An intelligent decision-making method for underground drainage network maintenance based on MOP-DL

Abstract

The present invention discloses a method for intelligent decision-making for underground drainage network maintenance based on MOP-DL. The method comprises the following steps: obtaining functional defects of the drainage network and regional flood disaster losses; constructing a multi-objective planning model for underground drainage network maintenance decisions using a multi-objective module as a constraint; solving the multi-objective planning model using a multi-objective group search algorithm module based on covariance evolution and chaos search to obtain a sample set of drainage network functional defects with maintenance decision labels; inputting the sample set of drainage network functional defects with maintenance decision labels into a deep learning module; and repeatedly iteratively training the deep learning module using the sample set of drainage network functional defects with maintenance decision labels to calibrate the parameters of the loss function, thereby outputting a maintenance decision result. The method accurately fits the regional disaster losses of urban waterlogging under the influence of functional defects, accurately quantifies the economic benefits of maintenance decisions, and provides highly reliable drainage network maintenance decision results.

Description

Technical Field

[0001] The present invention relates to the field of underground drainage pipe network maintenance, and in particular to an intelligent decision-making method for underground drainage pipe network maintenance based on MOP-DL. Background Art

[0002] In the study of urban drainage network maintenance decisions, some scholars have assessed network vulnerability risks and provided maintenance recommendations from the perspectives of structural degradation and hydraulic degradation, but have ignored the economic benefits of maintenance decisions and have not considered the hydrological and hydrodynamic coupling relationship between functional defects in drainage networks and urban waterlogging losses. Existing decision-making optimization mechanism designs focus on improving network defect maintenance technology and generalized pattern search algorithms under the constraint of minimum maintenance costs. These studies do not consider the inherent relationship between urban drainage network maintenance strategy costs, waterlogging losses, and constraints, nor the issue of intelligent decision-making. In summary, the existing urban drainage network maintenance decision-making system has the following shortcomings:

[0003] (1) The hydrological and hydrodynamic coupling relationship between functional damage of the drainage network and urban waterlogging losses was not considered;

[0004] (2) Ignoring the economic benefits of maintenance decisions;

[0005] (3) The intrinsic relationship between the cost of urban drainage network maintenance strategies, waterlogging losses and constraints, as well as the issue of intelligent decision-making, are not considered.

[0006] In order to solve the above problems, the inventors of the present invention proposed an intelligent decision-making method for underground drainage network maintenance based on MOP-DL. Summary of the Invention

[0007] In order to solve the above problems, the purpose of the present invention is to provide an intelligent decision-making method for underground drainage network maintenance based on MOP-DL, which can accurately fit the disaster losses in urban waterlogging areas under the influence of functional diseases, accurately quantify the economic benefits of maintenance decisions, and the drainage network maintenance decision results are highly intelligent and reliable.

[0008] Based on this, the present invention provides an intelligent decision-making method for underground drainage network maintenance based on MOP-DL, the method comprising:

[0009] Obtain information on functional damage to the drainage network and losses caused by regional flood disasters;

[0010] A multi-objective planning model for underground drainage network maintenance decision-making is constructed using multi-objective modules as constraints;

[0011] Based on the functional damage status of the drainage network and the regional flood disaster losses, a multi-objective group search algorithm module based on covariance evolution and chaos search is used to solve the multi-objective planning model to obtain a sample set of functional damage status of the drainage network with maintenance decision labels;

[0012] Inputting the functional disease condition sample set of the drainage network with maintenance decision labels into the deep learning module;

[0013] The deep learning module uses a sample set of functional disease conditions of the drainage network with maintenance decision labels to calibrate the parameters of the loss function to perform iterative training on itself and output the maintenance decision results.

[0014] The acquisition of drainage network functional disease conditions and regional flood disaster losses includes:

[0015] Input functional disease data into the urban pipe network disease initial sample input module;

[0016] Based on the hydrological and hydrodynamic coupling relationship model between functional diseases of drainage pipe networks and urban waterlogging and the PCSWMM module, an urban waterlogging early warning module was constructed that considers the coupling effects of functional diseases of pipe networks and urban waterlogging.

