Remote waterlogging warning method and system for water meter wells based on the Internet of Things

Abstract

The present application belongs to the field of intelligent early warning technology, and discloses a remote waterlogging early warning method and system for water meter wells based on the Internet of Things; the method comprises: integrating the collected water meter well measurement data through an Internet of Things gateway, comprehensively analyzing the measurement data, obtaining comprehensive features, using the measurement data and the comprehensive features as inputs of a waterlogging analysis model, and obtaining a waterlogging feature pattern; combining the acoustic positioning derivative features in the comprehensive features, and determining the coordinates of the leakage point through a fusion positioning algorithm; dynamically analyzing the measurement data, obtaining derivative dynamic features, using the measurement data and the derivative dynamic features as inputs of a dynamic feature model, and obtaining dynamically adjusted graded waterlogging early warning thresholds; and performing corresponding graded early warnings based on the relationship between the actual amount of waterlogging and the graded waterlogging early warning thresholds; the present application constructs a closed-loop system integrating real-time monitoring, abnormal diagnosis, precise positioning, and graded early warnings through multi-dimensional sensor data fusion and intelligent feature engineering.

Description

Technical Field

[0001] The present application relates to the field of intelligent early warning technology, and more specifically, to a method and system for remote waterlogging early warning in water meter wells based on the Internet of Things. Background Art

[0002] With the acceleration of urbanization, the scale of urban water supply pipelines continues to expand. Water meter wells, as key nodes in the pipeline network, face the hidden dangers of water accumulation caused by pipeline aging, external water seepage, drainage blockage, etc. Traditional manual inspections are inefficient and have poor real-time performance. The early warning method based on the static threshold of a single sensor is easily affected by environmental fluctuations, and there are problems such as high false alarm rate, vague cause identification, and rough leakage point positioning.

[0003] Chinese patent application publication number CN112712680A discloses a remote voice warning box, warning method, and smart water meter for a smart water meter: the remote voice warning box includes at least a warning broadcast module, an information processing module, and a communication module; the warning method includes: the communication module receives warning information sent by the smart water meter and sends the received warning information to the information processing module, the warning information including at least the remaining prepaid amount and / or remaining water volume; the information processing module receives the warning information, generates warning broadcast information based on the warning information, and sends the warning broadcast information to the warning broadcast module; the warning broadcast module receives the warning broadcast information and broadcasts it according to the content of the warning broadcast information to provide a warning reminder to the user. This invention effectively improves water quality and user experience, avoiding the difficulties or inconveniences caused by water outages.

[0004] Although the above method can meet most scenarios, research and practical application of the above method and existing technology have found that the above method and existing technology have at least the following defects:

[0005] The data perception dimension of the above method is single and cannot identify the cause of waterlogging.

[0006] In view of this, the present application proposes a remote waterlogging warning method and system for water meter wells based on the Internet of Things to solve the above problems. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present application provides the following technical solution: a remote waterlogging early warning method for water meter wells based on the Internet of Things, comprising the following steps:

[0008] The collected water meter well measurement data is integrated through the IoT gateway, and the measurement data is comprehensively analyzed to obtain comprehensive features. The measurement data and comprehensive features are used as inputs to the waterlogging analysis model to obtain the waterlogging feature pattern.

[0009] Comprehensive features include sensor cross-features, time series dynamic features, spatial correlation features, and environmental features; sensor cross-features include threshold trigger features, correlation features, and physical derivative features; time series dynamic features include short-term window features, frequency domain features, and state transition features; spatial correlation features include spatial gradient features and acoustic wave positioning derivative features;

[0010] Combined with the derived features of acoustic wave positioning, the coordinates of the leak point are determined through a fusion positioning algorithm;

[0011] The measured data are dynamically analyzed to obtain derived dynamic features. The measured data and derived dynamic features are used as inputs of the dynamic feature model to obtain dynamically adjusted graded waterlogging warning thresholds. Corresponding graded warnings are then made based on the relationship between the actual amount of waterlogging and the graded waterlogging warning thresholds.

[0012] Furthermore, the method for obtaining the threshold trigger feature includes:

[0013] Detect whether the pipeline pressure change value exceeds the preset pressure threshold, detect whether the difference between the total flow value and the user end flow value exceeds the preset flow threshold; detect whether the water level rise rate exceeds the preset rate threshold; if all three are exceeded at the same time, mark the pipeline rupture risk as 1, otherwise it is 0; use the pipeline rupture risk as the threshold trigger feature;

[0014] Methods for obtaining correlation features include:

[0015] The Pearson correlation coefficient was used to calculate the correlation between different sensors and obtain the correlation characteristics.

[0016] Furthermore, the method of obtaining the physical derivative feature includes:

[0017] Obtain the total flow at the user end and the flow at the pipeline inlet, and calculate the leakage rate; count the pressure change and flow change, and calculate the joint feature; concatenate the leakage rate and the joint feature to obtain the physical derivative feature.

[0018] Furthermore, the method for obtaining the state transition feature includes:

[0019] A threshold is preset to classify the sensor states into normal, warning, and fault, and the data is encoded in sequence. The sensor's time series data is converted into a state sequence. The number of transitions from state a to state b is counted, the total number of stays in state a is calculated, and the ratio of the number of transitions to the total number of stays is calculated to obtain a transition probability matrix. The continuous stay duration of each state is counted, and the transition probability matrix and the continuous stay duration are concatenated to obtain statistical features.

[0020] Count the average time interval from state a to state b, count the number of occurrences of the preset transition pattern, and concatenate the average time interval and the number of occurrences to obtain the temporal dependency feature;

[0021] Obtain the sensor's change rate and change amplitude during state transition, and whether the states of other sensors change when the sensor's change amplitude exceeds a preset change threshold; combine the sensor's change rate and change amplitude, and whether the states of other sensors change when the sensor's change amplitude exceeds a preset change threshold, to obtain dynamic trend characteristics;

[0022] Statistical features, time series dependency features, and dynamic trend features are combined to obtain state transition features.

