A Big Data-Based Method for Optimizing Urban Sanitation Operations

Abstract

This invention discloses a big data-based method for optimizing urban sanitation operations, relating to the field of intelligent sanitation optimization technology. The method includes collecting urban sanitation data to construct a pollution area model, abstracting the detection process of mobile detection equipment as a Poisson stochastic process, deriving the probability distribution function of pollution area detection, calculating the minimum detection frequency for each area, using a simulated annealing algorithm to calculate the optimal detection frequency for each area, and using an improved NSGA-III multi-objective optimization algorithm for global planning and scheduling of sanitation operation paths. By combining the Poisson stochastic process model and the simulated annealing algorithm, the scientific quantification and optimized allocation of detection frequencies are achieved, achieving both accurate detection and effectively improving the utilization rate of detection equipment. The improved NSGA-III algorithm enhances the completion rate of regional detection while reducing resource consumption from ineffective operations, making sanitation task execution more efficient.

Description

Technical Field

[0001] This invention relates to the field of intelligent sanitation optimization technology, and in particular to a method for optimizing urban sanitation operations based on big data. Background Technology

[0002] With the acceleration of urbanization, the scheduling and optimization of resources required for urban sanitation work has gradually become a complex systemic problem. Traditional sanitation operation models typically rely on manual experience and fixed execution plans, making it impossible to dynamically adjust tasks based on regional pollution conditions or rationally allocate resources to cover constantly changing polluted areas within the city. In recent years, the widespread application of big data technology has provided more precise means for urban management. By collecting and analyzing real-time environmental data, the dynamic changes in urban pollution can be grasped from both macro and micro perspectives, and sanitation tasks and equipment deployment can be optimized. Although existing technologies have made some progress in pollution monitoring, operation statistics, and route planning, there are still many shortcomings, making it difficult to achieve a balance between efficient operation and resource conservation in the urban sanitation field. For example, traditional route planning algorithms often ignore the dynamic requirements of task priority and regional detection frequency, relying only on static geographic information resources, resulting in relatively fixed route designs that are difficult to adapt to rapid changes in pollution conditions. In addition, equipment fault monitoring is basically still at the manual processing stage, and research on fault prediction modeling and automated decision support is relatively weak, affecting the reliability and scheduling efficiency of sanitation equipment.

[0003] Existing technologies mainly focus on the following aspects: First, in the processing and analysis of pollution monitoring data, traditional methods are mostly based on simple pollution index thresholds, lacking modeling methods based on stochastic process theory. This makes it impossible to scientifically evaluate the probability of regional pollution being detected, thus hindering the calculation of the minimum detection frequency and the optimization of the optimal detection frequency. Second, in sanitation route planning algorithms, some studies have introduced multi-objective optimization methods, but have failed to fully consider the dynamic scheduling requirements and the importance of regional tasks in route optimization. In particular, in vehicle scheduling, there is no dynamic adjustment mechanism based on regional pollution models, and the potential of modern optimization algorithms is not fully utilized. Third, for the detection and maintenance of sanitation equipment malfunctions, the currently used simple fault alarm mode cannot efficiently predict the probability of malfunctions and adjust task plans in a timely manner, often leading to high-risk equipment participating in tasks and further deteriorating the working conditions. In addition, the assessment of equipment deployment scale lacks a method for quantitatively balancing cost and efficiency, failing to ensure the optimal deployment of equipment resources. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a big data-based method for optimizing urban sanitation operations, which solves the problems of inaccurate pollution assessment, inaccurate detection frequency, and low route scheduling efficiency in existing urban sanitation operations.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for optimizing urban sanitation operations based on big data, which includes collecting urban sanitation data and preprocessing it, determining the pollution status label of each area based on the preprocessed data, including pollution status, pollution type and risk level, constructing a pollution area model based on the determination results, generating multiple types of sanitation operation tasks according to the pollution status labels of each area in the pollution area model, and adjusting the pollution status label of the pollution area model based on the operation completion status.

[0008] The pollution detection process of mobile detection equipment is abstracted as a Poisson stochastic process. After deriving the probability distribution function of the detection of polluted areas, the minimum detection frequency of each area is calculated. Based on the minimum detection frequency of the area, the simulated annealing algorithm is used to calculate the optimal detection frequency of each area. Based on the optimal detection frequency of the area, the improved NSGA-III multi-objective optimization algorithm is used to perform global planning and scheduling of sanitation operation paths.

[0009] During sanitation operation, an LSTM prediction model is used to predict the failure probability of sanitation equipment, screen out high-risk equipment that is not included in route scheduling, and automatically generate maintenance orders. At the same time, the cost-efficiency function method is used to calculate the deployment scale critical point, and sanitation equipment is deployed based on the calculation results.

