Sound pressure flow multi-mode sensing water supply pipe network leakage positioning and early warning system

Abstract

The invention relates to the technical field of water supply pipe network fault monitoring, in particular to a sound pressure flow multi-modal sensing water supply pipe network leakage positioning and early warning system, which comprises a multi-modal real-time acquisition module for acquiring multi-modal data of a water supply pipe network in real time; the preprocessing module is used for preprocessing the multi-modal data of the water supply network and carrying out feature fusion; the leakage detecting and positioning module is used for detecting the occurrence condition of a leakage event in real time and identifying and positioning the leakage event; the health early warning management module is used for constructing a health early warning model, performing health early warning and optimizing the leakage detection positioning module; and the intelligent control module is used for generating an optimal control strategy of the water supply pipe network and intelligently controlling the water supply pipe network according to the optimal control strategy of the water supply pipe network. The leakage rate of the pipe network is reduced, and the utilization efficiency of water resources and the operation stability of the pipe network are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water supply network fault monitoring, in particular to a sound pressure flow multi-modal sensing water supply network leakage positioning and early warning system. BACKGROUND

[0002] The water supply network is the "lifeline" of large industrial facilities such as coal-fired power plants and petrochemical parks, and its safe, stable and efficient operation is directly related to the continuity and economy of the entire production system. However, especially for underground pipe network systems that have been in service for more than 15 years, the leakage risk is increasingly prominent due to long-term corrosion, vibration and stress. Traditional manual inspection and single parameter monitoring mode has been difficult to meet the fine requirements of modern industry for pipe network safety and energy efficiency management.

[0003] Chinese Patent Publication No. CN110332467A discloses a water supply network leakage monitoring and early warning system, which includes a leakage hazard early warning module, a leakage alarm module, and an alarm and early warning comprehensive management platform. The leakage hazard early warning module and the leakage alarm module are respectively communicatively connected to the alarm and early warning comprehensive management platform. However, the information of the present application is summarized to the platform "for reference and viewing by operating personnel", and the decision and control still highly depend on manual experience, which is difficult to overcome the problems of high pipe network leakage rate, low water resource utilization efficiency and low pipe network operation stability in the prior art. SUMMARY

[0004] To this end, the present application provides a sound pressure flow multi-modal sensing water supply network leakage positioning and early warning system to overcome the problems of high pipe network leakage rate, low water resource utilization efficiency and low pipe network operation stability in the prior art.

[0005] To achieve the above-mentioned purpose, the present application provides a sound pressure flow multi-modal sensing water supply network leakage positioning and early warning system, which comprises: A multi-modal real-time acquisition module is used to acquire multi-modal data of the water supply network in real time to obtain multi-modal data of the water supply network. A preprocessing module is used to preprocess the multi-modal data of the water supply network to obtain preprocessed multi-modal data, and is also used to perform feature fusion on the preprocessed multi-modal data to obtain feature fusion data. A leakage detection and positioning module is used to detect the occurrence of a leakage event in real time according to the feature fusion data, and to identify and locate the leakage event to obtain a leakage detection result and leakage positioning data. A health early warning management module is used to construct a health early warning model, and is also used to perform health early warning according to the health early warning model and the feature fusion data to obtain a health early warning result, and to optimize the leakage detection and positioning module according to the health early warning result. The intelligent control module is configured to generate an optimal control strategy for the water supply network according to the leakage detection result and the leakage positioning data, and to intelligently control the water supply network according to the optimal control strategy for the water supply network.

[0006] Further, the preprocessing module is configured to preprocess the multi-modal data of the water supply network by using a preprocessing method. The preprocessing method comprises: Step A1: removing outliers from the multi-modal data of the water supply network to obtain outlier-removed data; Step A2: inputting the outlier-removed data into a signal denoising model to perform signal denoising to obtain signal denoised data; Step A3: inputting the signal denoised data into a space-time synchronization model to perform space-time synchronization to obtain preprocessed multi-modal data.

[0007] Further, in the step A1 of removing outliers from the multi-modal data of the water supply network, for the pressure data and the flow data: inputting the pressure data and the flow data into an outlier removal window, calculating the mean value of the pressure data and the flow data in the outlier removal window respectively, and the standard deviation B, data points exceeding the range of mean value ± 3 times the standard deviation are determined as outliers, and data points not exceeding the range of mean value ± 3 times the standard deviation are determined as valid data, and the median of the valid data before and after the outliers is used to replace the outliers to obtain outlier-removed pressure data and outlier-removed flow data, and the outlier removal window time Yt is set to 5 seconds, , wherein is the number of data points in the outlier removal window, is the i-th data point in the outlier removal window, is the summation symbol, is the sample size, is the deviation from the mean, is the square of the deviation from the mean, is the sum of squares of deviations from the mean, is the sample variance.

[0008] Further, for the acoustic wave data: according to the range of the acoustic wave sensor and the normal signal range, a preset upper limit amplitude F1 and a preset lower limit amplitude F2 are set, sampling points exceeding the preset upper limit amplitude F1 and the preset lower limit amplitude F2 are marked as acoustic wave outliers, and the acoustic wave outliers are set to zero to obtain outlier-removed acoustic wave data. The outlier-removed pressure data, the outlier-removed flow data, and the outlier-removed acoustic wave data are output as the outlier-removed data.

