Large model-based urban and rural water supply pipe network leakage monitoring method

Abstract

The invention relates to pipeline leakage detection, in particular to an urban and rural water supply network leakage monitoring method based on a large model, and the method comprises the steps: constructing a water supply network leakage detection large model, obtaining real-time audio signals collected by each monitoring point of a water supply network, converting the real-time audio signals into a frequency spectrum thermodynamic diagram, and inputting the frequency spectrum thermodynamic diagram into the water supply network leakage detection large model. Monitoring the leakage of the water supply network by the large water supply network leakage detection model; determining an objective function of water supply network leakage monitoring based on a detection result of the large water supply network leakage detection model, determining constraint conditions of water supply network leakage monitoring, constructing a water supply network leakage monitoring model according to the objective function and the constraint conditions of water supply network leakage monitoring, and solving the water supply network leakage monitoring model to obtain a water supply network leakage monitoring result. Determining a specific leakage position; according to the technical scheme provided by the invention, the defects in the prior art that the specific leakage position cannot be accurately determined under the condition that the layout density of the acoustic sensor is reduced, and external noise interference is easily caused can be overcome.

Description

Technical Field

[0001] The invention relates to pipeline leakage detection, and in particular to a leakage monitoring method for urban and rural water supply pipe networks based on a large model. Background Art

[0002] Leakage detection and positioning of water supply networks has become an important task in water supply management systems. Leakage monitoring of large water supply networks is a challenging task. Reliable leakage detection and positioning technology is crucial to reducing losses in water supply networks.

[0003] Traditional acoustic leak detection technology uses acoustic sensors installed on the pipeline to measure the propagation time of sound waves using correlation functions. If the leak is located between two acoustic sensors, the sensor will detect the noise emitted by the leak, and then the correlation function can be used to calculate the time difference caused by the same leak noise propagating to the two sensors. Once the sound speed in the pipeline is determined, the specific location of the leak can be calculated.

[0004] However, the realization of acoustic leak detection technology requires a large number of hardware devices, resulting in a high density of acoustic sensor deployment, high detection costs, and susceptibility to external noise interference. However, if the density of acoustic sensors is reduced, the specific location of the leak point cannot be accurately measured. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a leakage monitoring method for urban and rural water supply pipe networks based on a large model, which can effectively overcome the defects of the prior art that the specific leakage location cannot be accurately determined when the density of acoustic sensor deployment is reduced, and the method is easily affected by external noise interference.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A method for monitoring leakage of urban and rural water supply networks based on a large model, which constructs a large model for detecting leakage of water supply networks, obtains real-time audio signals collected at each monitoring point of the water supply network, and converts the real-time audio signals into a spectrum heat map and inputs it into the large model for detecting leakage of water supply networks, and the large model for detecting leakage of water supply networks is used to monitor leakage of water supply networks;

[0010] Based on the detection results of the large model of water supply network leakage detection, the objective function of water supply network leakage monitoring is determined, and the constraints of water supply network leakage monitoring are determined. According to the objective function and constraints of water supply network leakage monitoring, a water supply network leakage monitoring model is constructed, the water supply network leakage monitoring model is solved, and the specific leakage location is determined.

[0011] Preferably, after constructing the large model for water supply network leakage detection, the method further comprises:

[0012] S11, collecting historical audio signals collected from each monitoring point of the water supply network, manually annotating the historical audio signals, and establishing an audio signal sample library;

[0013] S12, converting all historical audio signals in the audio signal sample library into spectrum heat maps, and establishing a spectrum heat map data set;

[0014] S13, using the spectrum heat map data set to train the water supply network leakage detection large model, to obtain a trained water supply network leakage detection large model;

[0015] Among them, the large model for water supply network leakage detection is a large model based on a multi-layer convolutional neural network CNN.

