CN-121977167-A - Water supply network leakage monitoring and positioning system and method based on big data analysis

CN121977167ACN 121977167 ACN121977167 ACN 121977167ACN-121977167-A

Abstract

The invention relates to the technical field of water supply network monitoring and discloses a water supply network leakage monitoring and positioning system and method based on big data analysis, wherein the system comprises the steps of acquiring continuous audio stream data and network topology diagram data acquired by a distributed sensor, and calculating background baseline parameters; the method comprises the steps of applying multi-scale wavelet transformation decomposition to a current audio frame, calculating energy entropy characteristics and generating single-frame anomaly scores, applying a cumulative sum control graph algorithm to output cumulative anomaly indexes, generating enhanced anomaly scores through a graph attention network, and judging and positioning leakage based on the multi-sensor anomaly scores. The invention solves the technical problem of low sensitivity of weak leakage signal detection.

Inventors

Yao qingda
CAO YUYING
HUANG JINGLAN
SHI JIAXU

Assignees

吉林省讯达科技发展有限公司

Dates

Publication Date: 20260505
Application Date: 20260127

Claims (10)

1. The water supply network leakage monitoring and positioning method based on big data analysis is characterized by comprising the following steps of: Acquiring continuous audio stream data and pipe network topological graph data acquired by distributed sensors, and calculating background baseline parameters of each sensor node based on a historical energy entropy sequence in a sliding time window; Applying multi-scale wavelet transformation decomposition to a current audio frame, calculating energy entropy characteristics based on energy distribution of each scale detail coefficient, and comparing the energy entropy characteristics with background baseline parameters to generate single-frame anomaly scores; Applying a cumulative sum control graph algorithm to the single frame anomaly scores of the continuous multiframes to perform cumulative sum, and outputting cumulative anomaly indexes of each sensor node; the accumulated abnormal indexes of the sensor nodes are used as initial node characteristics to be input into a graph attention network, and enhanced abnormal scores fusing neighborhood information are generated through neighbor node characteristic aggregation; comparing the enhanced anomaly score with a detection threshold value to perform leakage judgment, determining the position coordinates of the leakage points under the topological constraint of the pipe network based on the spatial distribution of the anomaly scores of the multiple sensors, and outputting leakage monitoring results and the positioning coordinates of the leakage points; The attention network calculates attention coefficients between each node and the neighbor nodes of the node, weighting and aggregating the neighbor node characteristics based on the attention coefficients, and obtaining the attention coefficients through linear transformation and splicing of the node characteristic pairs, activating functions and normalizing calculation.
2. The method of claim 1, wherein calculating an energy entropy feature based on the energy distribution of scale detail coefficients comprises: Calculating the energy value of each scale detail coefficient; calculating normalized energy distribution probability based on each scale energy value, wherein the normalized energy distribution probability is the ratio between each scale energy value and the sum of all scale energy values; And calculating a shannon entropy value as an energy entropy characteristic based on the normalized energy distribution probability.
3. The method of claim 1, wherein said comparing the energy entropy signature to the background baseline parameter, generating a single frame anomaly score comprises: Calculating a difference value between the current energy entropy characteristic and a background baseline mean value; dividing the difference value by the standard deviation of the background baseline to obtain a normalized deviation degree as a single frame anomaly score.
4. The method of claim 1, wherein said applying a cumulative sum control graph algorithm to single frame anomaly scores for consecutive multiple frames comprises: Respectively calculating a positive accumulation sum and a negative accumulation sum for a single frame anomaly score sequence of continuous frames, wherein the positive accumulation sum is a result of taking a non-negative value after subtracting a relaxation parameter from a positive accumulation sum of a previous frame and an anomaly score of a current frame, and the negative accumulation sum is a result of taking a non-negative value after subtracting a relaxation parameter from a negative value of the anomaly score of the previous frame and the anomaly score of the current frame; and taking the maximum value between the positive accumulation and the negative accumulation as an accumulation abnormality index.
5. The method according to claim 1, wherein the method further comprises: obtaining ageing-related attribute data of each pipe section from a pipe network asset database, wherein the ageing-related attribute data comprise pipe types, service lives of pipes, historical failure times and soil corrosiveness indexes; And normalizing the aging-related attribute data to form an aging factor feature vector.
6. The method of claim 5, wherein the method further comprises: Calculating the ageing risk weight of each pipe section based on the correlation analysis of the historical leakage record and the ageing factor; calculating the ageing risk score of each pipe section according to the ageing risk weight and the ageing factor feature vector; calculating the average value of the ageing risk scores of the connecting pipe sections of the sensor nodes as a node ageing risk coefficient; And multiplying the enhanced anomaly score by taking the node aging risk coefficient as a modulation factor to generate a modulation anomaly score.
7. The method of claim 6, wherein the method further comprises: Performing differential adjustment on the detection threshold value based on the ageing risk coefficient of each node, wherein the differential adjustment is to multiply the reference detection threshold value by an adjustment coefficient which is inversely related to the ageing risk coefficient; and carrying out differential threshold judgment on the modulation anomaly score.
8. The method according to claim 1, wherein the method further comprises: when the node anomaly score exceeds a corresponding threshold value, acquiring the anomaly score of the node and the first-order neighbor node thereof; carrying out weighted summation on the basis of the abnormal scores of the node and the neighbor nodes, and calculating cooperative confidence, wherein the weight of the neighbor nodes is inversely related to the pipe segment distance between the nodes; and determining an early warning level based on the comparison of the cooperative confidence coefficient and a preset level threshold value, and outputting a cooperative confidence coefficient score and the early warning level.
9. The method of claim 1, wherein determining the leak location coordinates under the network topology constraints based on the spatial distribution of the multisensor anomaly scores comprises: Extracting the enhanced anomaly score and the geographic coordinates of the node triggering the alarm and the first-order neighbor node; And taking the reinforced abnormal score of each node as weight, and carrying out weighted average calculation on the geographic coordinates of each node to obtain the position coordinates of the leakage point.
10. A water supply network leakage monitoring and positioning system based on big data analysis, for executing the water supply network leakage monitoring and positioning method based on big data analysis according to any one of claims 1 to 9, characterized by comprising: The data acquisition and baseline calculation module is used for acquiring continuous audio stream data and pipe network topological graph data acquired by the distributed sensors and calculating background baseline parameters of each sensor node based on a historical energy entropy sequence in a sliding time window; The feature extraction module is used for applying multi-scale wavelet transformation decomposition to the current audio frame, calculating energy entropy features based on energy distribution of each scale detail coefficient, and comparing the energy entropy features with the background baseline parameters to generate single-frame anomaly scores; The time sequence accumulation module is used for carrying out accumulation summation on single frame anomaly scores of continuous multiframes by applying an accumulation and control graph algorithm and outputting accumulation anomaly indexes of each sensor node; The space aggregation module is used for inputting the accumulated abnormal indexes of each sensor node as initial node characteristics into the graph attention network, and generating enhanced abnormal scores fusing neighborhood information through neighbor node characteristic aggregation; and the leakage judging and positioning module is used for comparing the enhanced anomaly score with a detection threshold value to judge leakage, determining the position coordinates of the leakage points under the topological constraint of the pipe network based on the spatial distribution of the anomaly scores of the multiple sensors, and outputting a leakage monitoring result and the positioning coordinates of the leakage points.