[0009] Further, when the leakage detection and positioning module detects and locates the leakage event occurrence in real time according to the feature fusion data, the feature fusion data is compared with the leakage feature data set, the leakage event occurrence is judged according to the comparison result, and the judgment is output according to the judgment, wherein: When there is preset leakage feature fusion data consistent with the feature fusion data in the leakage feature data set, it is determined that the leakage event occurrence is a leakage event, the feature fusion data is marked as leakage event data, and the leakage event is identified and positioned according to the leakage event data; When there is no preset leakage feature fusion data consistent with the feature fusion data in the leakage feature data set, it is determined that the leakage event occurrence is not a leakage event, and the judgment is not output.

[0010] Further, when the leakage detection and positioning module identifies and locates the leakage event, the historical leakage database is divided into 80% leakage training set, 15% leakage verification set and 5% leakage test set, the leakage training set is input into the deep belief network model to train the deep belief network model to obtain the trained deep belief network model, the leakage verification set is input into the trained deep belief network model to iteratively optimize the trained deep belief network model to obtain the iteratively optimized deep belief network model, the leakage test set is input into the iteratively optimized deep belief network model for identification and positioning test to obtain the identification and positioning test result, the identification and positioning accuracy C is calculated according to the identification and positioning test result, the total number of identification and positioning test results is C1, the correct number in the identification and positioning test result is C2, C=C2 / C1x100%, the identification and positioning accuracy C is compared with the preset identification and positioning accuracy C0, the compliance of the iteratively optimized deep belief network model is judged according to the comparison result, and the iteratively optimized deep belief network model is output according to the judgment result, wherein: When C≥C0, it is determined that the compliance of the iteratively optimized deep belief network model is training compliance, the iteratively optimized deep belief network model is output as a leakage identification and positioning model, and the leakage event data is input into the leakage identification and positioning model for identification and positioning, and the leakage detection result and the leakage positioning data are output by the leakage identification and positioning model; When C<C0, it is determined that the compliance of the iteratively optimized deep belief network model is training non-compliance, the historical leakage database is updated, and the iteratively optimized deep belief network model is trained according to the updated historical leakage database until the compliance of the iteratively optimized deep belief network model is training compliance.

[0011] Further, when constructing the health early warning model, the health early warning management module divides the health early warning feature database into a 90% health training set and a 10% health test set, inputs the health training set into the hidden Markov model to train the hidden Markov model to obtain a trained hidden Markov model, inputs the health test set into the trained hidden Markov model to perform health early warning testing, obtains a health early warning test result, calculates a health early warning accuracy J according to the health early warning test result, sets a total number of health early warning test results as Jz, a correct number in the health early warning test results as Jq, J = Jq / Jz x 100%, compares the health early warning accuracy J with a preset health early warning accuracy J0, judges a compliance condition of the iteratively optimized hidden Markov model according to a comparison result, and outputs the iteratively optimized hidden Markov model according to a judgment result, wherein: When J ≥ J0, it is determined that the compliance condition of the iteratively optimized hidden Markov model is training compliance, and the trained hidden Markov model is output as the health early warning model; When J < J0, it is determined that the compliance condition of the iteratively optimized hidden Markov model is training non-compliance, the health early warning feature database is updated, and the hidden Markov model is trained according to the updated health early warning feature database until the compliance condition of the hidden Markov model is training compliance.

[0012] Further, when performing health early warning according to the health early warning model and the feature fusion data, the health early warning management module inputs the feature fusion data into the health early warning model, outputs a health early warning result from the health early warning model, and optimizes the leakage detection and positioning module according to the health early warning result, wherein: When the health early warning result is a health state in a warning period, the leakage detection and positioning module is not optimized; When the health early warning result is a crack accelerated expansion state in a warning period, the leakage detection and positioning module is optimized, and the feature fusion data corresponding to the health early warning result is added to the leakage and loss feature data set; When the health early warning result is a critical failure state in a warning period, the leakage detection and positioning module is optimized, the feature fusion data corresponding to the health early warning result is marked as special data and added to the leakage and loss feature data set, and an emergency protection strategy is triggered when the comparison result in the real-time detection and positioning process of the leakage detection and positioning module is special data.

[0013] Further, when generating the water supply network control strategy according to the leakage detection result and the leakage positioning data, the intelligent control module divides the leakage maintenance database into a 75% control training set and a 25% control test set, inputs the control training set into the decision tree model to train the decision tree model to obtain a trained decision tree model, inputs the control test set into the trained decision tree model to test the control strategy, obtains a control strategy test result, calculates a control strategy accuracy E according to the control strategy test result, sets a total number of the control strategy test result as Ez, a correct number in the control strategy test result as Eq, and E = Eq / Ez x 100%, compares the control strategy accuracy E with a preset control strategy accuracy E0, judges a standard reaching condition of the iterative optimization decision tree model according to the comparison result, and outputs the iterative optimization decision tree model according to the judgment result, wherein: when E ≥ E0, it is determined that the standard reaching condition of the iterative optimization decision tree model is training reaching standard, the trained decision tree model is output as an intelligent strategy model, and the leakage detection result and the leakage positioning data are input into the intelligent strategy model to output a first water supply network control strategy; when E < E0, it is determined that the standard reaching condition of the iterative optimization decision tree model is training not reaching standard, the leakage maintenance database is updated, and the decision tree model is trained according to the updated leakage maintenance database until the standard reaching condition of the decision tree model is training reaching standard.

