CN-121980409-A - Detection and positioning method for coping with multi-pipeline leakage of water supply network

CN121980409ACN 121980409 ACN121980409 ACN 121980409ACN-121980409-A

Abstract

The invention relates to the technical field of operation and maintenance detection of a water supply network, in particular to a detection and positioning method for coping with multi-pipeline leakage of the water supply network, which comprises the following steps: and collecting pressure time sequence data of monitoring points of the water supply network, processing the pressure time sequence data to generate a standardized pressure fluctuation sequence, analyzing the data by utilizing a pre-training leakage reasoning calculation model to obtain a leakage suspected degree score and leakage type classification, screening high-suspected degree leakage events, and obtaining corresponding positions and time stamps. And constructing a transient flow state inversion equation set by combining a pipe network topological structure and a hydraulic model, reversely deducing the hydraulic state change of the pipe section by using a standardized pressure sequence as a boundary condition through a state inversion algorithm, identifying a pressure gradient mutation point, positioning a corresponding pipe section, and fusing leakage type and positioning information to generate a complete detection positioning report. The method can accurately distinguish leakage types, accurately position the concurrent leakage pipe sections of multiple pipelines, reduce non-leakage signal interference and adapt to the leakage detection scene of the complex pipe network.

Inventors

CHEN BENYAO
TANG QINGCHAO
SHEN QIAOYUN
SUN CHANGJUN
LI XIUTING

Assignees

山东绿之源节水科技研究院有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. The utility model provides a detection and location method of reply water supply network multi-pipe leakage, which is characterized in that includes: Collecting and processing pressure time sequence data of selected monitoring points in a water supply network to generate a standardized pressure fluctuation sequence; analyzing pressure time sequence data of the water supply network based on a pre-trained leakage reasoning calculation model, and obtaining leakage suspected degree scores and leakage type classifications; Screening out suspected leakage events higher than a set threshold according to the leakage suspected degree score, and acquiring the positions and time stamps of monitoring points where the suspected leakage events occur; constructing a pipe network transient flow state inversion equation set by utilizing a topological structure model of a water supply pipe network, a pipeline hydraulic model and the occurrence time of the suspected leakage event; Solving the pipe network transient flow state inversion equation set by adopting a state inversion algorithm, wherein the state inversion algorithm takes the standardized pressure fluctuation sequence actually measured by the pressure sensor as a boundary condition, and reversely deduces the hydraulic state change of each pipe section in the pipe network in the suspected leakage event time window; According to the hydraulic state change deduced by the state inversion algorithm, identifying pipe network connection points with abrupt change of the pressure gradient, and positioning a specific pipe section associated with the abrupt change of the pressure gradient; And combining the leakage type classification output by the leakage reasoning calculation model and the specific pipe section positioned by the state inversion algorithm to generate a multi-pipe leakage detection and positioning report containing the leakage position, the leakage type and the occurrence time.
2. The method for detecting and locating multi-pipe leaks in a water supply network according to claim 1, wherein the step of collecting and processing pressure time series data of selected monitoring points in the water supply network to generate a standardized pressure fluctuation sequence includes: collecting pressure time sequence data of selected monitoring points in a water supply network, wherein the pressure time sequence data are continuously obtained through pressure sensors deployed at key nodes of the network; Performing signal preprocessing on the acquired pressure time sequence data to generate a standardized pressure fluctuation sequence, wherein the method specifically comprises the following steps of: Applying a wavelet threshold to the pressure time sequence data to remove noise, separating out high-frequency noise components and filtering the high-frequency noise components; A polynomial fitting method is adopted on the denoised pressure data, long-term trend items caused by the change of the water consumption mode are stripped, and a pressure residual error sequence after the trend items are removed is obtained; resampling the pressure residual sequence at equal intervals at fixed time intervals to form the normalized pressure fluctuation sequence; and carrying out normalization processing on the normalized pressure fluctuation sequence, and eliminating the magnitude difference caused by different reference pressures among different monitoring points.