Description

Water supply network leakage monitoring and positioning system and method based on big data analysis Technical Field The invention relates to the technical field of water supply network monitoring, in particular to a water supply network leakage monitoring and positioning system and method based on big data analysis. Background In the operation and maintenance process of the large-scale urban water supply network, the distributed sensor continuously collects mass pipeline audio data for leakage detection. The audio signal strength generated at the early stage of leakage is extremely weak, and the signal amplitude is close to or even lower than the noise floor level of the sensor. In the prior art, the traditional single-frame detection method based on the fixed threshold value is widely applied to pipe network leakage monitoring. According to the method, the audio signals of all the sensor nodes are independently analyzed and judged by setting a fixed detection threshold value. However, the conventional method has the defects that although the response capability to weak leakage signals can be improved by reducing the threshold value, background noise fluctuation can be misjudged as a leakage event, and the weak signals in the early stage of leakage can be missed by reducing misinformation by increasing the threshold value. In addition, in the existing method, each sensor node is used as an independent unit to perform detection analysis, the topology structure information of the pipe network is not utilized, the related anomalies of adjacent nodes cannot be mutually verified in a single-point independent detection mode, and the comprehensive judging capability of the detection system on weak leakage signals is weakened. Meanwhile, the traditional detection method adopts the same detection threshold value for all pipe sections, and does not conduct differentiation treatment according to the ageing risk of the pipeline, so that the sensitivity of detecting tiny leakage of the high-risk old pipe section is insufficient. Disclosure of Invention The invention provides a water supply network leakage monitoring and positioning system and method based on big data analysis, which solve the technical problems that weak leakage signal detection sensitivity and false alarm rate are difficult to balance, single-point detection cannot utilize spatial correlation of a pipeline network, and unified threshold value cannot distinguish ageing risk differences of the pipeline in the related technology. The invention provides a water supply network leakage monitoring and positioning method based on big data analysis, which comprises the following steps: 1. The water supply network leakage monitoring and positioning method based on big data analysis is characterized by comprising the following steps of: Acquiring continuous audio stream data and pipe network topological graph data acquired by distributed sensors, and calculating background baseline parameters of each sensor node based on a historical energy entropy sequence in a sliding time window; Applying multi-scale wavelet transformation decomposition to a current audio frame, calculating energy entropy characteristics based on energy distribution of each scale detail coefficient, and comparing the energy entropy characteristics with background baseline parameters to generate single-frame anomaly scores; Applying a cumulative sum control graph algorithm to the single frame anomaly scores of the continuous multiframes to perform cumulative sum, and outputting cumulative anomaly indexes of each sensor node; the accumulated abnormal indexes of the sensor nodes are used as initial node characteristics to be input into a graph attention network, and enhanced abnormal scores fusing neighborhood information are generated through neighbor node characteristic aggregation; comparing the enhanced anomaly score with a detection threshold value to perform leakage judgment, determining the position coordinates of the leakage points under the topological constraint of the pipe network based on the spatial distribution of the anomaly scores of the multiple sensors, and outputting leakage monitoring results and the positioning coordinates of the leakage points; The attention network calculates attention coefficients between each node and the neighbor nodes of the node, weighting and aggregating the neighbor node characteristics based on the attention coefficients, and obtaining the attention coefficients through linear transformation and splicing of the node characteristic pairs, activating functions and normalizing calculation. 2. The method of claim 1, wherein calculating an energy entropy feature based on the energy distribution of scale detail coefficients comprises: Calculating the energy value of each scale detail coefficient; calculating normalized energy distribution probability based on each scale energy value, wherein the normalized energy distribution probability is the ratio bet