[0016] Preferably, in S12, all historical audio signals in the audio signal sample library are converted into spectrum heat maps, and a spectrum heat map data set is established, including:

[0017] S121, pre-emphasize the historical audio signal:

[0018] y(t)=x(t)-α·x(t-1);

[0019] Wherein, x(t) is the historical audio signal at time t, y(t) is the historical audio signal after pre-emphasis at time t, x(t-1) is the historical audio signal at time t-1, α is the pre-emphasis coefficient, α∈(0.9,1);

[0020] S122, performing a short-time Fourier transform (STFT) on the pre-emphasized historical audio signal to obtain a corresponding spectrum graph, and converting the amplitude of the spectrum graph into a logarithmic scale;

[0021] S123, using time as the abscissa and frequency as the ordinate, and using different colors to represent spectrum amplitudes, so as to convert the historical audio signal into a spectrum heat map;

[0022] S124. Repeat S121 to S123 to complete the conversion of all historical audio signals into spectrum heat maps, and establish a spectrum heat map data set.

[0023] Preferably, in S13, the spectrum heat map data set is used to perform model training on the water supply network leakage detection large model to obtain the trained water supply network leakage detection large model, including:

[0024] S131, dividing the spectrum heat map data set into a training set, a validation set and a test set according to a preset ratio;

[0025] S132, setting the loss function and optimizer of the large model for water supply network leakage detection, and inputting the training set into the large model for water supply network leakage detection to perform model training;

[0026] S133, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information;

[0027] S134. If the loss value is less than the preset threshold, the model training ends, and the current water supply network leakage detection large model is the trained water supply network leakage detection large model. Otherwise, return to S132 and continue to use the training set for model training.

[0028] S135, inputting the validation set into the trained large model for water supply network leakage detection, evaluating the generalization ability of the model by observing the performance on the validation set, and tuning the hyperparameters and structure of the model;

[0029] S136. Input the test set into the optimized large model for water supply network leakage detection to evaluate the model performance.

[0030] Preferably, in S132, the training set is input into a large model for water supply network leakage detection to perform model training, including:

[0031] S1321. Obtain the detection results of the water supply network leakage detection large model on the training set, draw the ROC curve, calculate the AUC area under the ROC curve, and use the probability value corresponding to the point with the largest AUC area as the probability threshold P th ;

[0032] S1322, after the training set data is input into the water supply network leakage detection model, if the detection result is greater than the probability threshold P th , it indicates that the training set data corresponds to the leakage label, otherwise the training set data corresponds to the normal label.

[0033] Preferably, the objective function of water supply network leakage monitoring is determined based on the detection results of the water supply network leakage detection large model, including:

[0034] S21. Constructing a water pressure distribution model of the water supply network under normal working conditions according to the topological relationship and operation parameters of the water supply network;

[0035] S22, based on the abnormal monitoring points detected by the large model for water supply network leakage detection, the water supply pipelines between the abnormal monitoring points are taken as target water supply pipelines;

[0036] S23, setting a plurality of sub-monitoring points in the target water supply pipeline, and collecting the actual water pressure at each sub-monitoring point;

[0037] S24, simulating various water supply network leakage conditions in the water supply network water pressure distribution model, and obtaining simulated water pressures of each sub-monitoring point under each water supply network leakage condition;

[0038] S25. Determine the objective function of water supply network leakage monitoring based on the actual water pressure and simulated water pressure of each sub-monitoring point.

[0039] Preferably, in S25, the objective function of water supply network leakage monitoring is determined according to the actual water pressure and simulated water pressure of each sub-monitoring point, including:

[0040] According to the actual water pressure and simulated water pressure of each sub-monitoring point, the objective function F of water supply network leakage monitoring is determined by the following formula:

[0041]

[0042] Among them, Ps mn is the simulated water pressure of the nth sub-monitoring point under the mth water supply network leakage condition, Pr mn is the actual water pressure of the nth sub-monitoring point under the mth water supply network leakage condition, N1 is the number of sub-monitoring points, and there are M water supply network leakage conditions in total.