3. The method for detecting and locating multi-pipe leaks in a water supply network according to claim 2, wherein analyzing pressure time series data of the water supply network based on a pre-trained leak inference calculation model, obtaining a leak suspected score and a leak type classification, comprises: identifying an abnormal pressure event contained therein by a pattern recognition algorithm based on the normalized pressure fluctuation sequence, the abnormal pressure event exhibiting an abnormal dip or continuous drop segment of a pressure value; Extracting waveform characteristics of each abnormal pressure event; Inputting the extracted waveform characteristics into a pre-trained leakage inference calculation model, wherein the leakage inference calculation model outputs leakage suspected degree scores and classifications of leakage types, and the method specifically comprises the following steps of: the leakage reasoning calculation model is based on a gradient lifting decision tree architecture, and an input layer of the leakage reasoning calculation model receives a feature vector formed by the waveform features; splitting and judging the input feature vector by an implicit layer of the leakage reasoning calculation model through a multi-level decision tree, and extracting nonlinear feature combinations related to leakage layer by layer; Outputting the leakage suspected degree score by one branch at an output layer of the leakage reasoning calculation model, wherein the leakage suspected degree score is a continuous numerical value between zero and one; and outputting the classification probability of the leakage type by the other branch at the output layer of the leakage reasoning calculation model.
4. A method for detecting and locating a multi-pipe leak in a water supply network as defined in claim 3, wherein said constructing a system of pipe network transient flow inversion equations using a topology model of the water supply network, a hydraulic model of the pipe, and the time of occurrence of said suspected leak event comprises: The topological structure model defines the connection relation among all pipelines, nodes, water sources and monitoring points in the pipe network; the pipeline hydraulic model comprises the length, diameter, material, friction coefficient and valve state hydraulic attribute of the pipeline; Taking the occurrence time of the suspected leakage event as the starting moment of state inversion, and taking the data of the standardized pressure fluctuation sequence before and after the moment as the known observation value; Based on mass conservation and momentum conservation equations, establishing an equation set taking flow of all unknown nodes in a pipe network and pressure drop of the pipe section as state variables, wherein a coefficient matrix of the equation set is determined by the topological structure model and the pipeline hydraulic model; substituting the known observed value as a boundary condition into the equation set to form a closed pipe network transient flow state inversion equation set.
5. The method for detecting and locating multi-pipe leakage of water supply network according to claim 4, wherein said adopting a state inversion algorithm to solve said system of pipe network transient state inversion equations, said state inversion algorithm reversely deducing hydraulic state changes of pipe sections inside the pipe network within said suspected leakage event time window using said standardized pressure fluctuation sequence actually measured by the pressure sensor as boundary conditions, comprises: the state inversion algorithm adopts an accompanying method, and an objective function is constructed by minimizing the difference between the pipe network node pressure calculated by a model and the observed pressure in the standardized pressure fluctuation sequence; iteratively adjusting an unknown state variable in the pipe network transient state inversion equation set by solving a gradient of the objective function relative to the state variable; In each iteration, recalculating the pressure distribution of the whole pipe network by using the adjusted state variables, comparing the pressure distribution with an observed value, and updating the gradient direction; And stopping iteration when the value of the objective function is converged to be within a preset tolerance or the maximum iteration number is reached, wherein the state variable solution at the moment is the hydraulic state change of each pipe section in the pipe network obtained by inversion.
6. The method for detecting and locating multi-pipe leakage of a water supply network according to claim 5, wherein identifying pipe network connection points where abrupt changes in pressure gradient occur and locating a specific pipe section associated with the abrupt changes in pressure gradient according to the hydraulic state change deduced by the state inversion algorithm comprises: calculating the pressure gradient distribution of each pipeline along the path in the pipe network based on the hydraulic state solution finally obtained by the state inversion algorithm; Searching for spatial position points of which the pressure gradient value exceeds a normal fluctuation range threshold value in the pressure gradient distribution; mapping the searched space position points back to a topological structure model of the water supply network, and determining network connection points where the pressure gradient is suddenly changed; And in the topological structure model, all pipe sections directly connected with the pipe network connection point are searched, and the specific pipe section with leakage is determined by combining the direction of the pressure gradient.
7. The method for detecting and locating multi-pipeline leakage of a water supply network according to claim 6, wherein the step of combining the leakage type classification output by the leakage inference calculation model with the specific pipeline section located by the state inversion algorithm to generate a multi-pipeline leakage detection and locating report including the leakage position, the leakage type and the occurrence time comprises the steps of: Taking the serial numbers of the specific pipe sections, the initial node information and the termination node information positioned by the state inversion algorithm as leakage position information; Classifying the leakage type of the suspected leakage event which is output by the leakage reasoning calculation model and is related to the specific pipe section, and taking the leakage type as leakage type information; taking the occurrence time of the suspected leakage event as leakage occurrence time information; correlating and integrating the leakage position information, the leakage type information and the leakage occurrence time information; And filling the integrated information according to a preset report template, and attaching the pressure gradient distribution diagram for positioning analysis to form a final multi-pipeline leakage detection and positioning report.