[0014] Further, when generating the optimal water supply network control strategy according to the leakage detection result and the leakage positioning data, the intelligent control module further acquires a maintenance efficiency β of the first water supply network control strategy, compares the maintenance efficiency β of the first water supply network control strategy with a preset maintenance efficiency β0, judges an application condition of the first water supply network control strategy according to the comparison result, and outputs the first water supply network control strategy according to the judgment result, wherein: when β ≥ β0, it is determined that the application condition of the first water supply network control strategy is applicable, the first water supply network control strategy is output as the optimal water supply network control strategy, and the intelligent control module intelligently controls the water supply network according to the optimal water supply network control strategy; when β < β0, it is determined that the application condition of the first water supply network control strategy is not applicable, and the output mode of the first water supply network control strategy is that the preset control strategy accuracy E0 is adjusted, an adjusted preset control strategy accuracy is set as E0`, E0` = 1.46-0.34e （β-β0） , the intelligent control module is re-executed according to the adjusted preset control strategy accuracy E0` until the application condition of the first water supply network control strategy is applicable.

[0015] Compared with the prior art, the system provides precise and reliable raw data support for water supply network leakage detection, health early warning and intelligent control through the comprehensive and real-time data acquisition of the multi-modal real-time acquisition module, provides a high-quality data basis for subsequent core functions such as leakage detection and health early warning through the purification and quality improvement, space-time alignment and feature enhancement of the raw data by the preprocessing module, thereby reducing the subsequent processing complexity, and enhancing the adaptability of the system to different working conditions and different sensor models, so as to reduce the pipe network leakage rate, improve the water resource utilization efficiency and the pipe network operation stability, the system provides reliable decision basis for water supply network leakage prevention and control through the accuracy of leakage event identification and positioning and the self-adaptive evolution ability of the model of the leakage detection positioning module, thereby reducing the pipe network leakage rate, improving the water resource utilization efficiency and the pipe network operation stability, the system provides a closed-loop reinforcement of the health state of the water supply network through the health early warning management module, and the health state of the water supply network is transformed from passive repair to active prevention, thereby reducing the pipe network leakage rate, improving the water resource utilization efficiency and the pipe network operation stability, and the system provides core execution support for the rapid disposal after the occurrence of leakage events, the rational allocation of resources and the stable operation of the pipe network through the generation and optimization of the control strategy of the intelligent control module, so as to further reduce the pipe network leakage rate, improve the water resource utilization efficiency and the pipe network operation stability. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A structure schematic view of the water supply network leakage positioning and early warning system of the sound pressure flow multi-modal sensing of the embodiment; Figure 2 A flow schematic view of the preprocessing method of the embodiment. DETAILED DESCRIPTION

[0017] In order to make the purpose and advantages of the present application more clear and obvious, the present application is further described below in combination with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the protection scope of the present application.

[0018] The preferred embodiments of the present application are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present application, and are not used to limit the protection scope of the present application.

[0019] It should be noted that in the description of the present application, the terms "up", "down", "left", "right", "in", "out" and the like indicate the direction or positional relationship terms based on the direction or positional relationship shown in the drawings, which are only for the convenience of description, and are not indicative or imply that the device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0020] In addition, it should be noted that in the description of the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connecting" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integrally connected; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through intermediate medium, or internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0021] Please refer to Figure 1 The system includes: A multi-modal real-time acquisition module is configured to acquire multi-modal data of the water supply network in real time to obtain multi-modal data of the water supply network. A preprocessing module is configured to preprocess the multi-modal data of the water supply network to obtain preprocessed multi-modal data, and to perform feature fusion on the preprocessed multi-modal data to obtain feature fusion data. The preprocessing module is connected to the multi-modal real-time acquisition module. A leakage detection and positioning module is configured to detect leakage events in real time according to the feature fusion data, and to identify and locate the leakage events to obtain leakage detection results and leakage positioning data. The leakage detection and positioning module is connected to the preprocessing module. A health warning management module is configured to construct a health warning model, and to perform health warning according to the health warning model and the feature fusion data to obtain health warning results. The health warning management module is connected to the leakage detection and positioning module. An intelligent control module is configured to generate an optimal control strategy for the water supply network according to the leakage detection results and the leakage positioning data, and to perform intelligent control on the water supply network according to the optimal control strategy for the water supply network. The intelligent control module is connected to the health warning management module.