[0023] Furthermore, the method for obtaining the temporal dynamic characteristics includes:

[0024] The mean, standard deviation, and rate of change of the water level are calculated. The minimum value, fluctuation range, and number of times the pressure change per unit time in the pipeline pressure is lower than the pressure threshold are calculated. The mean, standard deviation, and rate of change of the water level, as well as the minimum value, fluctuation range, and number of times the pressure change per unit time in the pipeline pressure is lower than the pressure threshold are concatenated to obtain short-term window characteristics.

[0025] Statistically obtain the deviation between the actual temperature value and the historical average temperature for the same period, the deviation between the actual humidity value and the historical average humidity for the same period, and the proportion of low-frequency vibration energy below the preset low-frequency threshold. The deviation between the actual temperature value and the historical average temperature for the same period, the deviation between the actual humidity value and the historical average humidity for the same period, and the proportion of low-frequency vibration energy below the preset low-frequency threshold are spliced ​​as long-term window features;

[0026] The main frequency energy of the vibration signal and the high-frequency noise power of the pipeline pressure signal are extracted through Fourier transform, and the main frequency energy and high-frequency noise power are spliced ​​as frequency domain features;

[0027] The short-term window features, long-term window features, frequency domain features and state transition features are spliced ​​together to obtain the time series dynamic features.

[0028] Furthermore, the method of obtaining spatial correlation features includes:

[0029] Calculate the water level difference and spacing between adjacent water meter wells, calculate the ratio of the water level difference and spacing, and obtain the water level gradient;

[0030] Calculate the pressure difference and distance between the upstream water meter well and the downstream water meter well, calculate the ratio of the pressure difference to the distance, and obtain the pressure decay rate;

[0031] The water level gradient and pressure decay rate are spliced ​​together to obtain the spatial gradient characteristics;

[0032] Detect the time difference between any two water meter wells receiving the water leakage sound wave, calculate the ratio of the difference in received signal strength and the distance between the two water meter wells, and obtain the signal attenuation rate; splice the time difference and the signal attenuation rate to obtain the derived characteristics of the acoustic wave positioning;

[0033] The spatial gradient features and sonar localization derived features are spliced ​​together to obtain spatial correlation features.

[0034] Furthermore, the method for obtaining environmental characteristics includes:

[0035] Count the number of historical waterlogging events at the same location, and concatenate the coordinates corresponding to the location with the historical waterlogging events to serve as historical waterlogging features. Count the number of failures of similar equipment, and concatenate the equipment model and failure times as equipment failure features. Concatenate historical waterlogging features and equipment failure features to obtain environmental features.

[0036] Furthermore, the method for obtaining the leak point coordinates includes:

[0037] Obtain the coordinates of the water meter well and establish a hyperbolic positioning model based on the time difference of the acoustic wave positioning derivative features;

[0038] Obtain the signal strength received by the water meter well, establish an empirical attenuation model of signal strength with respect to distance, and calculate the distance from the leak point to the water meter well using the empirical attenuation model. Establish three circle equations using the coordinates of the three water meter wells and the corresponding distances.

[0039] The hyperbolic positioning model and the three circle equations are fused by the least squares method;

[0040] The maximum likelihood estimation or Kalman filter algorithm is used to iteratively optimize the coordinates of the leakage point.

[0041] Furthermore, the method for obtaining the derived dynamic features includes:

[0042] Obtain the water level change, temperature change, and humidity change within a preset time period, calculate the ratio of the water level change, temperature change, and humidity change to the preset time period, and obtain the water level change rate, temperature change rate, and humidity change rate;

[0043] Calculate the water leakage probability index by combining the pressure drop amplitude and flow difference;

[0044] Calculate the difference between the current pipeline pressure and the preset pipeline pressure to obtain the pressure change amplitude; calculate the ratio of the difference to the preset change time to obtain the pressure drop rate; and statistically obtain the standard deviation of pressure fluctuations;

[0045] Obtain the difference between the user-end traffic and the pipeline inlet traffic, calculate the ratio of the difference to the pipeline inlet traffic, and obtain the traffic difference ratio;

[0046] Obtain the absolute difference between the current user end traffic and the user end traffic at the previous moment, calculate the ratio of the absolute difference to the user end traffic at the previous moment, and obtain the traffic mutation coefficient;

[0047] Obtaining the amount of change in the accumulated water depth within a preset change time, calculating the ratio of the amount of change in the accumulated water depth to the preset change time, and obtaining the rate of change of the accumulated water depth;

[0048] Obtaining a high-frequency vibration signal amplitude when the vibration signal exceeds a preset high-frequency threshold;

[0049] Get whether the temperature change rate exceeds the preset temperature change rate threshold. If so, mark it as 1, otherwise it is 0, and obtain the temperature anomaly index;

[0050] Get whether the humidity change rate exceeds the preset humidity change rate threshold. If so, mark it as 1, otherwise it is 0, and obtain the humidity anomaly index;

[0051] The water level change rate, temperature change rate, humidity change rate, pressure fluctuation standard deviation, flow difference ratio, flow mutation coefficient, water accumulation depth change rate, high-frequency vibration signal amplitude, temperature anomaly index and humidity anomaly index are spliced ​​to obtain derived dynamic characteristics.

[0052] The remote waterlogging early warning system for water meter wells based on the Internet of Things implements the remote waterlogging early warning method for water meter wells based on the Internet of Things, including:

[0053] Feature analysis module: This module integrates the collected water meter well measurement data through the IoT gateway, performs comprehensive analysis on the measurement data, obtains comprehensive features, and uses the measurement data and comprehensive features as inputs to the waterlogging analysis model to obtain the waterlogging feature pattern.