8. The method for detecting and locating leakage of multiple pipelines in a water supply network according to claim 7, wherein the step of constructing a leakage inference calculation model comprises: Acquiring pressure time sequence data of a historical water supply network in a normal running state and various known leakage event occurrence states to form an initial training data set; executing the signal preprocessing step on pressure time sequence data corresponding to each known leakage event in the initial training data set to generate a corresponding standardized pressure fluctuation sequence; Extracting the abnormal pressure event and the waveform characteristics corresponding to the abnormal pressure event from the standardized pressure fluctuation sequence according to a marked time window of a known leakage event; correlating the extracted waveform characteristics with real labels corresponding to leakage events, wherein the real labels comprise leakage existence labels and leakage type labels, and form characteristic-label sample pairs for model training; Adopting a gradient lifting decision tree algorithm, taking waveform characteristics in the characteristic-label sample pair as input, and taking corresponding leakage existence labels and leakage type labels as supervision targets, and carrying out iterative optimization training on model parameters; In the model training process, dividing the feature-label sample pair into a training subset and a verification subset, updating model parameters by using the training subset, and evaluating the prediction performance of the model on leakage suspected degree scores and leakage type classifications by using the verification subset so as to prevent the model from being over-fitted; and stopping training when the prediction performance of the model on the verification subset meets the preset accuracy and recall index, and obtaining the leakage reasoning calculation model after training is completed.
9. The method for detecting and locating multi-pipe leakage of water supply network according to claim 8, wherein the establishing an equation set using all unknown node flow and pipe section pressure drop in the network as state variables based on mass conservation and momentum conservation equations comprises: Aiming at each node in the pipe network topological structure model, establishing a node flow balance equation according to a mass conservation equation; establishing a pipeline pressure drop equation according to a momentum conservation equation aiming at each section of pipeline in the pipe network hydraulic model; Combining the node flow balance equation established for all nodes and the pipeline pressure drop equation established for all pipe sections to form a nonlinear equation set taking all unknown node flows and all unknown pipe section pressure drops in the pipe network as state variables; The coefficient matrix of the nonlinear equation set is determined by the node-pipeline connection relation defined by the topological structure model and the pipeline attribute parameter defined by the pipeline hydraulic model.
10. The method for detecting and locating multi-pipe leaks in a water supply network of claim 9, wherein iteratively adjusting unknown state variables in the system of pipe network transient state inversion equations by solving for gradients of the objective function relative to the state variables includes: Defining the objective function as a two-norm square of the difference between the pipe network node pressure vector calculated by the model and the observed pressure vector in the measured standardized pressure fluctuation sequence; Calculating gradient vectors of the objective function on all unknown state variables in the pipe network transient state inversion equation set; determining a search direction for descending the objective function according to the gradient vector obtained by calculation; Updating the unknown state variable with a preset step length along the searching direction to obtain a group of new state variable estimated values; Substituting the updated state variable estimated value into the pipe network transient flow state inversion equation set, and recalculating the pressure distribution of the whole pipe network, so as to calculate a new model calculation pressure vector; Calculating a pressure vector and the observed pressure vector based on the new model, and recalculating the value of the objective function; judging whether the recalculated value of the objective function is smaller than a preset convergence tolerance or whether the iteration number reaches a preset maximum iteration number; If the stopping condition is not met, calculating the difference between the pressure vector and the observed pressure vector based on the new model, recalculating the gradient vector of the objective function relative to the updated state variable, and repeatedly executing the steps of determining the searching direction, updating the state variable, recalculating the pressure distribution and the objective function and judging whether the stopping condition is met; If the stopping condition is met, terminating the iterative process, and taking the state variable estimated value obtained by the last update as a final solution of the state inversion algorithm, namely inverting the hydraulic state change of each pipe section in the pipe network in the suspected leakage event time window.