[0017] Input the functional disease data in the urban pipe network disease initial sample input module into the urban waterlogging early warning module to obtain the functional disease status of the drainage network and the corresponding flood disaster status;

[0018] The flood disaster situation is input into the urban waterlogging loss calculation model to obtain the corresponding regional flood disaster loss situation.

[0019] Among them, the multi-objective module as the constraint condition includes: the constraint conditions include: maximum economic benefit, minimum maintenance cost and minimum damage caused by urban flooding.

[0020] The multi-objective planning model for building underground drainage network maintenance decisions specifically includes:

[0021] Objective functions for maximizing regional economic benefits, minimizing regional underground drainage network maintenance costs, and minimizing urban waterlogging losses;

[0022] The quantification formulas are as follows:

[0023]

[0024]

[0025]

[0026] in, represents the economic benefits of area i where waterlogging did not occur, is the economic benefit of the flooded area j, C k is the cost of taking pipeline maintenance measures in area k, P i is the property density of region i, S i is the flooded area of ​​region i, s is the flooded water level, m is the number of regions affected by flooding, n is the number of regions not affected by flooding, and l is the number of regions where pipe network maintenance measures are taken.

[0027] The multi-objective group search algorithm module based on covariance evolution and chaos search solves the multi-objective programming model specifically including:

[0028] The multi-objective group search algorithm module adopts a covariance matrix adaptive evolution strategy to build a multi-objective programming model, then introduces a chaos operator and uses it to update the position of the population center, so that the population has good global search capabilities, thereby obtaining relatively accurate solution results.

[0029] The construction process of the deep learning module includes:

[0030] A pipeline network functional disease condition training sample set with pipeline desilting and descaling labels is input, and data processing is performed on the pipeline network functional disease condition training sample set, wherein the data processing includes outlier detection and interpolation.

[0031] The deep learning module uses the functional disease sample set of the drainage network with maintenance decision labels to calibrate the parameters of the loss function to perform iterative training on itself, specifically including:

[0032] The parameters of the loss function are determined by multiple iterations to obtain the minimum gradient. The quantization formula of the minimum gradient is:

[0033]

[0034] Among them, P is the output pipeline maintenance decision result, and I is the input training dataset.

[0035] The present invention adopts the numerical simulation and fusion technology of "multi-objective programming (MOP) + deep learning (DL) + hydrological and hydrodynamic coupling relationship model of functional diseases of drainage network and waterlogging" to realize the intelligent decision-making method for underground drainage network maintenance based on MOP-DL and make intelligent decisions on drainage network maintenance under rainstorms with different recurrence periods. The purpose of the present invention is to design and implement a MOP-DL-based intelligent decision-making method for underground drainage network maintenance based on the coupling relationship model of functional diseases of drainage network and waterlogging, MOP and DL, which can meet the requirements of accurately fitting the disaster losses in urban waterlogging areas under the influence of functional diseases, accurately quantifying the economic benefits of maintenance decisions, high efficiency and high reliability, and overcome the many shortcomings of current underground drainage network maintenance methods. Based on this, the present invention has the following advantages:

[0036] (1) Accurately fit the disaster losses of urban waterlogging areas under the influence of functional diseases. The intelligent decision-making method for underground drainage network maintenance based on MOP-DL uses the hydrological and hydrodynamic coupling relationship model between functional diseases of drainage network and waterlogging to design and implement an urban flood warning module that considers the coupling effects of pipeline functional diseases and urban waterlogging. Then, the quantitative relationship between the functional diseases of pipeline networks in different regions and waterlogging losses is obtained. In conjunction with the existing quantitative assessment method for waterlogging losses, the disaster loss assessment of urban waterlogging areas caused by urban drainage network under the influence of functional diseases is realized.

[0037] (2) Accurate quantification of the economic benefits of maintenance decisions. By considering the coupled effects of pipeline functional diseases and urban waterlogging, the urban flood warning module dynamically simulates the one- and two-dimensional instantaneous water flow state transition patterns of the city under the influence of functional diseases of the drainage network, accurately calculates waterlogging indicators such as flooding depth, flooding area, and flooding duration, as well as the resulting disaster losses. Combined with the costs of different maintenance strategies, this solves the problem of quantifying the economic benefits of maintenance decisions that has not been considered in existing studies.