[0022] In particular, the system is applied to a water supply network intelligent operation and maintenance terminal, such as a water supply network operation monitoring center. The system collects multi-modal data of the water supply network in real time through a multi-modal real-time collection module, completes data preprocessing and feature fusion through a preprocessing module to provide a high-quality data basis for subsequent analysis, realizes real-time identification and accurate positioning of leakage events by a leakage detection and positioning module, solves the problems of low efficiency and lagging leakage discovery in traditional manual inspection, and at the same time, a health early warning management module builds a health early warning model to predict the health risk of the pipe network in advance and optimize the leakage detection and positioning module. Finally, an intelligent control module generates an optimal control strategy to realize active prevention and control and efficient operation and maintenance of water supply network leakage, so as to reduce the pipe network leakage rate, improve water resource utilization efficiency and pipe network operation stability. In particular, the comprehensiveness and real-time nature of data collection by the multi-modal real-time collection module provide accurate and reliable raw data support for water supply network leakage detection, health early warning and intelligent control. The preprocessing module purifies, upgrades, aligns in time and space, and enhances the features of the raw data to provide a high-quality data basis for subsequent core functions such as leakage detection and health early warning, thereby reducing the complexity of subsequent processing and enhancing the adaptability of the system to different working conditions and different sensor models to reduce the pipe network leakage rate, improve water resource utilization efficiency and pipe network operation stability. The leakage detection and positioning module provides reliable decision-making basis for water supply network leakage prevention and control by accurately identifying leakage events and accurately positioning the model to reduce the pipe network leakage rate, improve water resource utilization efficiency and pipe network operation stability. The health early warning management module provides a closed-loop reinforcement of dynamic optimization and risk prevention for the health state of the water supply network, transforming the operation and maintenance of the water supply network from passive repair to active prevention to reduce the pipe network leakage rate, improve water resource utilization efficiency and pipe network operation stability. The intelligent control module generates and optimizes the control strategy of the water supply network to provide core execution support for rapid disposal after leakage events, rational allocation of resources and stable operation of the pipe network to further reduce the pipe network leakage rate, improve water resource utilization efficiency and pipe network operation stability.

[0023] In particular, when the multi-modal real-time collection module collects multi-modal data of the water supply network, the multi-modal data of the water supply network includes sound wave data, pressure data and flow data.

[0024] Specifically, the acoustic data refers to the stress wave signal excited by fluid leakage and pipe wall vibration inside the pipe; the pressure data refers to the physical quantity reflecting the pressure of the fluid inside the pipe; and the flow rate data refers to the volume of fluid passing through the pipe cross-section per unit time. This embodiment does not limit the real-time acquisition method of the multimodal data of the water pipe network. For example, it can be set to acquire the acoustic data in the multimodal data of the water pipe network through a distributed fiber optic acoustic sensor array, acquire the pressure data in the multimodal data of the water pipe network through a high-frequency pressure transmitter, and acquire the flow rate data in the multimodal data of the water pipe network through an electromagnetic flow meter.

[0025] Specifically, the multimodal real-time acquisition module provides accurate and reliable raw data support for water supply network leakage detection, health early warning, and intelligent control through the comprehensiveness and real-time nature of data acquisition.

[0026] Specifically, when the preprocessing module preprocesses the multimodal data of the water supply network, it does so by using a preprocessing method.

[0027] Please see Figure 2 As shown, this is a flowchart illustrating the preprocessing method of this embodiment. The preprocessing method includes: Step A1: Remove outliers from the multimodal data of the water supply network to obtain outlier-removed data; Step A2: Input the outlier-removed data into the signal denoising model for signal denoising to obtain the denoised signal data; Step A3: Input the signal noise reduction data into the spatiotemporal synchronization model for spatiotemporal synchronization to obtain preprocessed multimodal data.

[0028] Specifically, in step A1, when removing outliers from the multimodal data of the water supply network, for pressure and flow data: the pressure and flow data are input into the outlier removal window, and the mean values ​​of the pressure and flow data are calculated separately within the outlier removal window. The sum and standard deviation B will exceed the mean. ±3 standard deviations Data points within the specified range are considered outliers and will be excluded if they do not exceed the mean. ±3 standard deviations Data points within the specified range are considered valid data, and the outlier is replaced by the median of the valid data before and after it, resulting in outlier removal pressure data and outlier removal flow data. The outlier removal window time is set to Yt, where Yt = 5 seconds. , ,in It is the number of data points within the outlier removal window. It is the i-th data point within the outlier removal window. It is the summation symbol. is a sample size, is a mean deviation, is a square of the mean deviation, is a sum of squares of the mean deviation, is a sample variance; For acoustic wave data: according to the range of the acoustic wave sensor and the normal signal range, a preset upper limit amplitude F1 and a preset lower limit amplitude F2 are set, the sampling points exceeding the preset upper limit amplitude F1 and the preset lower limit amplitude F2 are marked as acoustic wave outliers, and the acoustic wave outliers are zeroed to obtain outlier-removed acoustic wave data; The outlier-removed pressure data, the outlier-removed flow data and the outlier-removed acoustic wave data are output as outlier-removed data.

[0029] Specifically, the signal denoising model refers to a set of digital signal processing algorithms designed for different modal data characteristics, and the construction method of the signal denoising model is not limited in the embodiment, such as an adaptive weighted Kalman filtering algorithm can be used to construct the signal denoising model. The space-time synchronization model refers to a processing framework that ensures that heterogeneous data from different geographical locations and different sampling devices can be accurately aligned in time and space, and the construction method of the space-time synchronization model is not limited in the embodiment, such as a time series convolution network model can be used to construct the space-time synchronization model. The outlier removal window refers to a continuous and fixed time length of time data buffer used when performing outlier removal processing. The outlier removal window time refers to the length of time covered by the outlier removal window. The sample size refers to the total number of pressure or flow data points contained in the current outlier removal window. The mean deviation refers to the difference between the value of a data point in the outlier removal window and the arithmetic mean of all data points in the window. The square of the mean deviation refers to squaring the mean deviation. The sum of squares of the mean deviation refers to adding all the square values of the mean deviation of each data point in the outlier removal window.