[0010] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the step of using an improved NSGA-III multi-objective optimization algorithm to globally plan and schedule sanitation operation paths based on the regional optimal detection frequency refers to... Generate N scheduling schemes, each denoted as S, and the task allocation for each region must satisfy its optimal detection frequency.

[0011] For each scheduling scheme, calculate the optimization objective function value, including the function f1(S) for minimizing the total path length, the function f2(S) for maximizing detection coverage, the function f3(S) for maximizing task balance, and the function f4(S) for minimizing operating cost. The result of the multi-objective optimization calculation is recorded as the objective function value F(S).

[0012] Based on the objective function value F(S), all solutions in the solution set are Pareto ranked. If solution S i It is no worse than solution S across all optimization objectives. jAnd scheme S i It is significantly superior to solution S in at least one optimization objective. j Then scheme S i Pareto Domination Scheme j ;

[0013] Calculate the sparsity d of the set of nondominated schemes. crowd (S i Set sparsity threshold. If d crowd (S i Greater than or equal to Then scheme S i Add it to the set of preferred solutions; otherwise, do not use it.

[0014] Based on the selected set of preferred solutions, crossover and mutation operations are performed. Two solutions are randomly selected from the set of preferred solutions, and their partial task scheduling is swapped to generate a new sub-solution S. new And randomly select an indefinite number of preferred schemes, while meeting the regional detection frequency Under the given conditions, the task allocation of the scheme is randomly adjusted to generate new sub-schemes. The generated sub-schemes are then integrated with the preferred scheme to obtain a new set of candidate schemes W. new ;

[0015] Set the maximum number of iterations. For set W new The schemes in the solution are Pareto sorted, and the sparsity of the non-dominated schemes obtained after sorting is calculated. The non-dominated schemes with sparsity greater than a set threshold are set as a new set of preferred schemes. Before the number of iterations reaches the threshold, new sub-schemes are repeatedly generated, and a new set of preferred schemes is obtained after Pareto sorting and sparsity calculation.

[0016] When the maximum number of iterations is reached Then, output the optimal solution S that is non-dominated and has the highest sparsity. best As the optimal scheduling scheme.

[0017] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the step of generating multiple types of sanitation operation tasks based on the pollution status labels of each area in the pollution area model, and adjusting the pollution status labels of the pollution area model based on the operation completion status refers to reading the pollution status, pollution type, and risk level of each area in the pollution area model, and determining whether to generate sanitation operation tasks based on the read content:

[0018] If the pollution status is "polluted", a sanitation task will be generated; otherwise, it will not be generated.

[0019] If the pollution type is particulate matter, a water spraying task will be generated; if it is odor, a spraying task will be generated; if it is stockpiling, a garbage removal task will be generated.

[0020] The risk level is equivalent to the task priority, and the generated sanitation operation tasks are sorted according to the task priority and incorporated into the operation schedule.

[0021] The task completion status is uploaded through the vehicle terminal. If the task is completed, the pollution status of the corresponding area is set to "no pollution". If the task is not completed, the pollution status is maintained and sanitation operation tasks continue to be generated.

[0022] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the pollution detection process of the mobile detection device is abstracted as a Poisson random process. After deriving the probability distribution function of the detected polluted area, the minimum detection frequency of each area is calculated, which refers to the number of times λ of the i-th area needs to be detected per unit time. i ;

[0023] The detection behavior of mobile pollution detection equipment in each area is modeled as a Poisson distribution P;

[0024] The probability of the i-th region being detected at least once within the time interval ρ is calculated based on the Poisson distribution.

[0025] Set a probability threshold θ for the i-th region to be detected at least once within a time interval ρ, and calculate the minimum required detection frequency.

[0026] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the optimal detection frequency for each region is calculated using a simulated annealing algorithm based on the lowest regional detection frequency. Initialize to the current detection frequency

[0027] Based on energy consumption cost function With response performance function Calculate the current detection frequency The optimization objective function value

[0028] Candidate detection frequencies are generated in the neighborhood of the current detection frequency. and Greater than

[0029] Calculate the optimization objective function value of candidate frequencies Compare the optimization objective function values ​​of the candidate detection frequencies with the current detection frequency. Greater than or equal to Then maintain the current detection frequency, if Less than Then set the candidate detection frequency to the current detection frequency;

[0030] Calculate the difference between the current detection frequency and the candidate detection frequency in the optimized objective function value.

[0031] Set the threshold for changes in the objective function. like Greater than or equal to Then continue generating candidate detection frequencies, if Less than This indicates that the frequency has converged to the optimal frequency, and the output is... This is the optimal detection frequency.