[0038] (3) Intelligence and high efficiency. The MOP-DL-based intelligent decision-making method for underground drainage network maintenance uses a combination of multi-objective programming and deep learning to achieve intelligent and efficient output of drainage network maintenance decision results.

[0039] (4) High reliability. The present invention adopts the numerical simulation and fusion technology of "multi-objective programming (MOP) + deep learning (DL) + hydrological and hydrodynamic coupling relationship model of functional diseases of drainage network and waterlogging" to realize the intelligent decision-making method for underground drainage network maintenance based on MOP-DL. The deep learning method is used to train the drainage network maintenance optimization decision sample set obtained by multi-objective planning, and the training set is updated with a period T and the parameter value of the loss function is adjusted. With the cooperation of Python engineering implementation, DLL dynamic library linking, and Arcgis software, the reliability of the result output of the intelligent decision-making method for underground drainage network maintenance based on MOP-DL can be fully guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0041] Figure 1 This is a flowchart of an intelligent decision-making method for underground drainage network maintenance based on MOP-DL provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0043] Figure 1 Flowchart of an intelligent decision-making method for underground drainage network maintenance based on MOP-DL provided by an embodiment of the present invention, the method comprising:

[0044] S101. Obtain information on functional damage to the drainage network and losses caused by regional flood disasters;

[0045] The functional disease conditions of the drainage network and the losses caused by regional flood disasters are obtained. These data are used to create a dataset with maintenance decision labels. The dataset is used as the input of the trained deep learning module to obtain the final intelligent decision-making output for underground drainage network maintenance.

[0046] Among them, the functional disease conditions of the drainage network and the regional flood disaster losses include: the siltation conditions of pipelines in different sections (whether there is siltation), the degree of siltation, the length of siltation, the scaling conditions (whether there is scaling), the degree of scaling, and the length of scaling.

[0047] Input functional disease data into the urban pipe network disease initial sample input module;

[0048] Based on the hydrological and hydrodynamic coupling relationship model between functional diseases of drainage pipe networks and urban waterlogging and the PCSWMM module, an urban waterlogging early warning module was constructed that considers the coupling effects of functional diseases of pipe networks and urban waterlogging.

[0049] The urban waterlogging early warning module issues an early warning when water begins to accumulate on the ground. The coupling of pipeline functional defects with urban waterlogging in the urban waterlogging early warning module improves the accuracy of early warnings and enhances PCSWMM's ability to predict waterlogging.

[0050] The urban waterlogging early warning module is a secondary development of the PCSWMM module, which uses a dynamic library link to engineer an integrated hydrological and hydrodynamic coupling relationship model between functional diseases of the drainage network and waterlogging.

[0051] The hydrological and hydrodynamic coupling relationship model between functional diseases of the drainage network and waterlogging can be dynamically quantified using the one-dimensional clear-flow governing equation and the two-dimensional shallow-water equation, as follows:

[0052]

[0053]

[0054] Among them, Z represents the water level, A is the cross-sectional area of ​​the water flow, Q is the outlet flow of the cross section, and q L represents the lateral inflow, g is the gravitational acceleration, t and x represent the one-dimensional time and space coordinates, a represents the wave velocity, S f is the friction ratio, h L is the local head loss over the unit length, h is the water depth, t is the time, x, y and z are the coordinate systems, u and v are the velocity components in the x and y directions respectively, is the vertical average velocity, ρ is the fluid density, b is the bottom elevation, S ax and S ay are the bottom slope components in the x and y directions, S fx and S fy are the friction components in the x and y directions, τ zy and τ zy Both are lateral stresses.

[0055] Input the functional disease data in the urban pipe network disease initial sample input module into the urban waterlogging early warning module to obtain the functional disease status of the drainage network and the corresponding flood disaster status;

[0056] Input the flood disaster situation into the urban waterlogging loss calculation model, and then obtain the corresponding regional flood disaster loss situation;

[0057] The urban waterlogging damage calculation model uses the following formula to evaluate damage:

[0058] Damage value of waterlogging disaster = property density (yuan / km 2 )*Flooded area (km 2 )*Loss rate (%), Where s is the flooding water level.