[0030] Specifically, the preprocessing module purifies and upgrades the original data, aligns in time and space, and enhances features, thereby providing a high-quality data basis for subsequent leakage and loss detection, health warning and other core functions, reducing the complexity of subsequent processing, and enhancing the adaptability of the system to different working conditions and different sensor models, thereby reducing the pipe network leakage rate, improving water resource utilization efficiency and pipe network operation stability.

[0031] Specifically, when the leakage detection and positioning module detects and locates the leakage event occurrence according to the feature fusion data, the feature fusion data is compared with the leakage feature data set, the leakage event occurrence is judged according to the comparison result, and the judgment is output, wherein: When the preset leakage feature fusion data consistent with the feature fusion data exists in the leakage feature data set, it is determined that the leakage event occurrence condition is a leakage event occurrence, the feature fusion data is marked as a leakage event data, and the leakage event is identified and positioned according to the leakage event data; When the preset leakage feature fusion data consistent with the feature fusion data does not exist in the leakage feature data set, it is determined that the leakage event occurrence condition is that the leakage event does not occur, and the judgment is not output; When the leakage detection and positioning module identifies and positions the leakage event, the historical leakage database is divided into 80% leakage training set, 15% leakage verification set and 5% leakage test set, the leakage training set is input into the deep belief network model to train the deep belief network model to obtain the trained deep belief network model, the leakage verification set is input into the trained deep belief network model to iteratively optimize the trained deep belief network model to obtain the iteratively optimized deep belief network model, the leakage test set is input into the iteratively optimized deep belief network model for identification and positioning test to obtain the identification and positioning test result, the identification and positioning accuracy C is calculated according to the identification and positioning test result, the total number of identification and positioning test results is C1, the correct number in the identification and positioning test result is C2, C=C2 / C1x100%, the identification and positioning accuracy C is compared with the preset identification and positioning accuracy C0, the conformity of the iteratively optimized deep belief network model is judged according to the comparison result, and the iteratively optimized deep belief network model is output according to the judgment result, wherein: When C≥C0, it is determined that the conformity of the iteratively optimized deep belief network model is training conformity, the iteratively optimized deep belief network model is output as a leakage identification and positioning model, and the leakage event data is input into the leakage identification and positioning model for identification and positioning, and the leakage detection result and the leakage positioning data are output by the leakage identification and positioning model; When C<C0, it is determined that the conformity of the iteratively optimized deep belief network model is training non-conformity, the historical leakage database is updated, and the iteratively optimized deep belief network model is trained according to the updated historical leakage database until the conformity of the iteratively optimized deep belief network model is training conformity.

[0032] Specifically, the leakage feature dataset refers to a pre-constructed, standardized knowledge base that stores verified feature fusion data templates corresponding to various typical leakage events, the preset leakage feature fusion data refers to a single leakage feature template stored in the leakage feature dataset, the historical leakage database refers to a database that accumulates confirmed leakage event cases, the leakage training set refers to part of the data used to directly train the deep belief network model, the leakage validation set refers to part of the data used to independently evaluate the performance of the deep belief network model during the training process and guide iterative optimization, the leakage test set refers to part of the data used to perform a final one-time, unbiased evaluation of the iteratively optimized deep belief network model, the deep belief network model refers to a deep learning model stacked by multiple layers of restricted Boltzmann machines, the iterative optimization refers to a process of repeatedly adjusting the hyperparameters of the deep belief network model according to the performance on the leakage validation set to seek the best performance of the deep belief network model on the validation set, the recognition and positioning accuracy C refers to a core quantitative indicator for evaluating the performance of the deep belief network model, the preset recognition and positioning accuracy C0 refers to a pre-set minimum performance threshold that the deep belief network model must reach before being put into practical application, for example, the preset recognition and positioning accuracy C0 = 92%, the leakage detection result refers to a qualitative judgment of whether a leakage event has occurred output by the leakage recognition and positioning model, and the leakage positioning data refers to quantitative information about the spatial position of a leakage point output by the leakage recognition and positioning model, such as pipe segment number, distance from a reference point, and three-dimensional coordinates. The embodiment does not limit the way in which the historical leakage database is updated, and the historical leakage database can be updated through big data, for example.

[0033] Specifically, the leakage detection and positioning module provides reliable decision-making basis for water supply network leakage prevention and control through the accuracy of leakage event recognition, the accuracy of positioning, and the adaptive evolution ability of the model, thereby reducing the pipe network leakage rate, improving water resource utilization efficiency, and improving pipe network operation stability.