[0043] Preferably, solving the water supply network leakage monitoring model to determine the specific leakage location includes:

[0044] The krill swarm algorithm (KHA) is used to solve the objective function F of water supply network leakage monitoring and obtain the specific leakage location that meets the constraint conditions of water supply network leakage monitoring.

[0045] Preferably, the krill swarm algorithm KHA is used to solve the objective function F of the water supply network leakage monitoring to obtain a specific leakage location that meets the constraint conditions of the water supply network leakage monitoring, including:

[0046] S31, initializing parameters and randomly generating a krill swarm of a certain size, and taking the initial position of each krill individual in the krill swarm as an initial solution;

[0047] S32, updating the position of each krill individual according to the induction, foraging and random physical diffusion of other krill individuals;

[0048] S33, using the objective function F of water supply network leakage monitoring to evaluate each krill individual and calculate the corresponding fitness value;

[0049] S34, record and update the global optimal solution;

[0050] S35. Determine whether the iteration termination condition is met. If the iteration termination condition is not met, return to S32. Otherwise, use the current global optimal solution as the specific leakage position that meets the constraint conditions of water supply network leakage monitoring.

[0051] Preferably, in S32, the position of each krill individual is updated according to the induction, foraging and random physical diffusion of other krill individuals, including:

[0052] S321. The movement of krill individuals will be induced by other krill individuals around them, forming induced movement:

[0053] N i+1 =ε i ·v max +ω n ·N i ;

[0054] Among them, N i is the position of the i-th krill individual before induction, N i+1 is the position of the ith krill individual after induction, ε i is the induced motion coefficient of the i-th krill individual, v max is the maximum induced velocity, ω n is the induced motion inertia weight, ω n ∈[0,1];

[0055] is the global optimal krill guidance coefficient, σ1 is the influence coefficient of the global optimal krill on the i-th krill individual, rand is a random number between (0,1), T is the current iteration number, T max is the maximum number of iterations, is the normalized fitness value of the i-th krill individual and the global optimal krill, is the normalized position information of the i-th krill individual and the global optimal krill;

[0056] is the guidance coefficient of other surrounding krill, is the normalized fitness value of the ith krill individual and the surrounding jth krill individual, is the normalized position information of the ith krill individual and the jth krill individual around it, and N2 is the number of surrounding krill;

[0057] S322. The movement of krill individuals will be affected by food, forming foraging movements:

[0058] F i+1 =β i ·v food +ω f ·F i ;

[0059] Among them, F i is the position of the i-th krill individual before foraging, F i+1 is the position of the i-th krill individual after foraging, β i is the foraging motion coefficient of the i-th krill individual, v food is the foraging movement speed, ω f is the inertia weight of foraging motion, ω f ∈[0,1];

[0060] is the food effect coefficient, σ2 is the coefficient of the influence of food on the i-th krill individual, is the normalized fitness value of the i-th krill individual and the food, is the normalized position information of the i-th krill individual and the global optimal krill;

[0061] is the influence coefficient of the global optimal krill historical position,

[0062] S323. Krill individuals will perform random physical diffusion to increase population diversity and form free movement:

[0063]

[0064] Among them, D i+1 is the position of the i-th krill individual after free movement, v' max is the maximum random physical diffusion velocity, δ is the random direction vector;

[0065] S324. Update the position of each krill individual according to the induction, foraging and random physical diffusion of other krill individuals:

[0066]

[0067] Among them, L i (t) is the position of the i-th krill individual at time t, and Δt is the proportional factor of the speed.

