Sewage pipe network health condition analysis method for municipal drainage

Abstract

The invention discloses a sewage pipe network health condition analysis method for municipal drainage, and relates to the technical field of sewage pipe network analysis. The invention discloses a sewage pipe network health condition analysis method for municipal drainage. The sewage pipe network health condition analysis method comprises the steps of building sewage pipe network partitions, building an evaluation system and analyzing health conditions of the partitions. According to the invention, by constructing the partition and evaluation system, accurate evaluation and predictive maintenance of the municipal drainage system are realized; performing effective partitioning on a sewage pipe network by utilizing historical sewage discharge data and a partitioning model in combination with a deep clustering network; key pollution discharge indexes are screened through an AHP model, and the health condition of each partition is quantitatively analyzed in combination with a health condition analysis model constructed by a BP neural network; the health condition prediction model is used for predicting the health condition of the partition, and a scientific basis is provided for maintenance and management of a sewage pipe network, so that the operation efficiency and reliability of the system are improved, the fault risk is reduced, and resource allocation is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of sewage pipe network analysis, and in particular to a method for analyzing the health status of a sewage pipe network used for municipal drainage. Background Art

[0002] The health status analysis of the sewage network of municipal drainage is crucial to ensure the effective operation of the urban drainage system. Through this analysis, it can help identify structural and functional defects in the network, evaluate their impact on the environment and public safety, and provide decision support for the maintenance, repair and upgrading of the network; however, urban road conditions are complex, and there are significant differences in the sewage discharge characteristics of different areas. For example, there may be chemical emissions in industrial areas, while residential areas are mainly domestic sewage. Therefore, when evaluating the health status of the sewage network, it is necessary to formulate corresponding evaluation standards based on the characteristics of different areas; in addition, it is also crucial to identify and quantify non-sewage sources, such as rainwater, groundwater, etc.; this will help to comprehensively and comprehensively analyze the health status of the sewage network, so as to more effectively formulate maintenance and improvement measures. Summary of the invention

[0003] The present invention aims to provide a method for analyzing the health status of a sewage pipe network for municipal drainage, and to conduct a comprehensive and integrated analysis of the health status of the sewage pipe network.

[0004] A method for analyzing the health status of a sewage pipe network for municipal drainage comprises the following steps: Step S1. Constructing sewage pipe network partitions; Obtain historical sewage discharge data in the target area; the historical sewage discharge data contains the sewage pipe network number B of all sewage pipe networks in the target area n And the corresponding sewage network characteristic information X n , n=1, 2, ..., N; N is the total number of all sewage pipe networks in the target area; according to the historical sewage data and the sewage pipe network partition model, the target area partition result is obtained; the target area partition result includes I target area partitions F i , i=1, 2, …, I; The sewage pipe network partition model includes a data preprocessing layer, a regional feature extraction layer, a regional division layer and a result output layer, and is constructed based on the improved deep clustering network; Step S2. Constructing an evaluation system; For the target area partition F i , get the target area partition F i The pollution index factor dataset L i , pollution index factor data set L i It is obtained by screening by AHP model; according to the pollution index factor data set L iThe health status analysis result R of the target area is obtained by analyzing the health status analysis model of the sewage network. i The sewage network health status analysis model is constructed based on the improved BP neural network, including parameter calculation layer, health evaluation layer and analysis result output layer, which is used to partition the target area. i The pollution index factor dataset L i Conduct precise analysis; Step S3. Partition health status analysis; Based on the analysis results of the health status of the target area R i The health status prediction model of the sewage pipe network is used for prediction analysis to obtain the health status prediction result Y of the target area. i The sewage pipe network health status prediction model includes a data preprocessing layer, a health status feature recognition layer and a health status analysis result output layer; according to the health status analysis results R i and the predicted health status of the target area Y i Partition the target area F i Predictive maintenance of the sewage network within the city.