[0032] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the sanitation operation process uses an LSTM prediction model to predict the failure probability of sanitation equipment, filters out high-risk equipment from participating in path scheduling, and automatically generates a maintenance log to record equipment information once per hour, outputting the state sequence X of the equipment at the most recent n moments.

[0033] Based on the state sequence of the equipment, an LSTM model is used to predict the probability ξ of the equipment failing within the next τ time. A failure probability threshold is set. If the predicted failure probability is greater than or equal to the failure probability threshold, the equipment is identified as high-risk equipment, does not participate in the route planning of sanitation work, and a maintenance order is directly generated for equipment maintenance.

[0034] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the method uses a cost-efficiency function to calculate the deployment scale critical point, and deploys sanitation equipment based on the calculation results, indicating the required investment. Cost function of a device Constructing efficiency functions Calculate unit efficiency cost right Find the minimum point where the derivative is zero to determine the optimal deployment. Calculated The optimal number of devices to deploy.

[0035] As a preferred embodiment of the big data-based urban sanitation operation optimization method described in this invention, the following steps are taken: collecting urban sanitation data and preprocessing it, determining the pollution status label of each area based on the preprocessed data, and constructing a pollution area model based on the determination results. This involves deploying sensing devices on fixed monitoring stations and mobile detection equipment to collect environmental data in real time, including geographical location, particulate matter size, environmental images, abnormal gas concentration, and equipment number. The data is then preprocessed and uploaded to the database.

[0036] The pollution status label includes pollution status, pollution type, and risk level. Thresholds are set for air particle size, abnormal gas concentration, and solid waste area. Data in the database is compared with the set thresholds. If the particle size is large, the abnormal gas concentration is high, or the solid waste area identified by the image is greater than or equal to the set threshold, the pollution status of the area is marked as "polluted". The three pollution types are recorded as particulate matter type, odor type, and accumulation type, respectively. Otherwise, the pollution status of the area is marked as "unpolluted".

[0037] The pollution index value A is calculated using normalization, and the pollution risk level value (RiskScore) for the i-th region is calculated based on the pollution index value and the duration of pollution. i ;

[0038] Set a low-risk threshold, RiskScore low and high-risk threshold RiskScore high :

[0039] If RiskScore i Less than RiskScore low If the risk level of the i-th region is set to level one, then the risk level of the i-th region will be set to level one.

[0040] If RiskScore i Greater than or equal to RiskScore low And less than or equal to RiskScore high If so, then the risk level of the i-th region is set to level two;

[0041] If RiskScore i Greater than RiskScore high If so, the risk level of the i-th region is set to level three;

[0042] The city is divided into a grid model of size m, where nodes represent regions and edges represent adjacency relationships. The pollution status label of each region is embedded into the grid model to form a pollution region model.

[0043] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the big data-based urban sanitation operation optimization method described in the first aspect of the present invention.

[0044] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the big data-based urban sanitation operation optimization method described in the first aspect of the present invention.

[0045] The beneficial effects of this invention are as follows: By combining the Poisson stochastic process model and the simulated annealing algorithm, the scientific quantification and optimized allocation of detection frequency are achieved, which not only achieves the goal of accurate detection but also effectively improves the utilization rate of detection equipment and reduces the cost of detection resource allocation; the improved NSGA-III enables the dynamic and real-time nature of path planning, improves the completion rate of regional detection, and reduces the resource consumption of ineffective operation, making the execution of sanitation tasks more efficient; the use of the LSTM model to predict the probability of equipment failure and optimize equipment scheduling realizes the intelligentization of failure prediction and automated maintenance decision-making, which significantly improves the utilization efficiency of sanitation equipment, while further reducing the cost of equipment deployment and improving the reliability of task execution through scale optimization. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a flowchart of a big data-based urban sanitation operation optimization method in Example 1.

[0048] Figure 2 This is the core flowchart of a big data-based urban sanitation operation optimization method in Example 1. Detailed Implementation

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0050] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0051] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0052] Example 1, referring to Figure 1 and Figure 2This is the first embodiment of the present invention, which provides a method for optimizing urban sanitation operations based on big data, including the following steps:

[0053] S1: Collect urban sanitation data and preprocess it. Based on the preprocessed data, determine the pollution status label of each area, including pollution status, pollution type and risk level. Based on the determination results, construct a pollution area model. Based on the pollution status label of each area in the pollution area model, generate multiple types of sanitation operation tasks. Adjust the pollution status label of the pollution area model based on the operation completion status.