[0054] Comprehensive features include sensor cross-features, time series dynamic features, spatial correlation features, and environmental features; sensor cross-features include threshold trigger features, correlation features, and physical derivative features; time series dynamic features include short-term window features, frequency domain features, and state transition features; spatial correlation features include spatial gradient features and acoustic wave positioning derivative features;

[0055] Leak point location module: Combines the derived features of acoustic wave positioning and determines the coordinates of the leak point through a fusion positioning algorithm;

[0056] Waterlogging warning module: Dynamically analyzes the measurement data to obtain derived dynamic features, uses the measurement data and derived dynamic features as inputs to the dynamic feature model to obtain dynamically adjusted graded waterlogging warning thresholds; and performs corresponding graded warnings based on the relationship between the actual waterlogging volume and the graded waterlogging warning thresholds.

[0057] The technical effects and advantages of the IoT-based water meter well remote waterlogging warning method and system are as follows:

[0058] This application uses multi-dimensional sensor data fusion and intelligent feature engineering to build a closed-loop system that integrates real-time monitoring, abnormal diagnosis, precise positioning and graded warning. Through cross-sensor threshold joint judgment, correlation analysis and time series state transition modeling, it accurately identifies waterlogging patterns, improves the accuracy of cause diagnosis, and reduces false alarm and missed alarm rates. By integrating the hyperbola-triangulation positioning algorithm of acoustic wave time difference and signal strength, combined with least squares and Kalman filter optimization, it improves the accuracy of leak point positioning and shortens maintenance response time. Based on the dynamic graded threshold model of derived dynamic features, it automatically adapts to environmental changes such as season and load to achieve multi-level precise warning, synchronously associates historical high-frequency areas of waterlogging with equipment failure characteristics, promotes the concentration of operation and maintenance resources in high-risk areas, reduces inspection costs, and significantly improves the intelligent management level of urban water pipe networks and the efficiency of emergency response to waterlogging incidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 Schematic diagram of the remote waterlogging warning method for water meter wells based on the Internet of Things in this application;

[0060] Figure 2 This is a schematic diagram of the water accumulation location and warning process for this application;

[0061] Figure 3 This is the structural diagram of the water meter well remote waterlogging warning system based on the Internet of Things for this application. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0063] Example 1

[0064] See also Figure 1 、 Figure 2 As shown, this embodiment provides a remote waterlogging warning method for water meter wells based on the Internet of Things, including the following steps:

[0065] Combined with the measurement data collected by sensors from water meter wells, the measurement data are comprehensively analyzed to obtain comprehensive features. The measurement data and comprehensive features are used as inputs to the water accumulation analysis model to obtain water accumulation characteristic patterns; for example, pattern 1 (pipeline rupture): sudden pressure drop + abnormal flow + continuous water level rise; pattern 2 (drainage blockage): continuous water level rise + stable humidity + no pressure anomalies; pattern 3 (external water seepage): synchronous changes in temperature and humidity + no pipeline parameter anomalies.

[0066] By deploying multiple sensors within water meter wells, including water level, temperature, humidity, pressure, flow, and vibration, these sensors collect real-time multidimensional data, including well water depth, environmental parameters, and pipeline operating status. After data cleaning and feature engineering, comprehensive features are generated, including cross-sensor correlation, temporal dynamic trends, and spatial coordination. These raw data and comprehensive features are then fed into a waterlogging analysis model, enabling accurate identification of characteristic patterns such as pipe ruptures, drainage blockages, and external water seepage. A remote waterlogging early warning system, built on this foundation, leverages dynamic thresholds and a tiered early warning mechanism to match data with pre-set patterns in real time, rapidly locating leak coordinates and visualizing them on a GIS map. This system also triggers SMS / APP notifications and generates repair work orders. This technological system represents an upgrade from manual inspections to intelligent monitoring, significantly improving the accuracy of waterlogging diagnosis and response efficiency, effectively reducing risks such as water waste and road collapses caused by pipeline leaks. It provides data-driven decision support for refined management and preventive maintenance of urban water networks, and contributes to the development of a smart and intensive drainage safety assurance system.

[0067] Comprehensive features include sensor cross-features, time series dynamic features, and spatial correlation features; sensor cross-features include threshold trigger features, correlation features, and physical derivative features; time series dynamic features include short-term window features, frequency domain features, and state transition features; spatial correlation features include spatial gradient features and acoustic wave positioning derivative features;

[0068] Methods for obtaining sensor cross-signaling include:

[0069] Detect whether the pipeline pressure change value exceeds the preset pressure threshold, detect whether the difference between the total flow value and the user-end flow value exceeds the preset flow threshold; detect whether the water level rise rate exceeds the preset rate threshold; if all three are exceeded at the same time, mark the pipeline rupture risk as 1, otherwise it is 0; use the pipeline rupture risk as the threshold trigger feature; pipeline rupture risk is judged through multiple thresholds of pressure, flow, and water level, which can accurately capture the coupling relationship of key indicators in high-risk scenarios such as "pipeline rupture", avoid misjudgment of a single sensor, directly output a clear risk status, and provide a strong discriminative Boolean feature for the early warning model;

[0070] The Pearson correlation coefficient is used to calculate the correlation between different sensors and obtain correlation characteristics. The Pearson correlation coefficient is used to quantify the linear relationship between sensors, which can explore the potential dependencies between data, help distinguish the causes of waterlogging, and enrich the feature dimensions of model input.

[0071] Methods for obtaining physically derived features include:

[0072] Obtain the total flow at the user end and the flow at the pipeline inlet, and calculate the leakage rate; count the pressure change and flow change, and calculate the joint feature; concatenate the leakage rate and the joint feature to obtain the physical derivative feature.