[0005] As a preferred technical solution of the present invention, the sewage pipe network zoning model includes a data preprocessing layer, a regional feature extraction layer, a regional division layer and a result output layer; The data preprocessing layer is used to convert the sewage pipe network characteristic information X in the historical sewage discharge data n Perform pretreatment to obtain the pretreated sewage pipe network characteristic information X n '; The sewage pipe network number B in the historical sewage data n The topological distance matrix is ​​constructed using the Floyd algorithm to obtain the sewage pipe network matrix of the target area; the sewage pipe network matrix of the target area contains the sewage pipe network node B of the target area n '; The regional feature extraction layer is used to extract the pre-treated sewage pipe network feature information X n 'Perform feature extraction to obtain the feature information of the pre-treated sewage pipe network. n ; The regional division layer is used to use the improved deep clustering network to classify the target area sewage pipe network matrix and the pre-treated sewage pipe network feature information feature T n Clustering is performed to obtain the target area partition result; the target area partition result includes I target area partitions F i ; The result output layer is used to output the target area partitioning results.

[0006] As a preferred technical solution of the present invention, the specific steps of clustering in the region division layer include: The region division layer consists of an autoencoder and a graph convolutional network, and preset parameters of the autoencoder and the graph convolutional network are set; the maximum number of iterations is set to P, the current iteration number is denoted as p, and p = 1, 2,..., P; When the current iteration number p < P: Use the autoencoder to extract the feature information feature T of the preprocessed sewage pipe network n of the data representation, and obtain the first embedding representation S of the sewage pipe network ae ; use the graph convolutional network to extract the data representation of the sewage pipe network matrix in the target area, and obtain the second embedding representation S of the sewage pipe network fg and the third embedding representation S of the sewage pipe network tg ; Perform data fusion on the first embedding representation S of the sewage pipe network ae , the second embedding representation S of the sewage pipe network fg and the third embedding representation S of the sewage pipe network tg to obtain the fused sewage pipe network sample distribution H r ; Calculate the sample distribution of the first embedding representation S of the sewage pipe network ae to obtain the first sample distribution V1 and the second sample distribution V2 of the sewage pipe network; Minimize the KL divergence of the fused sewage pipe network sample distribution H r , the first sample distribution V1 and the second sample distribution V2 of the sewage pipe network to obtain the KL divergence D1 of the sewage pipe network, the KL divergence D2 of the sewage pipe network and the KL divergence D3 of the sewage pipe network; Use the preset sewage pipe network loss function to calculate the KL divergence D1 of the sewage pipe network, the KL divergence D2 of the sewage pipe network and the KL divergence D3 of the sewage pipe network, and update the preset parameters of the autoencoder and the graph convolutional network; When the maximum number of iterations P is reached, use the clustering algorithm to cluster the latest fused sewage pipe network sample distribution H r to obtain the target area partition result, and the target area partition result contains I target area partitions F i .

[0007] As a preferred technical solution of the present invention, the specific steps for obtaining the sewage discharge index factor data set L i of the target area partition F i include: Obtain an initial pollution discharge index factor index system; the initial pollution discharge index factor index system includes several pollution discharge index factors; introduce the sewage pipe network decision data set into the AHP model to obtain the sewage pipe network index AHP model; input the initial pollution discharge index factor index system into the sewage pipe network index AHP model for comparison to obtain the sewage pipe network index judgment matrix; perform consistency test on the sewage pipe network index judgment matrix to obtain the sewage pipe network index weight coefficient set; The sewage network index weight coefficient is concentrated to filter out the pollution index factors corresponding to the pollution index selection weight, and the pollution index factors to be eliminated are obtained; the initial pollution index factor index system is updated according to the pollution index factors to be eliminated, and the updated pollution index factor index system is obtained; the target area partition F is obtained according to the updated pollution index factor index system i The pollution index factor dataset L i .

[0008] As a preferred technical solution of the present invention, the sewage pipe network health status analysis model includes a parameter calculation layer, a health evaluation layer and an analysis result output layer; The parameter calculation layer is composed of a three-layer BP neural network model, including an input layer, a hidden layer, and an output layer, which is used to calculate the pollution index factor data set L i Conduct prediction analysis to obtain a set of sewage network defect parameter indicators; The health assessment layer is used to perform membership analysis based on the sewage network defect parameter indicator set to obtain the sewage network defect parameter health membership set; The analysis result output layer is used to calculate the health membership set of sewage network defect parameters using the sewage network comprehensive evaluation matrix to obtain the target area health status analysis result R i .

