighboring nodes can be compared, or the historical self-diagnostic data of the equipment (such as battery voltage and signal strength) can be checked to finally determine the terminal of the anomaly source. During further iteration and confirmation, if the problem is determined to be a device malfunction, the diagnostic results can suggest remote restarting, calibration, or triggering on-site maintenance. If the problem is determined to be a leak, more detailed transient pressure wave analysis or acoustic monitoring can be used for precise location.

[0091] Obviously, in step 60, for cases where the physical structure of the pipe pressure drop does not change abruptly, the various sub-pipe segments will overlap, i.e., cover or partially cover each other. For terminals A, B, C, and D that are deployed sequentially with the same unbranched pipe segment, sub-pipe segment AC obviously covers sub-pipe segment AB and partially covers sub-pipe segment BD. The above scoring system can also compare the feature vector obtained by comparing the pipe pressure drop characterization sequence of all sub-pipe segments associated with the initial abnormal pipe segment with the historical normal state characteristics of the corresponding sub-pipe segment. When the abnormal score of a sub-pipe segment and any sub-pipe segment covering or adjacent to that sub-pipe segment is greater than a threshold, the sub-pipe segment is marked as having an abnormal leak. When the abnormal score of all sub-pipe segments covering or adjacent to that sub-pipe segment is less than a threshold, the sensor of the monitoring terminal of that sub-pipe segment is marked as abnormal.

[0092] The foregoing examples illustrate the deployment and configuration of some hardware platforms for monitoring primary heating networks upon which the communication methods provided in this disclosure are based. These examples are intended only to assist those skilled in the art in understanding the technical effects achieved by the technical solutions of this disclosure and do not limit the direct or improved implementation of these technical solutions in other similar deployment and configuration schemes in the art. Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above-described device configuration and execution steps and system application methods.

[0093] It is easy to understand that the various specific embodiments disclosed herein include communication methods for monitoring long-distance heating pipelines. In different aspects of the specific embodiments, based on the emphasis of the described technical aspects, other contents that can be easily understood or known from the prior art or the content disclosed herein are omitted.

Claims

1. A method for monitoring a primary heating network, characterized in that, it enables data exchange between a monitoring terminal and a cloud server based on a low-power wide area network, and is further characterized in that, For each batch of pressure monitoring data, the node pressure sequence is clustered using one or more aligned time windows based on its server reception time. The pipe pressure drop characterization sequence is calculated only for the node pressure sequences within the same aligned time window. The anomaly judgment of the pipeline network is calculated based on the pipe pressure drop characterization sequence. The aligned time window is less than a preset threshold.

2. The method for monitoring primary heating pipelines according to claim 1, comprising the following steps: Step 10: Obtain first pressure monitoring data asynchronously reported by multiple monitoring terminals in the primary heating network; the first pressure monitoring data includes node pressure sequence pairs reported by different monitoring terminals located on the same unbranched pipe section, with a receiving time difference within a first alignment time window; each node pressure sequence pair corresponds to a sub-pipe section with its monitoring terminal deployment location as the starting and ending point; Step 20: For each unbranched pipe segment, select at least one sub-pipe segment's node pressure sequence pair, calculate the corresponding pipe pressure drop difference sequence, and generate a pipe pressure drop characterization sequence to characterize the hydraulic state of the corresponding sub-pipe segment; at the same time, use the pipe pressure drop characterization sequence of the sub-pipe segment to characterize the hydraulic state of the unbranched pipe segment. Step 30: Extract a feature vector representing the statistical stability of the hydraulic state based on the pressure drop characterization sequence of at least one pipe segment without branches. The eigenvalues ​​of the feature vector include the mean; Step 40: When the receiving time of the node pressure sequence reported by at least one monitoring terminal of the branchless pipe segment is not in the same first alignment time window, based on the degree of deviation between the feature vector and the historical normal state characteristics of the corresponding branchless pipe segment, or the degree of deviation between multiple feature vectors belonging to the same first alignment time window, identify whether it is a preliminary abnormal pipe segment or an abnormal pipe segment; when it is determined to be a preliminary abnormal pipe segment, proceed to step 50; when it is determined to be an abnormal pipe segment, proceed to step 60. Step 50: Send a synchronous monitoring instruction to a number of monitoring terminals associated with the initial abnormal pipe section in order to obtain the node pressure sequence of the receiving time difference of all monitoring terminals associated with the initial abnormal pipe section in the second alignment time window in the second pressure monitoring data subsequently reported by the monitoring terminals. Step 60: Extract feature vectors from the node pressure sequence pairs of all sub-pipe segments associated with the preliminary abnormal pipe segment and perform a consistency analysis based on spatial correlation to distinguish and locate physical leakage anomalies and monitoring equipment anomalies.

3. The method for monitoring primary heating pipelines according to claim 2, characterized in that, In step 10, each node pressure sequence pair is also used to generate a snapshot of the network status of its corresponding unbranched pipe segment in the first alignment time window.

4. The method for monitoring primary heating pipelines according to claim 3, characterized in that, If different network status snapshots within the same alignment time window correspond to unbranched pipe segments or interconnected sub-segments, they are merged into a single network status snapshot of the connected portion of the network.

5. The method for monitoring primary heating networks according to claim 4, characterized in that, When the network status snapshots of any connected part of the pipeline network in multiple reported batches do not cover a branchless pipeline segment, a synchronous monitoring instruction is issued to several monitoring terminals of that branchless pipeline segment.

6. The method for monitoring primary heating pipelines according to claim 2, characterized in that, Step 10 includes multiple first alignment time windows, and the constraint condition for each first alignment time window further includes: the reception time of the node pressure sequence reported by at least two monitoring terminals located in the same branchless pipeline network is within the first alignment time window.

7. The method for monitoring primary heating pipelines according to claim 6, characterized in that, Steps 20 to 40 are executed in parallel for each first alignment time window. The preliminary abnormal pipe segment selected before issuing the synchronization command in step 50 is integrated and determined according to the preliminary abnormal pipe segment identified in each step 40.

8. The method for monitoring primary heating networks according to claim 7, characterized in that, The integrated judgment method includes: if there is one pipeline status snapshot that does not indicate its abnormality based on the total pipeline status snapshot containing the initially abnormal pipeline segment, then its abnormal status is canceled.

9. The method for monitoring primary heating pipelines according to claim 2, characterized in that, In step 50, if one of the node pressure sequence pairs used to determine the preliminary abnormal pipe segment originates from a monitoring terminal located at a branch point, then a synchronization monitoring command is also issued to a monitoring terminal adjacent to that monitoring terminal that does not belong to the preliminary abnormal pipe segment; in step 60, the consistency analysis based on spatial correlation also includes the node pressure sequence pairs of the monitoring terminal and the adjacent monitoring terminals that do not belong to the preliminary abnormal pipe segment.

10. The method for monitoring primary heating pipelines according to claim 2 or 9, characterized in that, In step 60, the method for distinguishing and locating physical leakage anomalies and monitoring equipment anomalies includes the following steps: comparing the feature vector obtained by the pipe pressure drop characterization sequence of all sub-pipe segments associated with the initially abnormal pipe segment with the historical normal state characteristics of the corresponding sub-pipe segment; when the anomaly score of a sub-pipe segment and any sub-pipe segment covering or adjacent to that sub-pipe segment is greater than a threshold, the sub-pipe segment is marked as having an abnormal leakage; when the anomaly score of all sub-pipe segments covering or adjacent to that sub-pipe segment is less than a threshold, the monitoring terminal sensor of that sub-pipe segment is marked as having an anomaly.

Images