[0034] Specifically, when constructing the health early warning model, the health early warning management module divides the health early warning feature database into a 90% health training set and a 10% health test set, inputs the health training set into the hidden Markov model to train the hidden Markov model to obtain a trained hidden Markov model, inputs the health test set into the trained hidden Markov model to perform health early warning testing, obtains a health early warning test result, calculates a health early warning accuracy J according to the health early warning test result, sets a total number of health early warning test results as Jz, a correct number in the health early warning test results as Jq, J = Jq / Jz x 100%, compares the health early warning accuracy J with a preset health early warning accuracy J0, judges a compliance condition of the iteratively optimized hidden Markov model according to a comparison result, and outputs the iteratively optimized hidden Markov model according to a judgment result, wherein: When J ≥ J0, it is determined that the compliance condition of the iteratively optimized hidden Markov model is training compliance, and the trained hidden Markov model is output as the health early warning model; When J < J0, it is determined that the compliance condition of the iteratively optimized hidden Markov model is training non-compliance, the health early warning feature database is updated, and the hidden Markov model is trained according to the updated health early warning feature database until the compliance condition of the hidden Markov model is training compliance; When performing health early warning according to the health early warning model and the feature fusion data, the health early warning management module inputs the feature fusion data into the health early warning model, outputs a health early warning result from the health early warning model, and optimizes the leakage detection and positioning module according to the health early warning result, wherein: When the health early warning result is a health maintaining state in a warning period, the leakage detection and positioning module is not optimized; When the health early warning result is a crack accelerating expansion state in a warning period, the leakage detection and positioning module is optimized, and the feature fusion data corresponding to the health early warning result is added to the leakage and loss feature data set; When the health early warning result is a critical failure maintaining state in a warning period, the leakage detection and positioning module is optimized, the feature fusion data corresponding to the health early warning result is marked as special data and added to the leakage and loss feature data set, and an emergency protection strategy is triggered when the comparison result in the real-time detection and positioning process of the leakage detection and positioning module is special data.

[0035] Specifically, the health warning feature database refers to a database for training and evaluating a health warning model, the health training set refers to a subset of data for training parameters of a hidden Markov model, the health test set refers to a subset of data for final performance evaluation of the trained hidden Markov model, the hidden Markov model refers to a statistical model for describing a Markov process containing unknown parameters, the health warning accuracy J refers to a key indicator for evaluating the performance of the health warning model, and the preset health warning accuracy J0 refers to the minimum accuracy threshold that the trained hidden Markov model must reach to be determined as usable for actual warning, for example, the preset health warning accuracy J0 = 94%, and the embodiment does not limit the updating manner of the health warning feature database, which can be updated by an administrator.

[0036] Specifically, the health warning management module transforms the operation and maintenance of the water supply network from passive repair to active prevention through forward-looking prediction of the health status of the water supply network and dynamic optimization and closed-loop reinforcement of risk prevention, thereby reducing the leakage rate of the network, improving the efficiency of water resource utilization, and improving the stability of the network operation.

[0037] Specifically, when the intelligent control module generates the water supply network control strategy according to the leakage detection result and the leakage positioning data, the leakage maintenance database is divided into a 75% control training set and a 25% control test set, the control training set is input into the decision tree model to train the decision tree model to obtain a trained decision tree model, the control test set is input into the trained decision tree model to test the control strategy, a control strategy test result is obtained, the control strategy accuracy E is calculated according to the control strategy test result, the total number of control strategy test results is set as Ez, the correct number in the control strategy test result is set as Eq, E = Eq / Ez x 100%, the control strategy accuracy E is compared with the preset control strategy accuracy E0, the compliance of the iterative optimization decision tree model is judged according to the comparison result, and the iterative optimization decision tree model is output according to the judgment result, wherein: When E ≥ E0, it is determined that the compliance of the iterative optimization decision tree model is training compliance, the trained decision tree model is output as an intelligent strategy model, and the leakage detection result and the leakage positioning data are input into the intelligent strategy model to output a first water supply network control strategy; When E < E0, it is determined that the compliance of the iterative optimization decision tree model is training non-compliance, the leakage maintenance database is updated, and the decision tree model is trained according to the updated leakage maintenance database until the compliance of the decision tree model is training compliance. The intelligent control module also acquires the maintenance efficiency β of the first water supply network control strategy when generating the optimal control strategy of the water supply network according to the leakage detection result and the leakage positioning data, compares the maintenance efficiency β of the first water supply network control strategy with the preset maintenance efficiency β0, judges the applicability of the first water supply network control strategy according to the comparison result, and outputs the first water supply network control strategy according to the judgment result, wherein: When β≥β0, it is determined that the applicability of the first water supply network control strategy is applicable, and the first water supply network control strategy is output as the optimal control strategy of the water supply network, and the intelligent control module intelligently controls the water supply network according to the optimal control strategy of the water supply network; When β<β0, it is determined that the applicability of the first water supply network control strategy is not applicable, and the output mode of the first water supply network control strategy is that the preset control strategy accuracy E0 is adjusted, the adjusted preset control strategy accuracy is E0`, E0`=1.46-0.34e （β-β0） , and the intelligent control module is re-executed according to the adjusted preset control strategy accuracy E0` until the applicability of the first water supply network control strategy is applicable.