Description

Detection and positioning method for coping with multi-pipeline leakage of water supply network Technical Field The invention relates to the technical field of operation and maintenance detection of a water supply network, in particular to a detection and positioning method for coping with multi-pipeline leakage of the water supply network. Background The existing water supply network leakage detection and positioning mostly adopt a mode of combining single-point pressure monitoring with simple threshold judgment, part of techniques rely on a conventional hydraulic model to carry out leakage analysis, and the leakage analysis can only carry out primary identification on single pipeline leakage, can not quantitatively score the possibility of leakage occurrence, and can not carry out type division on leakage with different causes and different forms. When the pressure of the pipe network abnormally fluctuates, the technology is easy to judge the non-leakage interference signal as a leakage event, the space-time information corresponding to the effective leakage event cannot be accurately extracted, and under the scene that the pressure abnormality occurs in multiple pipelines at the same time, the situation of leakage judgment and misjudgment is easy to occur. The existing pipe network leakage positioning is carried out by adopting a forward hydraulic simulation mode to analyze, a general leakage area can be defined only according to pressure change of monitoring points, reverse deduction calculation can not be carried out by combining transient flow states, and abrupt change characteristics of pressure gradient of pipe sections in the pipe network are difficult to capture. The prior art cannot rely on a standardized pressure fluctuation sequence as a boundary condition to construct an inversion equation set to deduce the hydraulic state change of a pipe network, cannot combine leak classification information inferred by a model with pipe section positioning results, cannot form complete multi-pipe leak detection positioning information, and is difficult to adapt to complex pipe network operation and maintenance scenes with multi-pipe concurrent leaks. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a detection and positioning method for coping with multi-pipeline leakage of a water supply network. In order to achieve the purpose, the invention adopts the following technical scheme that the method for detecting and positioning the multi-pipeline leakage of the water supply network comprises the following steps: Collecting and processing pressure time sequence data of selected monitoring points in a water supply network to generate a standardized pressure fluctuation sequence; analyzing pressure time sequence data of the water supply network based on a pre-trained leakage reasoning calculation model, and obtaining leakage suspected degree scores and leakage type classifications; Screening out suspected leakage events higher than a set threshold according to the leakage suspected degree score, and acquiring the positions and time stamps of monitoring points where the suspected leakage events occur; constructing a pipe network transient flow state inversion equation set by utilizing a topological structure model of a water supply pipe network, a pipeline hydraulic model and the occurrence time of the suspected leakage event; Solving the pipe network transient flow state inversion equation set by adopting a state inversion algorithm, wherein the state inversion algorithm takes the standardized pressure fluctuation sequence actually measured by the pressure sensor as a boundary condition, and reversely deduces the hydraulic state change of each pipe section in the pipe network in the suspected leakage event time window; According to the hydraulic state change deduced by the state inversion algorithm, identifying pipe network connection points with abrupt change of the pressure gradient, and positioning a specific pipe section associated with the abrupt change of the pressure gradient; And combining the leakage type classification output by the leakage reasoning calculation model and the specific pipe section positioned by the state inversion algorithm to generate a multi-pipe leakage detection and positioning report containing the leakage position, the leakage type and the occurrence time. As a further aspect of the present invention, the collecting and processing pressure time series data of a selected monitoring point in a water supply network to generate a standardized pressure fluctuation sequence includes: collecting pressure time sequence data of selected monitoring points in a water supply network, wherein the pressure time sequence data are continuously obtained through pressure sensors deployed at key nodes of the network; Performing signal preprocessing on the acquired pressure time sequence data to generate a standardized pressure fluctuation sequence, whe