[0073] The threshold trigger feature uses multiple thresholds of pressure, flow, and water level to jointly judge the risk of pipeline rupture, directly outputting a clear risk status in the form of Boolean values, accurately capturing the coupling relationship of key indicators in high-risk scenarios, effectively avoiding misjudgment of a single sensor, providing strong discriminant features for the early warning model, and improving the ability to identify core risk scenarios such as "pipeline rupture"; the correlation feature uses the Pearson correlation coefficient to quantify the linear correlation between sensors, explore the potential dependencies of the data, and assist in distinguishing the causes of waterlogging (such as pipeline rupture, rainstorm waterlogging, and other different scenarios), enriching the feature dimensions of the model input, and enabling the early warning model to analyze the causes of waterlogging from a multivariate correlation perspective; the physical derivative feature calculates the joint features of leakage rate, pressure and flow changes and splices them to form an indicator with both physical meaning and comprehensive characterization capabilities, converting basic sensor data into in-depth features reflecting the operating status of the pipeline network, and enhancing the model's understanding of physical processes related to waterlogging (such as leakage degree and fluid dynamics changes). The three work together to build a multi-dimensional input system from the three levels of risk status identification, data correlation analysis, and physical process characterization, thereby improving the accuracy, robustness, and scenario differentiation capabilities of the early warning model, and providing a solid feature foundation for accurate early warning of remote waterlogging in water meter wells.

[0074] Methods for obtaining temporal dynamic characteristics include:

[0075] The mean, standard deviation, and rate of change of the water level are calculated. The minimum value, fluctuation range, and number of times the pressure change per unit time in the pipeline pressure is lower than the pressure threshold are calculated. The mean, standard deviation, and rate of change of the water level, as well as the minimum value, fluctuation range, and number of times the pressure change per unit time in the pipeline pressure is lower than the pressure threshold are concatenated to obtain short-term window characteristics.

[0076] The short-term window feature integrates real-time statistics of water level and pipeline pressure, accurately capturing the short-term fluctuation trend of water depth and the high-frequency sudden state of pressure anomalies, providing "immediate abnormal signals" for real-time warning;

[0077] Statistically obtain the deviation between the actual temperature value and the historical average temperature for the same period, the deviation between the actual humidity value and the historical average humidity for the same period, and the proportion of low-frequency vibration energy below the preset low-frequency threshold. The deviation between the actual temperature value and the historical average temperature for the same period, the deviation between the actual humidity value and the historical average humidity for the same period, and the proportion of low-frequency vibration energy below the preset low-frequency threshold are spliced ​​as long-term window features;

[0078] The long-term window feature filters out short-term noise and uncovers long-term trend anomalies by analyzing the deviation of temperature and humidity from the historical mean values ​​for the same period and the proportion of low-frequency vibration energy, improving the system's ability to identify gradually changing waterlogging scenarios.

[0079] The main frequency energy of the vibration signal and the high-frequency noise power of the pipeline pressure signal are extracted through Fourier transform, and the main frequency energy and high-frequency noise power are spliced ​​as frequency domain features;

[0080] Frequency domain features extract the vibration main frequency energy and pressure high-frequency noise power through Fourier transform, distinguish different abnormal modes from the frequency dimension, and enhance the feature decoupling capability of complex working conditions;

[0081] Short-term window features, long-term window features, frequency domain features, and state transition features are combined to obtain time series dynamic features. These features integrate short-term dynamics, long-term trends, frequency domain characteristics, and state transition processes (e.g., the temporal dependency from normal to warning). This allows the waterlogging analysis model to respond to both real-time anomalies (e.g., a sudden pressure drop and water level surge in a short-term window) and learn long-term evolution patterns (e.g., long-term humidity deviation and low-frequency vibration anomalies indicate pipeline aging and water seepage). This allows for more accurate identification of the temporal evolution paths of characteristic patterns such as "pipeline rupture-drainage blockage-external water seepage," effectively improving the early warning system's detection sensitivity for multi-timescale anomalies and its ability to dynamically diagnose the causes of waterlogging. This provides both real-time and predictive time series feature support for remote early warning.

[0082] Methods for obtaining state transition characteristics include:

[0083] A threshold is preset to classify sensor states into normal, warning, and fault, and the data is encoded sequentially. The sensor's time series data is converted into a state sequence. The number of transitions from state a to state b is counted, the total number of stays in state a is calculated, and the ratio of the number of transitions to the total number of stays is calculated to obtain a transition probability matrix. The continuous stay duration in each state is counted. The transition probability matrix and the continuous stay duration are concatenated to obtain statistical features. The statistical features quantify the sensor's transition patterns between states such as "normal-warning-fault". Combined with the state stay duration, they can help the model identify typical state sequence patterns that cause waterlogging.

[0084] The average time interval from state a to state b is calculated, as is the number of occurrences of a preset transition pattern. The average time interval and the number of occurrences are then combined to obtain a time-dependent feature. This feature captures the temporal logic of state transitions and, through the high frequency of preset transition patterns (such as the chain transition of "abnormal vibration → sudden pressure drop → sudden water level increase"), accurately locates the time chain of sudden water leaks, improving the real-time nature of early warnings.

[0085] The sensor's change rate and magnitude during state transitions are obtained, as well as whether other sensors' states change when the sensor's magnitude exceeds a preset threshold. The sensor's change rate and magnitude, as well as other sensors' states change when the sensor's magnitude exceeds a preset threshold, are combined to obtain dynamic trend features. Dynamic trend features verify the physical consistency of state transitions, ensuring that abnormal states are caused by real waterlogging events rather than noise interference, thereby enhancing the reliability of early warnings.

[0086] State transition features are obtained by combining statistical features, time-dependent features, and dynamic trend features. These features integrate the statistical laws, time-dependent relationships, and dynamic collaborative trends of state transitions, enabling the waterlogging analysis model to learn the complete evolutionary path from "normal operation → early warning → failure outbreak." This not only captures current state anomalies in real time but also predicts escalating waterlogging risks through state transition patterns. This enables the precise triggering of graded warnings (such as mild / moderate / severe), the rapid location of leak causes, and the early initiation of emergency responses. This significantly improves the warning system's adaptability to complex operating conditions and its ability to manage waterlogging incidents in a closed-loop manner.