[0054] Specifically, the process involves collecting and preprocessing urban sanitation data, determining the pollution status label of each area based on the preprocessed data, and constructing a pollution area model based on the determination results. This involves deploying various sensing devices, including particulate matter sensors, image acquisition devices, and gas detection devices, at fixed monitoring stations and mobile detection equipment to collect environmental data in real time, including geographical location, particulate matter size, environmental images, abnormal gas concentrations, and device numbers. The data is then preprocessed and uploaded to the database.

[0055] The pollution status label includes pollution status, pollution type, and risk level. Thresholds are set for air particle size, abnormal gas concentration, and solid waste area. Data in the database is compared with the set thresholds. If the particle size is large, the abnormal gas concentration is high, or the solid waste area identified by the image is greater than or equal to the set threshold, the pollution status of the area is marked as "polluted". The three pollution types are recorded as particulate matter type, odor type, and accumulation type, respectively. Otherwise, the pollution status of the area is marked as "unpolluted".

[0056] Normalized calculation of pollution index value A:

[0057]

[0058] Among them, a grain a air a solid These represent the actual pollution values ​​for airborne particle size, gas concentration, and solid waste area. The maximum tolerance values ​​set for airborne particle size, gas concentration, and solid waste area, ψ grain ψ air ψ solid Weight values ​​were assigned to the size of airborne particles, gas concentration, and area of ​​solid waste.

[0059] Calculate the pollution risk level value (RiskScorei) for the i-th region based on the pollution index value and the duration of pollution:

[0060] RiskScore i =ω1·A i+ω2·D i ,

[0061] Among them, A i Let D be the normalized pollution index value for the i-th region. i Let ω1 and ω2 be the duration of pollution in the i-th region, and let ω1 and ω2 be weight values ​​that can be optimized through training.

[0062] Set a low-risk threshold, RiskScore low and high-risk threshold RiskScore high :

[0063] If RiskScore i Less than RiskScore low If the risk level of the i-th region is set to level one, then the risk level of the i-th region will be set to level one.

[0064] If RiskScore i Greater than or equal to RiskScore low And less than or equal to RiskScore high If so, then the risk level of the i-th region is set to level two;

[0065] If RiskScore i Greater than RiskScore high If so, the risk level of the i-th region is set to level three;

[0066] The city is divided into a grid model of size m, where nodes represent regions and edges represent adjacency relationships. The pollution status label of each region is embedded into the grid model to form a pollution region model.

[0067] By comparing data collected from the database with set thresholds, pollution can be automatically classified, significantly improving the targeting of sanitation tasks. Normalized pollution index values ​​are calculated and regional risk levels are assessed, enabling dynamic risk analysis of polluted areas and effectively improving the efficiency of emergency response. The pollution area model comprehensively details the distribution of pollution in the city and provides a spatially correlated input structure, offering clear and complete data support for complex route planning.

[0068] Furthermore, based on the pollution status labels of each area in the pollution area model, multiple types of sanitation operation tasks are generated. The pollution status labels in the pollution area model are adjusted based on the task completion status. This involves reading the pollution status, pollution type, and risk level of each area in the pollution area model, and determining whether to generate a sanitation operation task based on the read information.

[0069] If the pollution status is "polluted", a sanitation task will be generated; otherwise, it will not be generated.

[0070] If the pollution type is particulate matter, a water spraying task will be generated; if it is odor, a spraying task will be generated; if it is stockpiling, a garbage removal task will be generated.

[0071] The risk level is equivalent to the task priority, and the generated sanitation operation tasks are sorted according to the task priority and incorporated into the operation schedule.

[0072] The task completion status is uploaded through the vehicle terminal. If the task is completed, the pollution status of the corresponding area is set to "no pollution". If the task is not completed, the pollution status is maintained and sanitation operation tasks continue to be generated.

[0073] By assessing the pollution status of a region, sanitation tasks can be precisely triggered, avoiding unnecessary waste of resources. Identifying the pollution type ensures that the task type matches the pollution situation, improving the professionalism and effectiveness of pollution treatment. Sanitation operations are prioritized and scheduled based on risk levels, allowing for the priority handling of tasks in high-risk areas and improving the efficiency and timeliness of task execution. The pollution area model is updated in real time through the completion status of operations, achieving closed-loop management of pollution status and strengthening the reliability of the entire sanitation task scheduling.

[0074] S2: The pollution detection process of mobile detection equipment is abstracted as a Poisson random process. After deriving the probability distribution function of the detected pollution area, the minimum detection frequency of each area is calculated. Based on the minimum detection frequency of the area, the simulated annealing algorithm is used to calculate the optimal detection frequency of each area. Based on the optimal detection frequency of the area, the improved NSGA-III multi-objective optimization algorithm is used to perform global planning and scheduling of sanitation operation paths.