[0009] As a preferred technical solution of the present invention, the specific steps of training the parameter calculation layer include: Set the range of hidden layer neurons; construct M hidden layer neurons individuals G m , m=1,2,…,M; the number of neurons in the hidden layer is G m According to the range of the number of hidden layer neurons, set the number of hidden layer neurons in M ​​individual G m Combine to obtain the iterative population of the number of neurons in the hidden layer; set the maximum number of parameter iterations to Q, and the current number of parameter iterations is represented by q, q=1, 2, …, Q; Construct parameter calculation training set; use parameter calculation training set and hidden layer neuron number individual G m Construct the simulation model and obtain the simulation parameter calculation layer C m ; Calculate the simulation parameters layer C mThe prediction accuracy function is used for comprehensive evaluation to obtain the prediction result A of the simulation parameter calculation layer. m ; The simulation parameter calculation layer prediction result A m As the number of neurons in the hidden layer, G m The fitness U m ; Introducing the inertia factor W, ; When performing population iteration, the inertia factor W is used to iterate the population of the number of neurons in the hidden layer; When the maximum number of parameter iterations is reached, the output fitness is the maximum corresponding hidden layer neuron number individual G m , which is the optimal number of hidden layer neurons; and the simulation parameter calculation layer C corresponding to the optimal number of hidden layer neurons m As the parameter calculation layer in the sewage network health status analysis model.

[0010] As a preferred technical solution of the present invention, the sewage pipe network health status prediction model includes a data preprocessing layer, a health status feature recognition layer and a health status analysis result output layer; The data preprocessing layer is used to analyze the health status of the target area. i Perform preprocessing to obtain the health status analysis result R of the preprocessing target area i '; The health status feature recognition layer is used to analyze the health status results R of the preprocessed target area. i 'Perform forecast analysis and obtain the target area health status forecast result Y i ; The health status analysis result output layer is used to output the health status prediction result Y of the target area. i .

[0011] As a preferred technical solution of the present invention, the specific steps of training the health status feature recognition layer include: Collecting several groups of health status feature recognition training samples, wherein the input value of each group of health status feature recognition training samples is the current health status analysis result, and the output value is the future health status analysis result of a fixed time step; combining several groups of health status feature recognition training samples to obtain a health status feature recognition training set; The health status feature recognition training set is input into the sewage network health status prediction model to train the health status feature recognition layer with the output value as the target, so as to obtain the initial health status feature recognition layer; the initial health status feature recognition layer is model evaluated to obtain the initial health status feature recognition layer model evaluation result; if the initial health status feature recognition layer model evaluation result is passed, the initial health status feature recognition layer is used as the health status feature recognition layer in the sewage network health status prediction model; otherwise, the model training is continued using the health status feature recognition training set.

[0012] The present invention has the following advantages: 1. The present invention realizes accurate evaluation and predictive maintenance of municipal drainage systems by constructing a zoning and evaluation system; effectively zoning the sewage pipe network by using historical sewage data and zoning models combined with a deep clustering network; quantitatively analyzing the health status of each zone by screening key sewage indicators through the AHP model and combining the health status analysis model constructed by the BP neural network; predicting the health status of the zones using the health status prediction model, providing a scientific basis for the maintenance and management of the sewage pipe network, thereby improving the operating efficiency and reliability of the system, reducing the risk of failures, and optimizing resource allocation.