[0087] Methods for obtaining spatial association features include:

[0088] Calculate the water level difference and spacing between adjacent water meter wells, calculate the ratio of the water level difference and spacing, and obtain the water level gradient;

[0089] Calculate the pressure difference and distance between the upstream water meter well and the downstream water meter well, calculate the ratio of the pressure difference to the distance, and obtain the pressure decay rate;

[0090] The water level gradient and pressure decay rate are combined to obtain spatial gradient features. Spatial gradient features can quantify the spatial distribution anomalies of water accumulation depth and pipeline pressure, helping the model identify the spatial clustering of abnormal water accumulation and along-the-pipeline changes in operating conditions at a regional scale, avoiding misjudgment of single-well data.

[0091] The time difference between any two water meter wells receiving the leak sound wave is detected, and the ratio of the difference in received signal strength to the distance between the two water meter wells is calculated to obtain the signal attenuation rate. The time difference and signal attenuation rate are then combined to obtain the derived acoustic wave positioning feature. The derived acoustic wave positioning feature is directly related to the propagation characteristics of the leak sound wave. By combining the time difference in sound wave arrival between multiple wells and the signal strength attenuation law, the coordinates of the leak point are accurately calculated using a triangulation positioning algorithm. The coordinates are then visually annotated on a GIS map, achieving the integration of "anomaly detection, cause analysis, and location positioning."

[0092] Spatial gradient features and acoustic positioning derivative features are combined to obtain spatial correlation features. Spatial correlation features fuse spatial distribution gradients with acoustic positioning information, enabling the waterlogging analysis model to both determine the spatial correlation of regional waterlogging at a macro level and accurately locate leaks at a micro level. This significantly improves the early warning system's adaptability to complex pipe network environments, providing a dual, precise basis of "spatial location + anomaly type" for rapid repair dispatch and mitigation of leak impacts, and promoting the efficient closed-loop upgrade of waterlogging early warning from "status-based warning" to "positioning and disposal."

[0093] Methods for obtaining environmental characteristics include:

[0094] Count the number of historical waterlogging events at the same location, and concatenate the coordinates corresponding to the location with the historical waterlogging events to serve as historical waterlogging features. Count the number of failures of similar equipment, and concatenate the equipment model and failure times as equipment failure features. Concatenate historical waterlogging features and equipment failure features to obtain environmental features.

[0095] Historical flooding characteristics quantify the historical risk level of a specific area, enabling the early warning system to dynamically adjust monitoring strategies for "old problem spots" and identify recurring flooding risks in advance. Equipment failure characteristics uncover common equipment defects, helping the model assign higher anomaly weights to specific equipment models, achieving "device-level precise risk assessment."

[0096] Feature splicing and integration combines the historical waterlogging patterns of spatial locations with the failure probability of equipment types, providing the waterlogging analysis model with a dual environmental context of "regional risk + equipment reliability". This allows the model to not only identify anomalies in current real-time data, but also provide early warnings based on historical high-incidence scenarios and inherent equipment risks, effectively reducing missed reports, helping to centrally allocate operation and maintenance resources to high-risk areas and high-defect equipment, and improving the economy of the early warning system and the efficiency of preventive maintenance.

[0097] The training methods for the water accumulation analysis model include:

[0098] Q groups of water accumulation training data are collected in advance, and the water accumulation training data includes measurement data, comprehensive features and water accumulation feature patterns.

[0099] The measured data and comprehensive features are used as the input of the waterlogging analysis model, and the waterlogging characteristic pattern is used as the output of the waterlogging analysis model. The goal is to minimize the error between the output waterlogging characteristic pattern and the actual waterlogging characteristic pattern. The network parameters of the waterlogging analysis model are optimized through a nature-inspired optimization algorithm to obtain the network parameters that minimize the error between the waterlogging characteristic pattern output by the waterlogging analysis model and the actual waterlogging characteristic pattern. The waterlogging analysis model constructed with the corresponding network parameters is used as the trained waterlogging analysis model.

[0100] Combined with the derived features of acoustic wave positioning, the coordinates of the leak point are determined through a fusion positioning algorithm;

[0101] Methods for obtaining leak point coordinates include:

[0102] Obtain the coordinates of the water meter well and establish a hyperbolic positioning model based on the time difference of the acoustic wave positioning derivative features;

[0103] Obtain the signal strength received by the water meter well, establish an empirical attenuation model of signal strength with respect to distance, and calculate the distance from the leak point to the water meter well using the empirical attenuation model. Establish three circle equations using the coordinates of the three water meter wells and the corresponding distances.

[0104] The hyperbolic positioning model and the three circle equations are integrated through the least squares method, such as minimizing the time difference error and the distance error from the leak point to the water meter well;

[0105] The maximum likelihood estimation or Kalman filter algorithm is used to iteratively optimize the coordinates of the leakage point.

[0106] The hyperbola positioning model constructs geometric constraints based on the time difference between the arrival of sound waves at multiple wells, which can achieve spatial positioning of leaks with meter-level accuracy, solving the problem of a single sensor "knowing the anomaly but not the location"; the empirical decay model combines empirical formulas to infer the distance to the leak, complementing the hyperbola model, providing auxiliary constraints when the time synchronization accuracy is insufficient or the sound wave signal is weak, and improving the positioning robustness in complex environments; the fusion algorithm balances the time difference and distance error through the least squares method, combined with maximum likelihood estimation or Kalman filtering iterative optimization, eliminates multiple solutions and reduces the impact of noise, ensuring that the coordinates of the leak are consistent with the pipeline network topology; the precise coordinates of the leak can be directly visualized on the GIS map, and a maintenance work order with location information is automatically generated, shortening the manual investigation time, helping the operation and maintenance team to quickly locate the leak and start repairs, reducing the spread of water accumulation, pipeline damage and water resource waste caused by leaks, and realizing the full process closed loop of "detection-positioning-disposal", significantly improving the engineering practicality and emergency response efficiency of the early warning system.