[0075] Specifically, the pollution detection process of mobile detection equipment is abstracted as a Poisson stochastic process. After deriving the probability distribution function of the detected pollution area, the minimum detection frequency of each area is calculated, which refers to the number of times λ of the i-th area needs to be detected per unit time. i :

[0076]

[0077] Where, λ i Let λ be the number of times the i-th region needs to be detected, and λ0 be the set global base detection frequency. The risk level weights are set accordingly;

[0078] The detection behavior of mobile pollution detection equipment in each area is modeled as a Poisson distribution P:

[0079]

[0080] Where, N i(ρ) represents the number of times the i-th region is detected within the time interval ρ, z represents the number of times the detection event occurs, and ρ represents the set fixed time.

[0081] The probability of the i-th region being detected at least once within the time interval ρ is calculated based on the Poisson distribution.

[0082]

[0083] Let θ be the probability threshold that the i-th region is detected at least once within the time interval ρ. Then:

[0084]

[0085] Calculate the minimum required detection frequency

[0086]

[0087] By calculating the number of times each area needs to be detected per unit time, and combining this with a scientific quantification of regional pollution needs, the regional pollution detection behavior is modeled as a Poisson process. This provides a scientific and quantifiable method for calculating key probability functions during the detection process. Based on the Poisson distribution, the probability that an area will be detected at least once per unit time is calculated, enabling the evaluation of detection coverage effectiveness. The calculation of the minimum detection frequency optimizes the spatiotemporal allocation efficiency of detection resources and improves the accuracy of detection planning.

[0088] Furthermore, based on the lowest detection frequency in a region, the simulated annealing algorithm is used to calculate the optimal detection frequency for each region, which is then used to initialize the current detection frequency.

[0089]

[0090] Calculate the current detection frequency The optimization objective function value

[0091]

[0092] in, The energy consumption cost function of the current detection frequency. Let α be the response performance function for the current detection frequency. F With β F These are energy consumption cost weights and response efficiency weights, respectively, set based on sanitation operation needs and budget constraints.

[0093] The energy consumption cost function Specifically, it is expressed as follows:

[0094]

[0095] Among them, v i b is the energy cost coefficient per unit frequency. i For fixed maintenance costs;

[0096] The response performance function Specifically, it is expressed as follows:

[0097]

[0098] Among them, g i This is the performance weighting coefficient;

[0099] Generate candidate detection frequencies in the neighborhood of the current detection frequency. and Greater than

[0100]

[0101] Where Δλ is the disturbance generated by a uniform normal distribution;

[0102] Calculate the optimization objective function value of candidate frequencies

[0103]

[0104] Compare the optimization objective function values ​​of the candidate detection frequencies with the current detection frequencies:

[0105] like Greater than or equal to Then maintain the current detection frequency;

[0106] like Less than Then set the candidate detection frequency to the current detection frequency;

[0107] Calculate the difference between the current detection frequency and the candidate detection frequency in the optimized objective function value.

[0108]

[0109] Set the threshold for changes in the objective function.

[0110] like Greater than or equal to Then continue generating candidate detection frequencies;

[0111] like Less than This indicates that the frequency has converged to the optimal frequency, and the output is... This is the optimal detection frequency.

[0112] By designing energy consumption cost function and response performance function, optimization objectives for resource cost and detection performance are constructed, ensuring that the optimization of detection frequency is both efficient and cost-effective. Candidate frequencies are generated in the frequency neighborhood to calculate the optimization objective value, which enhances the diversity of optimization search and greatly increases the possibility of escaping local optima. The optimization objective value is compared and high-quality detection frequencies are retained, ensuring that the objective function tends to the best global optimal detection frequency round by round.

[0113] Furthermore, based on the optimal regional detection frequency, an improved NSGA-III multi-objective optimization algorithm is used to perform global planning and scheduling of sanitation operation routes. Generate N scheduling schemes, each denoted as S:

[0114] S = {T1:[R1,R2,...],T2:[R...]} i ,R i+1 ],...,T j :[R m-1 ,R m ]},

[0115] Among them, T j For the j-th sanitation vehicle, R i The i-th region is the area to be inspected by the sanitation vehicle, where m is the number of rasterized regions in the pollution area model, and the task allocation for each region must satisfy its optimal detection frequency.