[0068] (III) Beneficial effects

[0069] Compared with the prior art, the leakage monitoring method for urban and rural water supply pipe networks based on a large model provided by the present invention has the following beneficial effects:

[0070] 1) Construct a large model for water supply network leakage detection, obtain real-time audio signals collected at each monitoring point of the water supply network, and convert the real-time audio signals into a spectrum heat map and input it into the large model for water supply network leakage detection. The large model for water supply network leakage detection monitors water supply network leakage. By converting the real-time audio signals into a spectrum heat map and using the large model for water supply network leakage detection based on a multi-layer convolutional neural network (CNN) for intelligent detection, it can avoid external noise interference and accurately detect whether the water supply network is leaking, as well as the target water supply pipeline where the leakage point is located;

[0071] 2) Based on the detection results of the large model of water supply network leakage detection, the objective function of water supply network leakage monitoring is determined, and the constraints of water supply network leakage monitoring are determined. According to the objective function and constraints of water supply network leakage monitoring, a water supply network leakage monitoring model is constructed, and the water supply network leakage monitoring model is solved to determine the specific leakage location. By constructing a water supply network leakage monitoring model and using the krill swarm algorithm to solve the objective function, the specific leakage location of the water supply network can be accurately and efficiently determined while reducing the density of acoustic sensor deployment. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0073] Figure 1 It is a schematic diagram of the process of the present invention;

[0074] Figure 2 It is a schematic diagram of the process of monitoring water supply network leakage by using a large water supply network leakage detection model in the present invention;

[0075] Figure 3 The present invention is a flow chart for solving the water supply network leakage monitoring model and determining the specific leakage location. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] A leakage monitoring method for urban and rural water supply networks based on a large model, such as Figure 1 and Figure 2 As shown, a large model for water supply network leakage detection is constructed to obtain real-time audio signals collected at each monitoring point in the water supply network, and the real-time audio signals are converted into spectrum heat maps and input into the large model for water supply network leakage detection. The large model for water supply network leakage detection is used to monitor water supply network leakage.

[0078] After building a large model for water supply network leakage detection, it includes:

[0079] S11, collecting historical audio signals collected from each monitoring point of the water supply network, manually annotating the historical audio signals, and establishing an audio signal sample library;

[0080] S12, converting all historical audio signals in the audio signal sample library into spectrum heat maps, and establishing a spectrum heat map data set;

[0081] S13, using the spectrum heat map data set to train the water supply network leakage detection large model, to obtain a trained water supply network leakage detection large model;

[0082] Among them, the large model for water supply network leakage detection is a large model based on a multi-layer convolutional neural network CNN.

[0083] 1) In S12, all historical audio signals in the audio signal sample library are converted into spectrum heat maps, and a spectrum heat map dataset is established, including:

[0084] S121, pre-emphasize the historical audio signal:

[0085] y(t)=x(t)-α·x(t-1);

[0086] Wherein, x(t) is the historical audio signal at time t, y(t) is the historical audio signal after pre-emphasis at time t, x(t-1) is the historical audio signal at time t-1, α is the pre-emphasis coefficient, α∈(0.9,1);

[0087] S122, performing a short-time Fourier transform (STFT) on the pre-emphasized historical audio signal to obtain a corresponding spectrum graph, and converting the amplitude of the spectrum graph into a logarithmic scale;

[0088] S123, using time as the abscissa and frequency as the ordinate, and using different colors to represent spectrum amplitudes, so as to convert the historical audio signal into a spectrum heat map;

[0089] S124. Repeat S121 to S123 to complete the conversion of all historical audio signals into spectrum heat maps, and establish a spectrum heat map data set.

[0090] 2) In S13, the spectrum heat map data set is used to train the water supply network leakage detection model to obtain a trained water supply network leakage detection model, including:

[0091] S131, dividing the spectrum heat map data set into a training set, a validation set and a test set according to a preset ratio;

[0092] S132, setting the loss function and optimizer of the water supply network leakage detection large model, and inputting the training set into the water supply network leakage detection large model for model training;

[0093] S133, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information;

[0094] S134. If the loss value is less than the preset threshold, the model training ends, and the current water supply network leakage detection large model is the trained water supply network leakage detection large model. Otherwise, return to S132 and continue to use the training set for model training.