[0107] Dynamic analysis of measurement data yields derived dynamic features. These data and derived dynamic features are then used as inputs to a dynamic feature model to obtain dynamically adjusted graded waterlogging warning thresholds. These thresholds are then adjusted based on the relationship between actual waterlogging volume and the graded waterlogging warning thresholds. Dynamically adjusting the graded waterlogging warning thresholds based on real-time data and historical patterns addresses the inability of traditional static thresholds to cope with environmental fluctuations, significantly reducing false alarm and missed alarm rates. The multi-level thresholds output by the dynamic feature model enable the system to precisely trigger differentiated responses based on the relationship between actual waterlogging volume and the thresholds. For example, a level one warning indicates "drainage anomaly, inspection recommended," a level two warning indicates "pipeline parameter anomaly, potential leak located," and a level three warning directly identifies the leak coordinates and initiates an emergency response, achieving an upgrade from "extensive warning" to "precise graded response." The dynamic feature model not only provides real-time thresholds but also predicts changes in risk levels through dynamically derived features, enabling operations and maintenance teams to pre-deploy resources, shorten response times, and minimize facility damage, traffic disruptions, and water waste caused by waterlogging. This provides the core technical support for the refined management of smart water services through the "data-driven, dynamic decision-making, and graded response" approach.

[0108] Methods for obtaining derived dynamic features include:

[0109] Obtain the water level change, temperature change, and humidity change within a preset time period, calculate the ratio of the water level change, temperature change, and humidity change to the preset time period, and obtain the water level change rate, temperature change rate, and humidity change rate;

[0110] Calculate the difference between the current pipeline pressure and the preset pipeline pressure to obtain the pressure change amplitude; calculate the ratio of the difference to the preset change time to obtain the pressure drop rate; and statistically obtain the standard deviation of pressure fluctuations;

[0111] Obtain the difference between the user-end traffic and the pipeline inlet traffic, calculate the ratio of the difference to the pipeline inlet traffic, and obtain the traffic difference ratio;

[0112] Obtain the absolute difference between the current user end traffic and the user end traffic at the previous moment, calculate the ratio of the absolute difference to the user end traffic at the previous moment, and obtain the traffic mutation coefficient;

[0113] Obtaining the amount of change in the accumulated water depth within a preset change time, calculating the ratio of the amount of change in the accumulated water depth to the preset change time, and obtaining the rate of change of the accumulated water depth;

[0114] Obtaining a high-frequency vibration signal amplitude when the vibration signal exceeds a preset high-frequency threshold;

[0115] Get whether the temperature change rate exceeds the preset temperature change rate threshold. If so, mark it as 1, otherwise it is 0, and obtain the temperature anomaly index;

[0116] Get whether the humidity change rate exceeds the preset humidity change rate threshold. If so, mark it as 1, otherwise it is 0, and obtain the humidity anomaly index;

[0117] The water level change rate, temperature change rate, humidity change rate, pressure fluctuation standard deviation, flow difference ratio, flow mutation coefficient, water accumulation depth change rate, high-frequency vibration signal amplitude, temperature anomaly index and humidity anomaly index are spliced ​​to obtain derived dynamic characteristics.

[0118] Features such as the water level change rate, temperature change rate, humidity change rate, and water depth change rate quantify the evolution rate of the environment and water accumulation in real time, helping the system to quickly identify abnormal development trends; features such as the flow difference ratio and flow mutation coefficient comprehensively judge the possibility of leakage from the pipeline operation status level to avoid misjudgment of a single indicator; features such as the pressure fluctuation standard deviation and high-frequency vibration signal amplitude identify abnormal pipeline vibration and pressure instability, and combined with temperature anomaly indicators and humidity anomaly indicators, accurately distinguish causes such as pipeline rupture and leakage, external water seepage, and drainage blockage; splicing continuous dynamic parameters with discrete anomaly indicators provides rich input dimensions for the dynamic feature model, enabling the system to dynamically adjust the classification threshold according to real-time features, achieving accurate response from anomaly detection to risk classification, such as mild warnings to indicate trend anomalies, moderate warnings to locate potential problems, and severe warnings to directly lock the leakage point and initiate emergency response, significantly improving the early warning system's real-time adaptability to complex working conditions and the efficiency of the full-process closed-loop management of water accumulation events.

[0119] The training methods for dynamic feature models include:

[0120] W groups of dynamic training data are collected in advance. The dynamic training data include measurement data, derived dynamic features and graded waterlogging warning thresholds.

[0121] The measured data and derived dynamic features are used as the input of the dynamic feature model, and the graded waterlogging warning threshold is used as the output of the dynamic feature model. The goal is to minimize the error between the output graded waterlogging warning threshold and the actual graded waterlogging depth. The network parameters of the dynamic feature model are optimized through a nature-inspired optimization algorithm to obtain the network parameters that minimize the error between the graded waterlogging warning threshold output by the dynamic feature model and the actual graded waterlogging warning threshold (the actual graded waterlogging warning threshold is obtained by reviewing historical waterlogging events, and extracting the critical waterlogging threshold under different working conditions through numerical simulation or machine learning classification. The real-time waterlogging monitoring data is also compared with the historical threshold library. The threshold is adjusted to adapt to the current working conditions (such as increasing the warning sensitivity in the rainy season and reducing the false alarm rate in the dry season) through rolling window updates or online learning). The dynamic feature model constructed with the corresponding network parameters is used as the trained dynamic feature model.