[0038] Specifically, the leakage maintenance database refers to a knowledge base storing historical leakage events and corresponding treatment schemes and effects, the control training set refers to a data subset for training a decision tree model, the control test set refers to a data subset for performance evaluation of the trained decision tree model, the decision tree model refers to a tree-structured classification and regression model, the control strategy accuracy E refers to an index for evaluating the performance of the decision tree model, and the first water supply network control strategy refers to a specific control instruction set generated by the intelligent strategy model according to real-time input leakage detection results and leakage positioning data, for example, reducing the inlet pressure set point of the target partition to 0.35 MPa, closing the isolation valve at the target partition, and notifying the maintenance team to go to the target partition for maintenance. The embodiment does not limit the updating method of the leakage maintenance database, which can be updated by integrating the work order management system to automatically record the complete process of each leakage maintenance back to the database. The maintenance efficiency β of the first water supply network control strategy refers to an index for quantitatively evaluating the comprehensive effect of the generated control strategy. The embodiment does not limit the acquisition method of the maintenance efficiency β of the first water supply network control strategy, which can be set to calculate the maintenance efficiency β of the first water supply network control strategy in real time by substituting the first water supply network control strategy into a preset formula. The preset maintenance efficiency β0 refers to a preset acceptable minimum maintenance efficiency threshold, for example, the preset maintenance efficiency β0 = 0.86. The embodiment does not limit the specific method of intelligent control of the water supply network according to the optimal control strategy of the water supply network, which can be set to issue the optimal control strategy to the programmable logic controller and frequency converter on site through the industrial control bus, and automatically execute the optimal control strategy by the programmable logic controller and frequency converter.

[0039] Specifically, the intelligent control module provides core execution support for rapid disposal after leakage, rational allocation of resources, and stable operation of the pipe network through generation and optimization of the water supply network control strategy, so as to further reduce the pipe network leakage rate, improve water resource utilization efficiency, and improve pipe network operation stability.

[0040] So far, the technical solutions of the present application have been described in conjunction with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will fall within the protection scope of the present application.

Claims

1. A water supply network leakage location and early warning system based on sound pressure flow multimodal sensing, characterized in that, The system includes: The multimodal real-time acquisition module is used to acquire multimodal data of the water supply network in real time and obtain multimodal data of the water supply network. The preprocessing module is used to preprocess the multimodal data of the water supply network to obtain preprocessed multimodal data, and also to perform feature fusion on the preprocessed multimodal data to obtain feature fused data. The leak detection and location module is used to detect the occurrence of leakage events in real time based on feature fusion data, identify and locate the leakage events, and obtain leakage detection results and leakage location data. The health early warning management module is used to build a health early warning model, and also to make health early warnings based on the health early warning model and feature fusion data, obtain health early warning results, and optimize the leak detection and location module based on the health early warning results; The intelligent control module is used to generate the optimal control strategy for the water supply network based on the leak detection results and leak location data, and to perform intelligent control of the water supply network according to the optimal control strategy.

2. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 1, characterized in that, When the preprocessing module preprocesses the multimodal data of the water supply network, it preprocesses the multimodal data of the water supply network using a preprocessing method. The preprocessing method includes: Step A1: Remove outliers from the multimodal data of the water supply network to obtain outlier-removed data; Step A2: Input the outlier-removed data into the signal denoising model for signal denoising to obtain the denoised signal data; Step A3: Input the signal noise reduction data into the spatiotemporal synchronization model for spatiotemporal synchronization to obtain preprocessed multimodal data.

3. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 2, characterized in that, In step A1, when removing outliers from the multimodal data of the water supply network, for pressure and flow data: the pressure and flow data are input into the outlier removal window, and the mean values ​​of the pressure and flow data are calculated separately within the outlier removal window. The sum and standard deviation B will exceed the mean. ±3 standard deviations Data points within the specified range are considered outliers and will be excluded if they do not exceed the mean. ±3 standard deviations Data points within the specified range are considered valid data, and the outlier is replaced by the median of the valid data before and after it, resulting in outlier removal pressure data and outlier removal flow data. The outlier removal window time is set to Yt, where Yt = 5 seconds. , ,in It is the number of data points within the outlier removal window. It is the i-th data point within the outlier removal window. It is the summation symbol. It is the sample size. It is the deviation from the mean. It is the square of the deviation from the mean. It is the sum of squared deviations from the mean. It is the sample variance.

4. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 2, characterized in that, For acoustic wave data: Based on the range of the acoustic wave sensor and the normal signal range, a preset upper limit F1 and a preset lower limit F2 are set. Sampling points that exceed the preset upper limit F1 and the preset lower limit F2 are marked as acoustic wave outliers, and the acoustic wave outliers are set to zero to obtain outlier-removed acoustic wave data. The outlier-removed pressure data, outlier-removed flow data, and outlier-removed acoustic data are output as outlier-removed data.

5. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 1, characterized in that, When the leakage detection and localization module performs real-time detection and localization of leakage events based on feature fusion data, it compares the feature fusion data with the leakage feature dataset, judges the occurrence of leakage events based on the comparison results, and outputs the judgment accordingly. When there is preset leakage feature fusion data in the leakage feature dataset that is consistent with the feature fusion data, the leakage event is determined to have occurred. The feature fusion data is marked as leakage event data, and the leakage event is identified and located based on the leakage event data. When there is no preset leak feature fusion data in the leak feature dataset that is consistent with the feature fusion data, the leak event is determined to have not occurred, and no judgment is output.

6. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 5, characterized in that, When the leakage detection and localization module identifies and locates leakage events, it divides the historical leakage database into an 80% training set, a 15% validation set, and a 5% test set. The training set is input into a deep belief network model to train the model, resulting in a trained deep belief network model. The validation set is input into the trained deep belief network model for iterative optimization, resulting in an iteratively optimized deep belief network model. The test set is input into the iteratively optimized deep belief network model for identification and localization testing, yielding the test results. The identification and localization accuracy C is calculated based on these results, with the total number of test results set as C1 and the number of correct results set as C2, where C = C2 / C1 × 100%. The accuracy C is compared to a preset accuracy C0. Based on the comparison results, the performance of the iteratively optimized deep belief network model is assessed, and the model is output based on the assessment results. When C≥C0, the training of the iteratively optimized deep belief network model is deemed to have met the target. The iteratively optimized deep belief network model is then used as the output of the leak identification and localization model. Leak event data is input into the leak identification and localization model for identification and localization. The leak identification and localization model then outputs the leak detection results and leak location data. When C < C0, the training failure status of the iteratively optimized deep belief network model is determined. The historical leakage database is updated, and the iteratively optimized deep belief network model is trained based on the updated historical leakage database until the training failure status of the iteratively optimized deep belief network model is determined.

7. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 1, characterized in that, When constructing the health early warning model, the health early warning feature database is divided into a 90% health training set and a 10% health test set. The health training set is input into the Hidden Markov Model (HMM) to train the HMM, resulting in a trained HMM. The health test set is input into the trained HMM to conduct health early warning tests, obtaining the test results. The health early warning accuracy J is calculated based on the test results, with the total number of health early warning test results set as Jz and the number of correct results set as Jq, where J = Jq / Jz × 100%. The health early warning accuracy J is compared with the preset health early warning accuracy J0. Based on the comparison results, the compliance of the iteratively optimized HMM is judged, and the iteratively optimized HMM is output based on the judgment results. When J≥J0, the training of the iteratively optimized Hidden Markov Model is considered to have met the target, and the trained Hidden Markov Model is output as a health warning model. When J < J0, the training failure is determined as the objective of the iterative optimization of the Hidden Markov Model. The health warning feature database is then updated, and the Hidden Markov Model is trained based on the updated health warning feature database until the training success is determined as the objective of the Hidden Markov Model.

8. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 7, characterized in that, When the health early warning management module issues a health early warning based on the health early warning model and feature fusion data, it inputs the feature fusion data into the health early warning model, which then outputs the health early warning result. Based on the health early warning result, the module optimizes the leak detection and location module, wherein: When the health warning result indicates that the system remains healthy during the warning period, the leak detection and location module will not be optimized. When the health warning result indicates that the crack is in an accelerated propagation state within the warning period, the leakage detection and localization module is optimized, and the feature fusion data corresponding to this health warning result is added to the leakage feature dataset. When the health warning result indicates a critical failure state within the warning period, the leak detection and location module is optimized. The feature fusion data corresponding to this health warning result is marked as special data and added to the leak feature dataset. When the comparison result during the real-time detection and location process of the leak detection and location module is special data, an emergency protection strategy is triggered.

9. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 1, characterized in that, When the intelligent control module generates a water supply network control strategy based on leak detection results and leak location data, it divides the leak repair database into a 75% control training set and a 25% control test set. The control training set is input into a decision tree model to train the model, resulting in a trained decision tree model. The control test set is input into the trained decision tree model to test the control strategy, obtaining the test results. The accuracy rate E of the control strategy is calculated based on these test results. The total number of control strategy test results is set as Ez, and the number of correct results is set as Eq, where E = Eq / Ez × 100%. The accuracy rate E is compared with the preset accuracy rate E0. Based on the comparison results, the compliance of the iteratively optimized decision tree model is judged, and the iteratively optimized decision tree model is output based on the judgment result. When E≥E0, the success of the iterative optimization decision tree model is determined to be training success. The trained decision tree model is output as the intelligent strategy model, and the leakage detection results and leakage location data are input into the intelligent strategy model. The intelligent strategy model outputs the first water supply network control strategy. When E < E0, the training of the iteratively optimized decision tree model is deemed unsatisfactory. The leakage repair database is then updated, and the decision tree model is trained based on the updated leakage repair database until the training of the decision tree model is deemed satisfactory.

10. The water supply network leakage location and early warning system based on sound pressure flow multimodal sensing according to claim 9, characterized in that, When the intelligent control module generates the optimal control strategy for the water supply network based on the leak detection results and leak location data, it also acquires the maintenance efficiency β of the first water supply network control strategy, compares the maintenance efficiency β of the first water supply network control strategy with the preset maintenance efficiency β0, judges the applicability of the first water supply network control strategy based on the comparison result, and outputs the first water supply network control strategy based on the judgment result, wherein: When β≥β0, the first water supply network control strategy is deemed applicable, and the first water supply network control strategy is output as the optimal control strategy for the water supply network. The intelligent control module performs intelligent control of the water supply network according to the optimal control strategy for the water supply network. When β < β0, the first water supply network control strategy is deemed inapplicable. The output method for the first water supply network control strategy is as follows: the accuracy E0 of the preset control strategy is adjusted, and the adjusted accuracy is set to E0`, where E0` = 1.46 - 0.34e （β-β0） The intelligent control module is re-executed based on the accuracy E0` of the adjusted preset control strategy until the application of the first water supply network control strategy is deemed applicable.

Images