[0059] S102. Using the multi-objective module as a constraint, a multi-objective planning model for underground drainage network maintenance decision-making is constructed;

[0060] Among them, the multi-objective module as the constraint condition includes: the constraint conditions include: maximum economic benefit, minimum maintenance cost and minimum damage caused by urban flooding.

[0061] The multi-objective planning model for building underground drainage network maintenance decisions specifically includes:

[0062] Objective function for maximizing regional economic benefits, objective function for minimizing regional underground drainage network maintenance costs, and objective function for minimizing urban flooding losses.

[0063] The quantification formulas are as follows:

[0064]

[0065]

[0066]

[0067] in, represents the economic benefits of area i where waterlogging did not occur, is the economic benefit of the flooded area j, C k is the cost of taking pipeline maintenance measures in area k, P i is the property density of region i, S i is the flooded area of ​​region i, s is the flooded water level, m is the number of regions affected by flooding, n is the number of regions not affected by flooding, and l is the number of regions where pipe network maintenance measures are taken.

[0068] S103. Based on the functional damage of the drainage network and the regional flood disaster losses, a multi-objective group search algorithm module based on covariance evolution and chaos search is used to solve the multi-objective planning model to obtain a sample set of functional damage conditions of the drainage network with maintenance decision labels;

[0069] The multi-objective group search algorithm module based on covariance evolution and chaos search solves the multi-objective programming model specifically including:

[0070] The multi-objective group search algorithm module constructs a covariance matrix adaptive evolutionary strategy for the multi-objective programming model, then introduces a chaos operator and uses it to update the position of the population center, so that the population has good global search capabilities, thereby obtaining relatively accurate solution results.

[0071] The search strategy of the leader is as shown in Equations (1)-(3), the search strategy of the follower is as shown in Equation (4), and the search strategy of the targetless is as shown in Equations (5)-(6).

[0072]

[0073]

[0074]

[0075]

[0076] R i+1 =μ·R i ·(1-R i ) (5)

[0077] N i+1 =N i +R i+1 ·(N i -ε) (6)

[0078] Among them, S c 、S l and S r are the updated targets for the leader’s search domain after the update, respectively, is the starting position of the p-th leader in the i-th round of search, r1 and r2 are random numbers that satisfy the standard normal distribution and random sequences that satisfy the (0, 1) uniform distribution, respectively, l max represents the maximum search distance, is the search direction, γ i represents the search angle, θ max is the maximum search angle, F i+1Indicates the starting position of the follower in the i+1th round of search, and Represent the mean and covariance matrix of the i-th round, v i is the step length, It means that the mean is 0 and the variance is The covariance matrix, R i 、R i+1 They represent the random generation sequences of rounds i and i+1 respectively, μ is the control parameter, N i 、N i+1 They represent the starting positions of the targetless player in rounds i and i+1 respectively, and ε is the Pareto optimal solution selected from the Pareto solution set.

[0079] 1) Each leader executes the optimal solution search strategy according to equations (1)-(3);

[0080] 2) In addition to the leader, select 70% of the nodes from the search population as followers, update the covariance matrix according to formula (4), and determine the evolution path and update step size of the followers;

[0081] 3) The remaining nodes are targetless and the chaotic search operations of equations (5) and (6) are performed;

[0082] 4) When the update round reaches the set threshold M, a Pareto optimal solution set is generated.

[0083] S104: Inputting the drainage network functional disease condition sample set with the maintenance decision label into a deep learning module;

[0084] The construction process of the deep learning module includes:

[0085] Input a training sample set of pipe network functional disease (siltation, scaling) conditions with pipeline desilting and descaling labels, and perform data processing on the training sample set of pipe network functional disease conditions. The data processing includes: outlier detection and interpolation. The specific steps of outlier detection and interpolation are as follows:

[0086] Outlier detection algorithm using a training sample set of functional defects in pipeline networks with maintenance decision labels:

[0087] Step 1: The kth distance d of the indicator data p of the pipeline network functional disease training sample set with maintenance decision labels k (p), which is defined as d k (p) = d(p,o), where the distance can be either time dimension or space dimension, satisfying:

[0088] (1) There are at least k points q∈C{x≠p} in the set, not including p, such that d(p,q)≤d(p,o);

[0089] (2) There are at most k-1 points q∈C{x≠p} in the set, not including p, such that d(p,q)<d(p,o);