[0122] Example 2

[0123] This embodiment provides a method for automatically tracing the source of accumulated water diffusion trends applied to Example 1, including the following steps:

[0124] The pipeline is divided into N×N grids, and real-time measurement data and dynamic derived features are obtained within the grids. These measurement data and dynamic derived features are used as inputs to the waterlogging analysis model to determine the probability of waterlogging. Grids with waterlogging probabilities exceeding a preset probability threshold are used as the starting point for tracing back to the source. A graph neural network is used to analyze the hydraulic connectivity between grids to obtain a hydraulic connectivity weight matrix. Combined with the hydraulic connectivity, a bidirectional long-short-term memory network is used to predict the waterlogging diffusion trend. Based on the hydraulic connectivity weight matrix and a time decay factor, a reverse diffusion tree is constructed from the current moment to the past. The reverse diffusion tree is used to simulate diffusion data at different leak locations. Spatial correlation analysis is performed between the actual diffusion data and the predicted waterlogging diffusion trend. If the spatial correlation exceeds the correlation threshold, the corresponding leak location is determined. If multiple leaks exist, a Kalman filter is used to fuse multi-sensor data for secondary screening to determine the leak location.

[0125] Example 3

[0126] See also Figure 3 As shown, this embodiment provides a water meter well remote waterlogging early warning system based on the Internet of Things, including:

[0127] Feature analysis module: This module integrates the collected water meter well measurement data through the IoT gateway, performs comprehensive analysis on the measurement data, obtains comprehensive features, and uses the measurement data and comprehensive features as inputs to the waterlogging analysis model to obtain the waterlogging feature pattern.

[0128] Comprehensive features include sensor cross-features, time series dynamic features, spatial correlation features, and environmental features; sensor cross-features include threshold trigger features, correlation features, and physical derivative features; time series dynamic features include short-term window features, frequency domain features, and state transition features; spatial correlation features include spatial gradient features and acoustic wave positioning derivative features;

[0129] Leak point location module: Combines the derived features of acoustic wave positioning and determines the coordinates of the leak point through a fusion positioning algorithm;

[0130] Waterlogging warning module: Dynamically analyzes the measurement data to obtain derived dynamic features, uses the measurement data and derived dynamic features as inputs to the dynamic feature model to obtain dynamically adjusted graded waterlogging warning thresholds; and performs corresponding graded warnings based on the relationship between the actual waterlogging volume and the graded waterlogging warning thresholds.

[0131] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0132] Finally: The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A remote waterlogging warning method for water meter wells based on the Internet of Things, characterized in that: The steps include: The collected water meter well measurement data is integrated through the IoT gateway, and the measurement data is comprehensively analyzed to obtain comprehensive features. The measurement data and comprehensive features are used as inputs to the waterlogging analysis model to obtain the waterlogging feature pattern. Comprehensive features include sensor cross-features, time series dynamic features, spatial correlation features, and environmental features; sensor cross-features include threshold trigger features, correlation features, and physical derivative features; time series dynamic features include short-term window features, frequency domain features, and state transition features; spatial correlation features include spatial gradient features and acoustic wave positioning derivative features; Combined with the derived features of acoustic wave positioning, the coordinates of the leak point are determined through a fusion positioning algorithm; Perform dynamic analysis on the measured data to obtain derived dynamic features, use the measured data and derived dynamic features as inputs to the dynamic feature model, and obtain dynamically adjusted graded waterlogging warning thresholds; And make corresponding graded warnings based on the relationship between the actual amount of water accumulation and the graded water accumulation warning threshold; Methods for obtaining threshold trigger characteristics include: Detect whether the pipeline pressure change value exceeds the preset pressure threshold, detect whether the difference between the total flow value and the user end flow value exceeds the preset flow threshold; detect whether the water level rise rate exceeds the preset rate threshold; if all three are exceeded at the same time, mark the pipeline rupture risk as 1, otherwise it is 0; use the pipeline rupture risk as the threshold trigger feature; Methods for obtaining correlation features include: The Pearson correlation coefficient was used to calculate the correlation between different sensors and obtain the correlation characteristics; Methods for obtaining spatial association features include: Calculate the water level difference and spacing between adjacent water meter wells, calculate the ratio of the water level difference and spacing, and obtain the water level gradient; Calculate the pressure difference and distance between the upstream water meter well and the downstream water meter well, calculate the ratio of the pressure difference to the distance, and obtain the pressure decay rate; The water level gradient and pressure decay rate are spliced ​​together to obtain the spatial gradient characteristics; Detect the time difference between any two water meter wells receiving the water leakage sound wave, calculate the ratio of the difference in received signal strength and the distance between the two water meter wells, and obtain the signal attenuation rate; splice the time difference and the signal attenuation rate to obtain the derived characteristics of the acoustic wave positioning; The spatial gradient features and sonar localization derived features are spliced ​​together to obtain spatial correlation features.

2. The remote waterlogging early warning method for water meter wells based on the Internet of Things according to claim 1 is characterized in that: Methods for obtaining physically derived features include: Obtain the total flow at the user end and the flow at the pipeline inlet, and calculate the leakage rate; count the pressure change and flow change, and calculate the joint feature; concatenate the leakage rate and the joint feature to obtain the physical derivative feature.

3. The remote waterlogging early warning method for water meter wells based on the Internet of Things according to claim 1 is characterized in that: Methods for obtaining state transition characteristics include: A threshold is preset to classify the sensor states into normal, warning, and fault, and the data is encoded in sequence. The sensor's time series data is converted into a state sequence. The number of transitions from state a to state b is counted, the total number of stays in state a is calculated, and the ratio of the number of transitions to the total number of stays is calculated to obtain a transition probability matrix. The continuous stay duration of each state is counted, and the transition probability matrix and the continuous stay duration are concatenated to obtain statistical features. Count the average time interval from state a to state b, count the number of occurrences of the preset transition pattern, and concatenate the average time interval and the number of occurrences to obtain the temporal dependency feature; Obtain the sensor's change rate and change amplitude during state transition, and whether the states of other sensors change when the sensor's change amplitude exceeds a preset change threshold; combine the sensor's change rate and change amplitude, and whether the states of other sensors change when the sensor's change amplitude exceeds a preset change threshold, to obtain dynamic trend characteristics; Statistical features, time series dependency features, and dynamic trend features are combined to obtain state transition features.