[0013] 2. The present invention realizes efficient zoning and health status assessment of the sewage pipe network system by integrating multi-stage processing of data preprocessing, feature extraction, deep clustering and result output; through the autoencoder and graph convolutional network, the model can extract and fuse the key features of the sewage pipe network and generate embedded representation, which enhances the understanding of the structure and characteristics of the pipe network; KL divergence is used for optimization to ensure the consistency of feature distribution and the effectiveness of clustering; the iterative update mechanism further improves the accuracy and stability of the model, and can provide municipal drainage management personnel with accurate zoning results and health status predictions, guide predictive maintenance, improve pipe network management efficiency, reduce operational risks, and provide decision support for optimal resource allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flowchart of a method for analyzing the health status of a sewage pipe network for municipal drainage, which is adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0016] A method for analyzing the health status of sewage pipe networks for municipal drainage, see Figure 1 As shown, the following steps are included: Step S1. Constructing sewage pipe network partitions; Obtain historical sewage discharge data in the target area; the historical sewage discharge data contains the sewage pipe network number B of all sewage pipe networks in the target area n And the corresponding sewage network characteristic information X n , n=1, 2, ..., N; N is the total number of all sewage pipe networks in the target area; according to the historical sewage data and the sewage pipe network partition model, the target area partition result is obtained; the target area partition result includes I target area partitions F i , i=1, 2, …, I; The sewage pipe network partition model includes a data preprocessing layer, a regional feature extraction layer, a regional division layer and a result output layer, and is constructed based on the improved deep clustering network; The characteristic information of the sewage network includes information on the material of the network, rainfall information near the network, and information on businesses near the network. Different materials of the network require different methods and materials for maintenance and replacement. Rainfall has a direct impact on the flow and pressure of the sewage network. During the rainy season, due to increased rainfall and surface runoff, the load on the sewage network may increase, leading to overflow or waterlogging. The type and number of businesses may affect the chemical composition and pollution level of sewage. For example, the catering industry may discharge more grease and solid waste, while industrial areas may contain more toxic chemicals. This information helps to assess the ability of the sewage network to handle specific types of wastewater and determine whether special treatment measures are required. The sewage pipe network zoning model includes a data preprocessing layer, a regional feature extraction layer, a regional division layer and a result output layer; The data preprocessing layer is used to convert the sewage pipe network characteristic information X in the historical sewage discharge data n Perform pretreatment to obtain the pretreated sewage pipe network characteristic information X n '; The sewage pipe network number B in the historical sewage data n The topological distance matrix is ​​constructed using the Floyd algorithm to obtain the sewage pipe network matrix of the target area; the sewage pipe network matrix of the target area contains the sewage pipe network node B of the target area n '; The regional feature extraction layer is used to extract the pre-treated sewage pipe network feature information X n 'Perform feature extraction to obtain the feature information of the pre-treated sewage pipe network. n ; The regional division layer is used to use the improved deep clustering network to classify the target area sewage pipe network matrix and the pre-treated sewage pipe network feature information feature T n Clustering is performed to obtain the target area partition result; the target area partition result includes I target area partitions F i ; The result output layer is used to output the target area partition results; The specific steps for clustering in the region division layer include: The region division layer is composed of an autoencoder and a graph convolutional network, and the preset parameters of the autoencoder and the graph convolutional network are set; the maximum number of iterations is set to P, the current number of iterations is denoted as p, and p = 1, 2,..., P; When the current number of iterations p < P: Use the autoencoder to extract the feature information feature T of the preprocessed sewage pipe network n of the data representation, and obtain the first embedding representation S of the sewage pipe network ae ; Use the graph convolutional network to extract the data representation of the sewage pipe network matrix in the target area, and obtain the second embedding representation S of the sewage pipe network fg and the third embedding representation S of the sewage pipe network tg ; Perform data fusion on the first embedding representation S of the sewage pipe network ae , the second embedding representation S of the sewage pipe network fg and the third embedding representation S of the sewage pipe network tg to obtain the fused sewage pipe network sample distribution H r ; Calculate the sample distribution of the first embedding representation S of the sewage pipe network ae to obtain the first sample distribution V1 and the second sample distribution V2 of the sewage pipe network; Minimize the KL divergence of the fused sewage pipe network sample distribution H r , the first sample distribution V1 and the second sample distribution V2 of the sewage pipe network to obtain the KL divergence D1 of the sewage pipe network, the KL divergence D2 of the sewage pipe network, and the KL divergence D3 of the sewage pipe network; Use the preset sewage pipe network loss function to calculate the KL divergence D1 of the sewage pipe network, the KL divergence D2 of the sewage pipe network, and the KL divergence D3 of the sewage pipe network, and update the preset parameters of the autoencoder and the graph convolutional network; When the maximum number of iterations P is reached, use the clustering algorithm to cluster the latest fused sewage pipe network sample distribution H r to obtain the target area partition result, and the target area partition result contains I target area partitions F i ; ​​​​​​​​By subdividing the sewage pipe network into smaller areas, each area can be managed more finely; the model can extract key features from historical sewage data, providing rich information for subsequent analysis; the topological distance matrix constructed by the Floyd algorithm helps to understand the connection relationship between pipe network nodes and provide a basis for regional division; the improved deep clustering network improves the accuracy of clustering, making the partition results more consistent with the characteristics of the actual pipe network; by integrating multiple sewage pipe network embedding representations, the model's understanding and expression of pipe network characteristics are enhanced; by minimizing KL divergence and iterative optimization, the model can continuously adjust parameters to obtain better clustering effects; Step S2. Constructing an evaluation system; For the target area partition F i , get the target area partition F i The pollution index factor dataset L i , pollution index factor data set L i It is obtained by screening by AHP model; according to the pollution index factor data set L i The health status analysis result R of the target area is obtained by analyzing the health status analysis model of the sewage network. i The sewage pipe network health status analysis model is constructed based on the improved BP neural network, including parameter calculation layer, health evaluation layer and analysis result output layer, which is used to partition the target area. i The pollution index factor dataset L i Conduct precise analysis; Get the target area partition F i The pollution index factor dataset L i The specific steps include: Obtain an initial pollution discharge index factor index system; the initial pollution discharge index factor index system includes several pollution discharge index factors; introduce the sewage pipe network decision data set into the AHP model to obtain the sewage pipe network index AHP model; input the initial pollution discharge index factor index system into the sewage pipe network