[0095] S135, inputting the validation set into the trained large model for water supply network leakage detection, evaluating the generalization ability of the model by observing the performance on the validation set, and tuning the hyperparameters and structure of the model;

[0096] S136. Input the test set into the optimized large model for water supply network leakage detection to evaluate the model performance.

[0097] Specifically, in S132, the training set is input into the water supply network leakage detection large model for model training, including:

[0098] S1321, obtain the detection results of the water supply network leakage detection large model on the training set, draw the ROC curve, calculate the AUC area under the ROC curve, and use the probability value corresponding to the point with the largest AUC area as the probability threshold P th ;

[0099] S1322, after the training set data is input into the water supply network leakage detection model, if the detection result is greater than the probability threshold P th , it indicates that the training set data corresponds to the leakage label, otherwise the training set data corresponds to the normal label.

[0100] The above technical scheme constructs a large model for water supply network leakage detection, obtains real-time audio signals collected at each monitoring point of the water supply network, and converts the real-time audio signals into frequency spectrum heat maps and inputs them into the large model for water supply network leakage detection. The large model for water supply network leakage detection monitors water supply network leakage. By converting real-time audio signals into frequency spectrum heat maps and using a large model for water supply network leakage detection based on a multi-layer convolutional neural network CNN for intelligent detection, it can avoid interference from external noise and accurately detect whether the water supply network is leaking, as well as the target water supply pipeline where the leakage point is located.

[0101] like Figure 1 and Figure 3 As shown, based on the detection results of the large model of water supply network leakage detection, the objective function of water supply network leakage monitoring is determined, and the constraints of water supply network leakage monitoring are determined. According to the objective function and constraints of water supply network leakage monitoring, a water supply network leakage monitoring model is constructed, the water supply network leakage monitoring model is solved, and the specific leakage location is determined.

[0102] 1) Determine the objective function of water supply network leakage monitoring based on the detection results of the large model of water supply network leakage detection, including:

[0103] S21. Constructing a water pressure distribution model of the water supply network under normal working conditions according to the topological relationship and operation parameters of the water supply network;

[0104] S22, based on the abnormal monitoring points detected by the large model for water supply network leakage detection, the water supply pipelines between the abnormal monitoring points are taken as target water supply pipelines;

[0105] S23, setting a plurality of sub-monitoring points in the target water supply pipeline, and collecting the actual water pressure at each sub-monitoring point;

[0106] S24, simulating various water supply network leakage conditions in the water supply network water pressure distribution model, and obtaining simulated water pressures of each sub-monitoring point under each water supply network leakage condition;

[0107] S25. Determine the objective function of water supply network leakage monitoring based on the actual water pressure and simulated water pressure of each sub-monitoring point.

[0108] Specifically, in S25, the objective function of water supply network leakage monitoring is determined according to the actual water pressure and simulated water pressure of each sub-monitoring point, including:

[0109] According to the actual water pressure and simulated water pressure of each sub-monitoring point, the objective function F of water supply network leakage monitoring is determined by the following formula:

[0110]

[0111] Among them, Ps mnis the simulated water pressure of the nth sub-monitoring point under the mth water supply network leakage condition, Pr mn is the actual water pressure of the nth sub-monitoring point under the mth water supply network leakage condition, N1 is the number of sub-monitoring points, and there are M water supply network leakage conditions in total.

[0112] 2) Solve the water supply network leakage monitoring model to determine the specific leakage location, including:

[0113] The krill swarm algorithm (KHA) is used to solve the objective function F of water supply network leakage monitoring, and the specific leakage location that meets the constraint conditions of water supply network leakage monitoring is obtained, including:

[0114] S31, initializing parameters and randomly generating a krill swarm of a certain size, and taking the initial position of each krill individual in the krill swarm as an initial solution;

[0115] S32, updating the position of each krill individual according to the induction, foraging and random physical diffusion of other krill individuals;

[0116] S33, using the objective function F of water supply network leakage monitoring to evaluate each krill individual and calculate the corresponding fitness value;

[0117] S34, record and update the global optimal solution;

[0118] S35. Determine whether the iteration termination condition is met. If the iteration termination condition is not met, return to S32. Otherwise, use the current global optimal solution as the specific leakage position that meets the constraint conditions of water supply network leakage monitoring.