[0116] For each scheduling scheme, calculate the optimization objective function value, including total path length, detection coverage, task balancing, and operating cost:

[0117] The function f1(S) minimizes the total path length:

[0118]

[0119] Among them, G j Let j be the path of the j-th sanitation vehicle. p represents the total number of sanitation vehicles. k p l For the task point, y(p) k ,p l () represents the distance between task points;

[0120] The detection coverage maximization function f2(S) is:

[0121]

[0122] Where, δ i The indicator function is defined as follows: if the detection frequency of the i-th region reaches... Then δi =1, otherwise δ i =0;

[0123] The task balance maximization function f3(S) is:

[0124]

[0125] in, Let be the working time of the j-th sanitation vehicle, and VAR be the working time variance.

[0126] The function to minimize the running cost is f4(S):

[0127]

[0128] in, For the operating costs of sanitation vehicles, The resource cost for the vehicle to complete the task;

[0129] The result of the multi-objective optimization calculation is denoted as the objective function value F(S):

[0130] F(S)=[f1(S),f2(S),f3(S),f4(S)],

[0131] Based on the objective function value F(S), all solutions in the solution set are Pareto ranked.

[0132]

[0133] in, For the first in the plan An optimization objective, if solution S i It is no worse than solution S across all optimization objectives. j And scheme S i It is significantly superior to solution S in at least one optimization objective. j Then scheme S i Pareto Domination Scheme j ;

[0134] Calculate the sparsity d of the set of nondominated schemes. crowd (S i ):

[0135]

[0136] in, This represents the total number of optimization objectives in the plan. and These are schemes S i In the The target values ​​of the left and right neighbor solutions for each optimization objective;

[0137] Set sparsity threshold If d crowd (S i Greater than or equal to Then scheme S i Add it to the set of preferred solutions; otherwise, do not use it.

[0138] Based on the selected set of preferred solutions, crossover and mutation operations are performed to randomly select two solutions S from the set of preferred solutions. i and S j By exchanging the task scheduling of some of the two schemes, a new sub-scheme S is generated. new And randomly select an indefinite number of preferred schemes, while meeting the regional detection frequency Under the given conditions, the task allocation of the scheme is randomly adjusted to generate new sub-schemes. The generated sub-schemes are then integrated with the preferred scheme to obtain a new set of candidate schemes W. new ;

[0139] Set the maximum number of iterations. For set W new The schemes in the solution are Pareto sorted, and the sparsity of the non-dominated schemes obtained after sorting is calculated. The non-dominated schemes with sparsity greater than a set threshold are set as a new set of preferred schemes. Before the number of iterations reaches the threshold, new sub-schemes are repeatedly generated, and a new set of preferred schemes is obtained after Pareto sorting and sparsity calculation.

[0140] When the maximum number of iterations is reached Then, output the optimal solution S that is non-dominated and has the highest sparsity. best As the optimal scheduling scheme.

[0141] By generating a scheduling scheme based on the optimal detection frequency in each region, the planning of sanitation tasks is ensured to meet the detection needs of each region, optimizing the accuracy of resource allocation. The scheme achieves multi-objective optimization values ​​for total path length, coverage, task balance, and operating cost, making the scheme comprehensively optimal across multiple objectives. Pareto sorting and sparsity calculation are used to screen non-dominated schemes and optimize the distribution performance of the scheduling scheme, enhancing the representativeness and diversity of the preferred scheme set. Crossover and mutation operations are used to free the scheme optimization from local optima limitations, enhancing the global optimization capability of the final scheme. Iterative optimization further improves the quality of the scheme, ultimately outputting the non-dominated scheme with the highest sparsity as the final scheme, ensuring that the scheme achieves balance across multi-dimensional optimization objectives and possesses high efficiency and low cost in task execution.

[0142] S3: During the sanitation operation, the LSTM prediction model is used to predict the failure probability of sanitation equipment, screen out high-risk equipment that does not participate in route scheduling, and automatically generate maintenance orders. At the same time, the cost-efficiency function method is used to calculate the deployment scale critical point, and sanitation equipment is deployed based on the calculation results.

[0143] Specifically, during sanitation operations, an LSTM prediction model is used to predict the failure probability of sanitation equipment, filter out high-risk equipment from route scheduling, and automatically generate maintenance logs that record equipment information once per hour, outputting the state sequence X of the equipment at the most recent n moments:

[0144] X = [x t-n x t-n+1 x t-n+2 , ..., x t ],

[0145] Where, x t The state of the device at time t is specifically represented as follows:

[0146] x t =[Q t O t U t State t ],

[0147] Among them, Q t Let O be the temperature of the device at time t. t Let U be the vibration intensity of the equipment at time t. t State represents the power consumption of the device at time t. t State represents the abnormal state of the device at time t. State = 0 indicates that the device is in normal condition, and State = 1 indicates that the device is in abnormal condition.