index AHP model for comparison to obtain the sewage pipe network index judgment matrix; perform consistency test on the sewage pipe network index judgment matrix to obtain the sewage pipe network index weight coefficient set; The sewage network index weight coefficient is concentrated to filter out the pollution index factors corresponding to the pollution index selection weight, and the pollution index factors to be eliminated are obtained; the initial pollution index factor index system is updated according to the pollution index factors to be eliminated, and the updated pollution index factor index system is obtained; the target area partition F is obtained according to the updated pollution index factor index system i The pollution index factor dataset L i ; The initial pollution discharge index factor index system includes a variety of evaluation indicators. The key pollution discharge index factors are screened by the AHP model to ensure the scientificity and rationality of the evaluation indicators and provide a scientific basis for sewage pipe network management. i The acquisition of the target area partition F i The management of sewage pipe network is more refined, which helps to identify and solve specific problems; the improved BP neural network model can process data more efficiently, quickly obtain analysis results, and improve decision-making efficiency; the evaluation system covers many aspects of the sewage pipe network, including service performance, construction status, maintenance management, etc., and realizes a comprehensive assessment of the health status of the sewage pipe network; The sewage network health status analysis model includes parameter calculation layer, health assessment layer and analysis result output layer; The parameter calculation layer is composed of a three-layer BP neural network model, including an input layer, a hidden layer, and an output layer, which is used to calculate the pollution index factor data set L i Conduct prediction analysis to obtain a set of sewage network defect parameter indicators; The health assessment layer is used to perform membership analysis based on the sewage network defect parameter indicator set to obtain the sewage network defect parameter health membership set; The analysis result output layer is used to calculate the health membership set of sewage network defect parameters using the sewage network comprehensive evaluation matrix to obtain the target area health status analysis result R i ; The specific steps of training parameter calculation layer include: Set the range of hidden layer neurons; construct M hidden layer neurons individuals G m , m=1,2,…,M; the number of neurons in the hidden layer is G m According to the range of the number of hidden layer neurons, set the number of hidden layer neurons in M ​​individual G m Combine to obtain the iterative population of the number of neurons in the hidden layer; set the maximum number of parameter iterations to Q, and the current number of parameter iterations is represented by q, q=1, 2, …, Q; Construct parameter calculation training set; use parameter calculation training set and hidden layer neuron number individual G m Construct the simulation model and obtain the simulation parameter calculation layer C m ; Calculate the simulation parameters layer C m The prediction accuracy function is used for comprehensive evaluation to obtain the prediction result A of the simulation parameter calculation layer. m ; The simulation parameter calculation layer prediction result A m As the number of neurons in the hidden layer, G m The fitness U m ; Introducing the inertia factor W, ; When performing population iteration, the inertia factor W is used to iterate the population of the number of neurons in the hidden layer; When the maximum number of parameter iterations is reached, the output fitness is the maximum corresponding hidden layer neuron number individual G m , which is the optimal number of hidden layer neurons; and the simulation parameter calculation layer C corresponding to the optimal number of hidden layer neurons m As the parameter calculation layer in the sewage network health analysis model; Through the three-layer BP neural network model, the pollution index factor data set L i The accurate prediction analysis of the sewage pipe network defect parameter index set is improved; the optimal network structure is found by iteratively optimizing the number of neurons in the hidden layer, thereby improving the prediction ability and efficiency of the model; the inertia factor W is introduced, so that the model can adaptively adjust the learning pace during the training process, enhance the generalization ability of the model, and automatically adjust the iteration range; the membership analysis of the health assessment layer provides a quantitative health status evaluation for the sewage pipe network defect parameters, making the evaluation results more intuitive and reliable; the sewage pipe network comprehensive evaluation matrix is ​​used to comprehensively consider multiple factors to obtain a more comprehensive and objective health status analysis result; the health status analysis result R provided by the analysis result output layer is i , helps to rationally allocate maintenance resources and prioritize key problem areas; Step S3. Partition health status analysis; Based on the analysis results of the health status of the target area R i The health status prediction model of the sewage pipe network is used for prediction analysis to obtain the health status prediction result Y of the target area. i The sewage pipe network health status prediction model includes a data preprocessing layer, a health status feature recognition layer and a health status analysis result output layer; according to the health status analysis results R i and the predicted health status of the target area Y i Partition the target area F i Predictive maintenance of the sewage network within: The sewage pipe network health status prediction model includes a data preprocessing layer, a health status feature recognition layer and a health status analysis result output layer; The data preprocessing layer is used to analyze the health status of the target area. i Perform preprocessing to obtain the health status analysis result R of the preprocessing target area i '; The health status feature recognition layer is used to analyze the health status results R of the preprocessed target area. i 'Perform forecast analysis and obtain the target area health status forecast result Y i ; The health status analysis result output layer is used to output the health status prediction result Y of the target area. i ; The specific steps for training the health status feature recognition layer include: Collecting several groups of health status feature recognition training samples, wherein the input value of each group of health status feature recognition training samples is the current health status analysis result, and the output value is the future health status analysis result of a fixed time step; combining several groups of health status feature recognition training samples to obtain a health status feature recognition training set; The health status feature recognition training set is input into the sewage pipe network health status prediction model to train the health status feature recognition layer with the output value as the target, and obtain the initial health status feature recognition layer; the initial health status feature recognition layer is model evaluated to obtain the initial health status feature recognition layer model evaluation result; if the initial health status feature recognition layer model evaluation result is passed, the initial health status feature recognition layer is used as the health status feature recognition layer in the sewage pipe network health status prediction model; otherwise, the model training is continued using the health status feature recognition training set; Through predictive analysis, predictive maintenance of the sewage network can be achieved, so as to prevent problems before they occur and reduce the need for emergency repairs; the fixed time step is set by professional technicians according to actual conditions. If it is necessary to predict the health of the sewage network for a longer period of time, the fixed time step can be adjusted, and re-training and prediction can be carried out; historical data and predictive models are used to provide data support for maintenance decisions to improve the scientific nature and accuracy of decisions; through automated predictive models, the time and cost of manual analysis can be reduced, and the efficiency of maintenance work can be improved; through timely predictive maintenance, the service life of the sewage network can be extended and the overall replacement cost can be reduced.