[0119] Specifically, in S32, the position of each krill individual is updated according to the induction, foraging and random physical diffusion of other krill individuals, including:

[0120] S321. The movement of krill individuals will be induced by other krill individuals around them, forming induced movement:

[0121] N i+1 =ε i ·v max +ω n ·N i ;

[0122] Among them, N i is the position of the i-th krill individual before induction, N i+1 is the position of the ith krill individual after induction, ε i is the induced motion coefficient of the i-th krill individual, v max is the maximum induced velocity, ω n is the induced motion inertia weight, ω n∈[0,1];

[0123] is the global optimal krill guidance coefficient, σ1 is the influence coefficient of the global optimal krill on the i-th krill individual, rand is a random number between (0,1), T is the current iteration number, T max is the maximum number of iterations, is the normalized fitness value of the i-th krill individual and the global optimal krill, is the normalized position information of the i-th krill individual and the global optimal krill;

[0124] is the guidance coefficient of other surrounding krill, is the normalized fitness value of the ith krill individual and the surrounding jth krill individual, is the normalized position information of the ith krill individual and the jth krill individual around it, and N2 is the number of surrounding krill;

[0125] S322. The movement of krill individuals will be affected by food, forming foraging movements:

[0126] F i+1 =β i ·v food +ω f ·F i ;

[0127] Among them, F i is the position of the i-th krill individual before foraging, F i+1 is the position of the i-th krill individual after foraging, β i is the foraging motion coefficient of the i-th krill individual, v food is the foraging movement speed, ω f is the inertia weight of foraging motion, ω f ∈[0,1];

[0128] is the food effect coefficient, σ2 is the coefficient of the influence of food on the i-th krill individual, is the normalized fitness value of the i-th krill individual and the food, is the normalized position information of the i-th krill individual and the global optimal krill;

[0129] is the influence coefficient of the global optimal krill historical position,

[0130] S323. Krill individuals will perform random physical diffusion to increase population diversity and form free movement:

[0131]

[0132] Among them, D i+1 is the position of the i-th krill individual after free movement, v' max is the maximum random physical diffusion velocity, δ is the random direction vector;

[0133] S324. Update the position of each krill individual according to the induction, foraging and random physical diffusion of other krill individuals:

[0134]

[0135] Among them, L i (t) is the position of the i-th krill individual at time t, and Δt is the proportional factor of the speed.

[0136] The above technical scheme determines the objective function of water supply network leakage monitoring based on the detection results of the large model of water supply network leakage detection, and determines the constraints of water supply network leakage monitoring, constructs a water supply network leakage monitoring model according to the objective function and constraints of water supply network leakage monitoring, solves the water supply network leakage monitoring model, and determines the specific leakage location. By constructing a water supply network leakage monitoring model and using the krill swarm algorithm to solve the objective function, the specific leakage location of the water supply network can be accurately and efficiently determined while reducing the density of acoustic sensor deployment.

[0137] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring leakage of urban and rural water supply networks based on a large model, characterized in that: Build a large model for water supply network leakage detection, obtain real-time audio signals collected at each monitoring point of the water supply network, and convert the real-time audio signals into spectrum heat maps and input them into the large model for water supply network leakage detection. The large model for water supply network leakage detection will monitor water supply network leakage. Based on the detection results of the large model of water supply network leakage detection, the objective function of water supply network leakage monitoring is determined, and the constraints of water supply network leakage monitoring are determined. According to the objective function and constraints of water supply network leakage monitoring, a water supply network leakage monitoring model is constructed, the water supply network leakage monitoring model is solved, and the specific leakage location is determined.

2. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 1 is characterized in that: After the large model of water supply network leakage detection is constructed, the following steps are included: S11, collecting historical audio signals collected from each monitoring point of the water supply network, manually annotating the historical audio signals, and establishing an audio signal sample library; S12, converting all historical audio signals in the audio signal sample library into spectrum heat maps, and establishing a spectrum heat map data set; S13, using the spectrum heat map data set to train the water supply network leakage detection large model, to obtain a trained water supply network leakage detection large model; Among them, the large model for water supply network leakage detection is a large model based on a multi-layer convolutional neural network CNN.

3. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 2 is characterized in that: In S12, all historical audio signals in the audio signal sample library are converted into spectrum heat maps, and a spectrum heat map dataset is established, including: S121, pre-emphasize the historical audio signal: y(t)=x(t)-α·x(t-1); Wherein, x(t) is the historical audio signal at time t, y(t) is the historical audio signal after pre-emphasis at time t, x(t-1) is the historical audio signal at time t-1, α is the pre-emphasis coefficient, α∈(0.9,1); S122, performing a short-time Fourier transform (STFT) on the pre-emphasized historical audio signal to obtain a corresponding spectrum graph, and converting the amplitude of the spectrum graph into a logarithmic scale; S123, using time as the abscissa and frequency as the ordinate, and using different colors to represent spectrum amplitudes, so as to convert the historical audio signal into a spectrum heat map; S124. Repeat S121 to S123 to complete the conversion of all historical audio signals into spectrum heat maps, and establish a spectrum heat map data set.

4. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 2 is characterized in that: In S13, the spectrum heat map data set is used to train the water supply network leakage detection model to obtain a trained water supply network leakage detection model, including: S131, dividing the spectrum heat map data set into a training set, a validation set and a test set according to a preset ratio; S132, setting the loss function and optimizer of the water supply network leakage detection large model, and inputting the training set into the water supply network leakage detection large model for model training; S133, calculating the loss value based on the loss function, and the optimizer updates the model parameters according to the loss value and the network gradient information; S134. If the loss value is less than the preset threshold, the model training ends, and the current water supply network leakage detection large model is the trained water supply network leakage detection large model. Otherwise, return to S132 and continue to use the training set for model training. S135, inputting the validation set into the trained large model for water supply network leakage detection, evaluating the generalization ability of the model by observing the performance on the validation set, and tuning the hyperparameters and structure of the model; S136. Input the test set into the optimized large model for water supply network leakage detection to evaluate the model performance.

5. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 4 is characterized in that: In S132, the training set is input into the large model for water supply network leakage detection to perform model training, including: S1321, obtain the detection results of the water supply network leakage detection large model on the training set, draw the ROC curve, calculate the AUC area under the ROC curve, and use the probability value corresponding to the point with the largest AUC area as the probability threshold P th ; S1322, after the training set data is input into the water supply network leakage detection model, if the detection result is greater than the probability threshold P th , it indicates that the training set data corresponds to the leakage label, otherwise the training set data corresponds to the normal label.

6. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 1 is characterized in that: The objective function of water supply network leakage monitoring is determined based on the detection results of the large water supply network leakage detection model, including: S21. Constructing a water pressure distribution model of the water supply network under normal working conditions according to the topological relationship and operation parameters of the water supply network; S22, based on the abnormal monitoring points detected by the large model for water supply network leakage detection, the water supply pipelines between the abnormal monitoring points are taken as target water supply pipelines; S23, setting a plurality of sub-monitoring points in the target water supply pipeline, and collecting the actual water pressure at each sub-monitoring point; S24, simulating various water supply network leakage conditions in the water supply network water pressure distribution model, and obtaining simulated water pressures of each sub-monitoring point under each water supply network leakage condition; S25. Determine the objective function of water supply network leakage monitoring based on the actual water pressure and simulated water pressure of each sub-monitoring point.

7. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 6 is characterized in that: In S25, the objective function of water supply network leakage monitoring is determined according to the actual water pressure and simulated water pressure of each sub-monitoring point, including: According to the actual water pressure and simulated water pressure of each sub-monitoring point, the objective function F of water supply network leakage monitoring is determined by the following formula: Among them, Ps mn is the simulated water pressure of the nth sub-monitoring point under the mth water supply network leakage condition, Pr mn is the actual water pressure of the nth sub-monitoring point under the mth water supply network leakage condition, N1 is the number of sub-monitoring points, and there are M water supply network leakage conditions in total.

8. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 7 is characterized in that: Solving the water supply network leakage monitoring model to determine the specific leakage location includes: The krill swarm algorithm (KHA) is used to solve the objective function F of water supply network leakage monitoring and obtain the specific leakage location that meets the constraint conditions of water supply network leakage monitoring.

9. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 8 is characterized in that: The krill swarm algorithm KHA is used to solve the objective function F of the water supply network leakage monitoring to obtain the specific leakage location that meets the constraint conditions of the water supply network leakage monitoring, including: S31, initializing parameters and randomly generating a krill swarm of a certain size, and taking the initial position of each krill individual in the krill swarm as an initial solution; S32, updating the position of each krill individual according to the induction, foraging and random physical diffusion of other krill individuals; S33, using the objective function F of water supply network leakage monitoring to evaluate each krill individual and calculate the corresponding fitness value; S34, record and update the global optimal solution; S35. Determine whether the iteration termination condition is met. If the iteration termination condition is not met, return to S32. Otherwise, use the current global optimal solution as the specific leakage position that meets the constraint conditions of water supply network leakage monitoring.

10. The method for monitoring leakage of urban and rural water supply pipe networks based on a large model according to claim 9, characterized in that: In S32, the position of each krill individual is updated according to the induction, foraging and random physical diffusion of other krill individuals, including: S321. The movement of krill individuals will be induced by other krill individuals around them, forming induced movement: N i+1 =e i ·v max +oh n ·N i ; Among them, N i is the position of the i-th krill individual before induction, N i+1 is the position of the ith krill individual after induction, ε i is the induced motion coefficient of the i-th krill individual, v max is the maximum induced velocity, ω n is the induced motion inertia weight, ω n ∈[0,1]; is the global optimal krill guidance coefficient, σ1 is the influence coefficient of the global optimal krill on the i-th krill individual, rand is a random number between (0,1), T is the current iteration number, T max is the maximum number of iterations, is the normalized fitness value of the i-th krill individual and the global optimal krill, is the normalized position information of the i-th krill individual and the global optimal krill; is the guidance coefficient of other surrounding krill, is the normalized fitness value of the ith krill individual and the surrounding jth krill individual, is the normalized position information of the ith krill individual and the jth krill individual around it, and N2 is the number of surrounding krill; S322. The movement of krill individuals will be affected by food, forming foraging movements: F i+1 =b i ·v food +oh f ·F i ; Among them, F i is the position of the i-th krill individual before foraging, F i+1 is the position of the i-th krill individual after foraging, β i is the foraging motion coefficient of the i-th krill individual, v food is the foraging movement speed, ω f is the inertia weight of foraging motion, ω f ∈[0,1]; is the food effect coefficient, σ2 is the coefficient of the influence of food on the i-th krill individual, is the normalized fitness value of the i-th krill individual and the food, is the normalized position information of the i-th krill individual and the global optimal krill; is the influence coefficient of the global optimal krill historical position, S323. Krill individuals will perform random physical diffusion to increase population diversity and form free movement: Among them, D i+1 is the position of the i-th krill individual after free movement, v' max is the maximum random physical diffusion velocity, δ is the random direction vector; S324. Update the position of each krill individual according to the induction, foraging and random physical diffusion of other krill individuals: Among them, L i (t) is the position of the i-th krill individual at time t, and Δt is the proportional factor of the speed.

Images