[0148] Based on the device's state sequence, an LSTM model is used to predict the probability ξ of the device failing within the next τ time interval:

[0149] ξ = LSTM(X),

[0150] Where ξ∈[0,1], the closer it is to 1, the higher the probability that the device will fail in the next τ time period;

[0151] The use of the LSTM model specifically refers to first using the LSTM_cell basic unit in LSTM to calculate the hidden state h of the device. t :

[0152] h t =LSTM_cell(x t h t-1 ),

[0153] Among them, ht h represents the hidden state of the device's final output. t-1 The hidden state of the device at the previous time step; LSTM_cell is a basic computational unit in the LSTM model.

[0154] Based on the device's final hidden state, the probability of the device failing within the next τ time interval is calculated using the Sigmoid function:

[0155]

[0156] Where Sigmoid is the function set in the last layer of the LSTM model, and κ is the weight obtained by training based on the set prediction time τ. The bias is obtained by training based on a set prediction time τ;

[0157] Set a failure probability threshold ξ θ If ξ is greater than or equal to ξ θ If the equipment is deemed high-risk, it will not be included in the route planning for sanitation work, and a maintenance order will be generated directly for equipment maintenance.

[0158] By generating state sequences of equipment, the time dynamic characteristics of equipment operation status are captured, providing a data foundation for fault probability prediction. The LSTM model is used to integrate short-term and long-term data of equipment operation, calculate the hidden state of the equipment, and use the Sigmoid function based on the hidden state to transform the complex equipment state into an intuitive fault probability. This provides a quantitative evaluation standard for equipment management, screens out high-risk equipment from participating in sanitation tasks and repairs it in advance, avoids the impact of risky equipment on sanitation tasks, and improves the efficiency of equipment maintenance.

[0159] Furthermore, the cost-efficiency function method is used to calculate the critical point for deployment scale, and the deployment of sanitation equipment is based on the calculation results to determine the required investment. Cost function of a device

[0160]

[0161] Where Γ0 represents fixed operating costs, Γ u Cost per unit of equipment;

[0162] Constructing efficiency functions

[0163]

[0164] in, ι is the basic efficiency constant determined by the structure of the operating system, γ is the positive exponent representing the efficiency improvement brought about by equipment deployment, and γ is the interference parameter reflecting the efficiency decline caused by excessive deployment.

[0165] Calculate unit efficiency cost

[0166]

[0167] right Find the minimum point where the derivative is zero to determine the optimal deployment.

[0168]

[0169] Calculated The optimal number of devices to deploy.

[0170] By constructing an input cost and efficiency function, the impact of the number of equipment on cost efficiency is quantified, providing a quantitative basis for the deployment planning of sanitation equipment. Using unit efficiency cost as a metric, the derivative of unit efficiency cost is used to find the minimum point, minimizing the cost expenditure corresponding to unit efficiency while ensuring that the deployment plan has sufficient task completion capability.

[0171] This embodiment also provides a computer device applicable to a big data-based urban sanitation operation optimization method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the big data-based urban sanitation operation optimization method proposed in the above embodiment.

[0172] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0173] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a big data-based urban sanitation operation optimization method as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0174] In summary, this invention achieves scientific quantification and optimized allocation of detection frequency by combining the Poisson stochastic process model and simulated annealing algorithm. This not only achieves accurate detection but also effectively improves the utilization rate of detection equipment while reducing the cost of detection resource allocation. The improved NSGA-III enables dynamic and real-time path planning, improving the completion rate of regional detection and reducing resource consumption from ineffective operations, thus making sanitation tasks more efficient. Furthermore, the use of an LSTM model to predict equipment failure probabilities and optimize equipment scheduling realizes intelligent decision-making for fault prediction and automated maintenance, significantly improving the utilization efficiency of sanitation equipment. Simultaneously, scale optimization further reduces equipment deployment costs and improves the reliability of task execution.

[0175] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for optimizing urban sanitation operations based on big data, characterized in that: include, Collect urban sanitation data and preprocess it. Based on the preprocessed data, determine the pollution status label of each area, including pollution status, pollution type and risk level. Based on the determination results, construct a pollution area model. Based on the pollution status label of each area in the pollution area model, generate multiple types of sanitation operation tasks. Adjust the pollution status label of the pollution area model based on the operation completion status. The pollution detection process of mobile detection equipment is abstracted as a Poisson stochastic process. After deriving the probability distribution function of the detection of polluted areas, the minimum detection frequency of each area is calculated. Based on the minimum detection frequency of the area, the simulated annealing algorithm is used to calculate the optimal detection frequency of each area. Based on the optimal detection frequency of the area, the improved NSGA-III multi-objective optimization algorithm is used to perform global planning and scheduling of sanitation operation paths. During sanitation operation, an LSTM prediction model is used to predict the failure probability of sanitation equipment, screen out high-risk equipment that is not included in route scheduling, and automatically generate maintenance orders. At the same time, the cost-efficiency function method is used to calculate the deployment scale critical point, and sanitation equipment is deployed based on the calculation results.