[0017] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for analyzing the health status of a sewage pipe network for municipal drainage, characterized in that: It includes the following steps: Step S1. Construct a sewage pipe network partition; Obtain historical sewage discharge data in the target area; the historical sewage discharge data contains the sewage pipe network number B of all sewage pipe networks in the target area n And the corresponding sewage network characteristic information X n , n=1, 2, …, N; N is the total number of all sewage pipe networks in the target area; Analyze according to historical sewage discharge data and the sewage pipe network partition model to obtain the partition result of the target area; The target area partition result includes I target area partitions F i , i=1,2,…,I; The sewage pipe network partition model includes a data preprocessing layer, a regional feature extraction layer, a regional division layer, and a result output layer, and is constructed based on an improved deep clustering network; Step S2. Construct an evaluation system; For the target area partition F i , get the target area partition F i The pollution index factor dataset L i , pollution index factor data set L i It is obtained by screening through AHP model; According to the pollution index factor data set L i The health status analysis result R of the target area is obtained by analyzing the health status analysis model of the sewage network. i ; The sewage network health status analysis model is constructed based on the improved BP neural network, including parameter calculation layer, health evaluation layer and analysis result output layer, which is used to partition the target area. i The pollution index factor dataset L i Conduct precise analysis; Step S3. Analyze the health status of the partition; Based on the analysis results of the health status of the target area R i The health status prediction model of the sewage pipe network is used for prediction analysis to obtain the health status prediction result Y of the target area. i ; The sewage pipe network health status prediction model includes a data preprocessing layer, a health status feature recognition layer, and a health status analysis result output layer; According to the health status analysis results of the target area R i and the predicted health status of the target area Y i Partition the target area F i Predictive maintenance of the sewage network within the city.

2. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 1, characterized in that: The sewage pipe network partition model includes a data preprocessing layer, a regional feature extraction layer, a regional division layer, and a result output layer; The data preprocessing layer is used to convert the sewage pipe network characteristic information X in the historical sewage discharge data n Perform pretreatment to obtain the pretreated sewage pipe network characteristic information X n '; The sewage pipe network number B in the historical sewage data n The topological distance matrix is ​​constructed using the Floyd algorithm to obtain the sewage pipe network matrix of the target area; the sewage pipe network matrix of the target area contains the sewage pipe network node B of the target area n '; The regional feature extraction layer is used to extract the pre-treated sewage pipe network feature information X n 'Perform feature extraction to obtain the feature information of the pre-treated sewage pipe network. n ; The regional division layer is used to use the improved deep clustering network to classify the target area sewage pipe network matrix and the pre-treated sewage pipe network feature information feature T n Perform clustering to obtain the target area partition results; The target area partition result includes I target area partitions F i ; The result output layer is used to output the partition result of the target area.

3. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 2, characterized in that: The specific steps for clustering in the regional division layer include: The regional division layer is composed of an autoencoder and a graph convolutional network, and set the preset parameters of the autoencoder and the graph convolutional network; set the maximum number of iterations as P, and the current number of iterations is represented as p, where p = 1, 2,..., P; When the current number of iterations p < P: Extracting characteristic information of pre-treated sewage pipe network using autoencoder n The data representation is used to obtain the first embedding representation S of the sewage network. ae ; Use graph convolutional networks to extract the data representation of the sewage pipe network matrix in the target area and obtain the second embedding representation S of the sewage pipe network fg and the third embedding representation of the sewage network S tg ; The first embedding representation S for the sewage network ae , the second embedded representation of the sewage network S fg and the third embedding representation of the sewage network S tg Perform data fusion to obtain the fused sewage network sample distribution H r ; Calculate the first embedding representation S of the sewage network ae The sample distribution of the sewage pipe network is obtained by obtaining the first sample distribution V1 of the sewage pipe network and the second sample distribution V2 of the sewage pipe network; Minimize the fusion of sewage network sample distribution H r , the KL divergence of the first sample distribution V1 of the sewage pipe network and the second sample distribution V2 of the sewage pipe network, to obtain the KL divergence D1 of the sewage pipe network, the KL divergence D2 of the sewage pipe network and the KL divergence D3 of the sewage pipe network; Calculate the KL divergence D1 of the sewage pipe network, the KL divergence D2 of the sewage pipe network, and the KL divergence D3 of the sewage pipe network using the preset sewage pipe network loss function, and update the preset parameters of the autoencoder and the graph convolutional network; When the maximum number of iterations P is reached, the clustering algorithm is used to distribute the latest fused sewage network samples H r Clustering is performed to obtain the target area partition result, which contains I target area partitions F i .

4. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 3, characterized in that: Get the target area partition F i The pollution index factor dataset L i The specific steps include: Obtain the initial sewage discharge index factor index system; the initial sewage discharge index factor index system contains several sewage discharge index factors; introduce the sewage pipe network decision data set into the AHP model to obtain the sewage pipe network index AHP model; input the initial sewage discharge index factor index system into the sewage pipe network index AHP model for comparison to obtain the sewage pipe network index judgment matrix; conduct a consistency test on the sewage pipe network index judgment matrix to obtain the sewage pipe network index weight coefficient set; The sewage network index weight coefficient is concentrated to filter out the pollution index factors corresponding to the pollution index selection weight, and the pollution index factors to be eliminated are obtained; the initial pollution index factor index system is updated according to the pollution index factors to be eliminated, and the updated pollution index factor index system is obtained; the target area partition F is obtained according to the updated pollution index factor index system i The pollution index factor dataset L i .

5. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 4, characterized in that: The sewage pipe network health status analysis model includes a parameter calculation layer, a health degree evaluation layer, and an analysis result output layer; The parameter calculation layer is composed of a three-layer BP neural network model, including an input layer, a hidden layer, and an output layer, which is used to calculate the pollution index factor data set L i Conduct prediction analysis to obtain a set of sewage network defect parameter indicators; The health degree evaluation layer is used to conduct membership analysis according to the sewage pipe network defect parameter index set to obtain the sewage pipe network defect parameter health membership set; The analysis result output layer is used to calculate the health membership set of sewage network defect parameters using the sewage network comprehensive evaluation matrix to obtain the target area health status analysis result R i .

6. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 5, characterized in that: The specific steps for training the parameter calculation layer include: Set the range of hidden layer neurons; construct M hidden layer neurons individuals G m , m=1,2,…,M; the number of neurons in the hidden layer is G m According to the range of the number of hidden layer neurons, set the number of hidden layer neurons in M ​​individual G m Combine to obtain the iterative population of the number of neurons in the hidden layer; set the maximum number of parameter iterations to Q, and the current number of parameter iterations is represented by q, q=1, 2, …, Q; Construct parameter calculation training set; use parameter calculation training set and hidden layer neuron number individual G m Construct the simulation model and obtain the simulation parameter calculation layer C m ; Calculate the simulation parameters layer C m The prediction accuracy function is used for comprehensive evaluation to obtain the prediction result A of the simulation parameter calculation layer. m ; The simulation parameter calculation layer prediction result A m As the number of neurons in the hidden layer, G m The fitness U m ; Introducing the inertia factor W, ; When performing population iteration, the inertia factor W is used to iterate the population of the number of neurons in the hidden layer; When the maximum number of parameter iterations is reached, the output fitness is the maximum corresponding hidden layer neuron number individual G m , which is the optimal number of hidden layer neurons; and the simulation parameter calculation layer C corresponding to the optimal number of hidden layer neurons m As the parameter calculation layer in the sewage network health status analysis model.

7. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 6, characterized in that: The sewage pipe network health status prediction model includes a data preprocessing layer, a health status feature recognition layer, and a health status analysis result output layer; The data preprocessing layer is used to analyze the health status of the target area. i Perform preprocessing to obtain the health status analysis result R of the preprocessing target area i '; Health The condition feature recognition layer is used to analyze the health status of the preprocessed target area. i 'Perform forecast analysis and obtain the target area health status forecast result Y i ; The health status analysis result output layer is used to output the health status prediction result Y of the target area. i .

8. A method for analyzing the health status of a sewage pipe network for municipal drainage according to claim 7, characterized in that: The specific steps for training the health status feature recognition layer include: Collect several groups of health status feature recognition training samples. In each group of health status feature recognition training samples, the input value is the current health status analysis result, and the output value is the future health status analysis result at a fixed time step; combine several groups of health status feature recognition training samples to obtain the health status feature recognition training set; The health status feature recognition training set is input into the sewage network health status prediction model to train the health status feature recognition layer with the output value as the target, so as to obtain the initial health status feature recognition layer; the initial health status feature recognition layer is model evaluated to obtain the initial health status feature recognition layer model evaluation result; if the initial health status feature recognition layer model evaluation result is passed, the initial health status feature recognition layer is used as the health status feature recognition layer in the sewage network health status prediction model; otherwise, the model training is continued using the health status feature recognition training set.

Images