[0090] Step 2: Calculate the kth distance field N of the indicator data p of the pipeline network functional disease training sample set with maintenance decision labels k (p), the number of k-th domain points N that satisfy p k (p)≥k;

[0091] Step 3: Calculate the kth reachable distance from the indicator data o to the data p of the pipeline network functional disease training sample set with maintenance decision labels:

[0092] d k (p,o)=max{d k (o),d(p,o)}

[0093] Step 4: Calculate the local reachability density of the indicator data p of the pipeline network functional disease training sample set with maintenance decision labels:

[0094]

[0095] Step 5: Calculate the local anomaly factor of the indicator data p of the pipeline network functional disease training sample set with maintenance decision labels:

[0096]

[0097] In step 6, a local outlier factor detection algorithm is used to calculate a LOF factor for each sample data set containing outliers. The calculated LOF factor and the set threshold are then used to determine whether the sample is an outlier. Samples with LOF factors exceeding the set threshold are classified as outliers, while those with LOF factors below the threshold are classified as normal.

[0098] Interpolation quantification method for pipeline network functional disease training sample set with maintenance decision labels:

[0099]

[0100] V prac (x0)=V fit (x0)+ε(x0)

[0101]

[0102]

[0103] V IDS (x i )=V fit (xi )+ε IDS (x i )

[0104] V CK (x i )=V fit (x i )+ε CK (x i )

[0105] Among them, V fit (x i ) is the multivariate regression fitting value of the interpolation point after removing outliers (the training sample set of functional disease of the pipeline network with maintenance decision labels); V prac (x0), V fit (x0) and ε(x0) are the actual values, multivariate regression fitting values ​​and residual values ​​of the functional disease training sample set of pipeline networks with maintenance decision labels at known points; m0, m1, m2, m3, m4 and m5 are the regression coefficients of the constant term and the fitting value respectively; λ, t, g and r represent the concentration of particulate matter near the pipeline, the average flow velocity of the pipeline water flow, the average flow rate of the pipeline and other index factors respectively; Y(x j ) are secondary variables, including the concentration of particulate matter near the pipeline, the average flow velocity of the pipeline water flow, the average flow rate of the pipeline and other index factors; n and m represent the interpolation points and the number of secondary variables respectively; d i is the distance from the predicted index data to the known index data i; ε IDS (x i ),ε CK (x i ), V IDS (x i ) and V CK (x i ) are the residual values ​​of the estimated points after IDS and CK interpolation and the interpolation values ​​corresponding to the relevant methods, respectively.

[0106] S105, the deep learning module uses the drainage network functional disease sample set with maintenance decision labels to calibrate the parameters of the loss function to perform iterative training on itself and output the maintenance decision result;

[0107] The deep learning module uses the functional disease sample set of the drainage network with maintenance decision labels to calibrate the parameters of the loss function to perform iterative training on itself, specifically including:

[0108] Grouped input of training dataset, grouped iterative unsupervised training, grouped iterative small-world transformation, and grouped iterative supervised learning.

[0109] The parameters of the loss function are determined by multiple iterations to obtain the minimum gradient. The quantization formula of the minimum gradient is: Where P is the output pipeline maintenance decision result, and I is the input training dataset.

[0110] The present invention adopts the intelligent decision-making system for underground drainage network maintenance of "multi-objective programming (MOP) + deep learning (DL) + hydrological and hydrodynamic coupling relationship model of functional diseases of drainage network and waterlogging", which fully considers the impact of the coupling relationship between functional diseases of underground drainage network and hydrological and hydrodynamic coupling of waterlogging on drainage network maintenance decision-making, and combines multi-objective programming, multi-objective group search algorithm based on covariance evolution and chaos search and deep learning inversion technology to improve the output efficiency and accuracy of the results of intelligent decision-making for underground drainage network maintenance. Therefore, the present invention has the advantages of high prediction efficiency, high accuracy and high reliability. The new output regional drainage network maintenance strategy results of the intelligent decision-making method for underground drainage network maintenance based on MOP-DL and the input samples of functional diseases of the network together constitute a labeled sample set, and the deep learning inversion is error corrected with a period T. Therefore, the present invention has the advantage that the accuracy of intelligent decision-making gradually increases with the increase of the number of iterations.