4. The remote waterlogging early warning method for water meter wells based on the Internet of Things according to claim 1 is characterized in that: Methods for obtaining temporal dynamic characteristics include: The mean, standard deviation, and rate of change of the water level are calculated. The minimum value, fluctuation range, and number of times the pressure change per unit time in the pipeline pressure is lower than the pressure threshold are calculated. The mean, standard deviation, and rate of change of the water level, as well as the minimum value, fluctuation range, and number of times the pressure change per unit time in the pipeline pressure is lower than the pressure threshold are concatenated to obtain short-term window characteristics. Statistically obtain the deviation between the actual temperature value and the historical average temperature for the same period, the deviation between the actual humidity value and the historical average humidity for the same period, and the proportion of low-frequency vibration energy below the preset low-frequency threshold. The deviation between the actual temperature value and the historical average temperature for the same period, the deviation between the actual humidity value and the historical average humidity for the same period, and the proportion of low-frequency vibration energy below the preset low-frequency threshold are spliced ​​as long-term window features; The main frequency energy of the vibration signal and the high-frequency noise power of the pipeline pressure signal are extracted through Fourier transform, and the main frequency energy and high-frequency noise power are spliced ​​as frequency domain features; The short-term window features, long-term window features, frequency domain features and state transition features are spliced ​​together to obtain the time series dynamic features.

5. The remote waterlogging early warning method for water meter wells based on the Internet of Things according to claim 1 is characterized in that: Methods for obtaining environmental characteristics include: Count the number of historical waterlogging events at the same location, and concatenate the coordinates corresponding to the location with the historical waterlogging events to serve as historical waterlogging features. Count the number of failures of similar equipment, and concatenate the equipment model and failure times as equipment failure features. Concatenate historical waterlogging features and equipment failure features to obtain environmental features.

6. The remote waterlogging early warning method for water meter wells based on the Internet of Things according to claim 1 is characterized in that: Methods for obtaining leak point coordinates include: Obtain the coordinates of the water meter well and establish a hyperbolic positioning model based on the time difference of the acoustic wave positioning derivative features; Obtain the signal strength received by the water meter well, establish an empirical attenuation model of signal strength with respect to distance, and calculate the distance from the leak point to the water meter well using the empirical attenuation model. Establish three circle equations using the coordinates of the three water meter wells and the corresponding distances. The hyperbolic positioning model and the three circle equations are fused by the least squares method; The maximum likelihood estimation or Kalman filter algorithm is used to iteratively optimize the coordinates of the leakage point.

7. The remote waterlogging early warning method for water meter wells based on the Internet of Things according to claim 1 is characterized in that: Methods for obtaining derived dynamic features include: Obtain the water level change, temperature change, and humidity change within a preset time period, calculate the ratio of the water level change, temperature change, and humidity change to the preset time period, and obtain the water level change rate, temperature change rate, and humidity change rate; Calculate the difference between the current pipeline pressure and the preset pipeline pressure to obtain the pressure change amplitude; calculate the ratio of the difference to the preset change time to obtain the pressure drop rate; and statistically obtain the standard deviation of pressure fluctuations; Obtain the difference between the user-end traffic and the pipeline inlet traffic, calculate the ratio of the difference to the pipeline inlet traffic, and obtain the traffic difference ratio; Obtain the absolute difference between the current user end traffic and the user end traffic at the previous moment, calculate the ratio of the absolute difference to the user end traffic at the previous moment, and obtain the traffic mutation coefficient; Obtaining the amount of change in the accumulated water depth within a preset change time, calculating the ratio of the amount of change in the accumulated water depth to the preset change time, and obtaining the rate of change of the accumulated water depth; Obtaining a high-frequency vibration signal amplitude when the vibration signal exceeds a preset high-frequency threshold; Get whether the temperature change rate exceeds the preset temperature change rate threshold. If so, mark it as 1, otherwise it is 0, and obtain the temperature anomaly index; Get whether the humidity change rate exceeds the preset humidity change rate threshold. If so, mark it as 1, otherwise it is 0, and obtain the humidity anomaly index; The water level change rate, temperature change rate, humidity change rate, pressure fluctuation standard deviation, flow difference ratio, flow mutation coefficient, water accumulation depth change rate, high-frequency vibration signal amplitude, temperature anomaly index and humidity anomaly index are spliced ​​to obtain derived dynamic characteristics.

8. A water meter well remote waterlogging early warning system based on the Internet of Things, implementing the water meter well remote waterlogging early warning method based on the Internet of Things according to any one of claims 1 to 7, characterized in that: include: Feature analysis module: This module integrates the collected water meter well measurement data through the IoT gateway, performs comprehensive analysis on the measurement data, obtains comprehensive features, and uses the measurement data and comprehensive features as inputs to the waterlogging analysis model to obtain the waterlogging feature pattern. Comprehensive features include sensor cross-features, time series dynamic features, spatial correlation features, and environmental features; sensor cross-features include threshold trigger features, correlation features, and physical derivative features; time series dynamic features include short-term window features, frequency domain features, and state transition features; spatial correlation features include spatial gradient features and acoustic wave positioning derivative features; Leak point location module: Combines the derived features of acoustic wave positioning and determines the coordinates of the leak point through a fusion positioning algorithm; Waterlogging warning module: Dynamically analyzes measurement data to obtain derived dynamic features. The measurement data and derived dynamic features are used as inputs to the dynamic feature model to obtain dynamically adjusted graded waterlogging warning thresholds. And according to the relationship between the actual amount of water accumulation and the graded water accumulation warning threshold, corresponding graded warnings will be issued.

Images