[0111] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention. These improvements and substitutions should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent decision-making method for underground drainage network maintenance based on MOP-DL, characterized by: include: Obtaining functional damage to the drainage network and regional flood disaster losses, said obtaining functional damage to the drainage network and regional flood disaster losses includes: Input functional disease data into the urban pipe network disease initial sample input module; Based on the hydrological and hydrodynamic coupling relationship model between functional defects of drainage pipe networks and urban waterlogging and the PCSWMM module, an urban waterlogging early warning module was constructed that considers the coupled impact of functional defects of pipe networks and urban waterlogging. The hydrological and hydrodynamic coupling relationship model between functional defects of drainage pipe networks and urban waterlogging is dynamically quantified using the one-dimensional clear-flow governing equation and the two-dimensional shallow-water equation. Input the functional disease data in the urban pipe network disease initial sample input module into the urban waterlogging early warning module to obtain the functional disease status of the drainage network and the corresponding flood disaster status; The flood disaster situation is input into the urban waterlogging loss calculation model to obtain the corresponding regional flood disaster loss situation. The urban waterlogging loss calculation model is: Damage value of urban flooding = property density (yuan / km 2 )*Flooded area (km 2 )*Loss rate (%), Loss rate = , where s is the height of the flooded water level; A multi-objective planning model for underground drainage network maintenance decision-making is constructed using multi-objective modules as constraints; According to the functional disease conditions of the drainage network and the regional flood disaster losses, a multi-objective group search algorithm module based on covariance evolution and chaos search is used to solve the multi-objective programming model to obtain a sample set of functional disease conditions of the drainage network with maintenance decision labels; Inputting the functional disease condition sample set of the drainage network with maintenance decision labels into the deep learning module; The deep learning module uses a sample set of functional disease conditions of the drainage network with maintenance decision labels to calibrate the parameters of the loss function to perform iterative training on itself and output the maintenance decision results.

2. The intelligent decision-making method for underground drainage network maintenance based on MOP-DL according to claim 1, characterized in that: The multi-objective module as the constraint condition includes: the constraint conditions include: maximum economic benefit, minimum maintenance cost and minimum damage caused by urban flooding.

3. The intelligent decision-making method for underground drainage network maintenance based on MOP-DL according to claim 2, characterized in that: The multi-objective planning model for underground drainage network maintenance decision-making specifically includes: Objective functions for maximizing regional economic benefits, minimizing regional underground drainage network maintenance costs, and minimizing urban waterlogging losses; The quantification formulas are as follows: ； ； ； in, represents the economic benefits of area i where waterlogging did not occur, is the economic benefit of the flooded area j, C k is the cost of taking pipeline maintenance measures in area k, is the property density of region i, S i is the flooded area of ​​region i, s is the flooded water level, m is the number of regions affected by flooding, n is the number of regions not affected by flooding, and l is the number of regions where pipe network maintenance measures are taken.

4. The intelligent decision-making method for underground drainage network maintenance based on MOP-DL according to claim 1, characterized in that: The multi-objective group search algorithm module based on covariance evolution and chaos search solves the multi-objective programming model specifically including: The multi-objective group search algorithm module adopts a covariance matrix adaptive evolution strategy to build a multi-objective programming model, then introduces a chaos operator and uses it to update the position of the population center, so that the population has good global search capabilities, thereby obtaining relatively accurate solution results.

5. The intelligent decision-making method for underground drainage network maintenance based on MOP-DL according to claim 1, characterized in that: The construction process of the deep learning module includes: A pipeline network functional disease condition training sample set with pipeline desilting and descaling labels is input, and data processing is performed on the pipeline network functional disease condition training sample set, wherein the data processing includes: outlier detection and interpolation.

6. The intelligent decision-making method for underground drainage network maintenance based on MOP-DL according to claim 1, characterized in that: The deep learning module uses the functional disease sample set of the drainage network with maintenance decision labels to calibrate the parameters of the loss function to perform iterative training on itself, specifically including: The parameters of the loss function are determined by obtaining the minimum gradient through multiple iterations. The quantization formula of the minimum gradient is: ； in, P To output the pipeline network maintenance decision results, I is the input training data set.

Images