2. The urban sanitation operation optimization method based on big data as described in claim 1, characterized in that: The improved NSGA-III multi-objective optimization algorithm, based on the optimal regional detection frequency, is used for global planning and scheduling of sanitation operation routes. Generate N scheduling schemes, each denoted as S, and ensure that the task allocation for each region satisfies its optimal detection frequency. ; For each scheduling scheme, calculate the optimization objective function value, including the function for minimizing the total path length. Detection coverage maximization function Task balance maximization function and the function for minimizing operating costs The result of the multi-objective optimization calculation is recorded as the objective function value. ; Based on the objective function value Perform Pareto sort on all solutions in the solution set. If the solution... The solution is no worse than the previous one across all optimization objectives. And the plan It is significantly superior to the proposed solution in at least one optimization objective. Then the solution Pareto Domination Program ; Calculate the sparsity of the set of nondominated schemes. Set sparsity threshold ,like Greater than or equal to Then the plan Add it to the set of preferred solutions; otherwise, do not use it. Based on the selected set of optimal solutions, crossover and mutation operations are performed. Two solutions are randomly selected from the set of optimal solutions, and their partial task scheduling is swapped to generate new sub-solutions. And randomly select an indefinite number of preferred schemes, while meeting the regional detection frequency Under the given conditions, the task allocation of the scheme is randomly adjusted to generate new sub-schemes. The generated sub-schemes are then integrated with the preferred scheme to obtain a new set of candidate schemes. ; Set the maximum number of iterations. For sets The schemes in the solution are Pareto sorted, and the sparsity of the non-dominated schemes obtained after sorting is calculated. The non-dominated schemes with sparsity greater than a set threshold are set as a new set of preferred schemes. Before the number of iterations reaches the threshold, new sub-schemes are repeatedly generated, and a new set of preferred schemes is obtained after Pareto sorting and sparsity calculation. When the maximum number of iterations is reached Then, output the optimal solution that is non-dominated and has the highest sparsity. As the optimal scheduling scheme.

3. The urban sanitation operation optimization method based on big data as described in claim 2, characterized in that: The process of generating multiple types of sanitation operation tasks based on the pollution status labels of each area in the pollution area model, and adjusting the pollution status labels of the pollution area model based on the task completion status, refers to reading the pollution status, pollution type, and risk level of each area in the pollution area model, and determining whether to generate sanitation operation tasks based on the read content. If the pollution status is "polluted", a sanitation task will be generated; otherwise, it will not be generated. If the pollution type is particulate matter, a water spraying task will be generated; if it is odor, a spraying task will be generated; if it is stockpiling, a garbage removal task will be generated. The risk level is equivalent to the task priority, and the generated sanitation operation tasks are sorted according to the task priority and incorporated into the operation schedule. The task completion status is uploaded through the vehicle terminal. If the task is completed, the pollution status of the corresponding area is set to "no pollution". If the task is not completed, the pollution status is maintained and sanitation operation tasks continue to be generated.

4. The urban sanitation operation optimization method based on big data as described in claim 3, characterized in that: The process of pollution detection by mobile detection equipment is abstracted as a Poisson random process. After deriving the probability distribution function of the detected pollution area, the minimum detection frequency of each area is calculated, which refers to the number of times the i-th area needs to be detected per unit time. ; The detection behavior of mobile pollution detection equipment in each area is modeled as a Poisson distribution P; Calculate the i-th region in the time period based on the Poisson distribution. The probability of being detected at least once. ; Set the i-th region in the time interval The probability threshold for being detected at least once within a given timeframe Calculate the minimum required detection frequency .

5. The urban sanitation operation optimization method based on big data as described in claim 4, characterized in that: The optimal detection frequency for each region is calculated using a simulated annealing algorithm based on the region's lowest detection frequency. Initialize to the current detection frequency ; Based on energy consumption cost function With response performance function Calculate the current detection frequency The optimization objective function value ; Candidate detection frequencies are generated in the neighborhood of the current detection frequency. ,and Greater than ; Calculate the optimization objective function value of candidate frequencies Compare the optimization objective function values ​​of the candidate detection frequencies with the current detection frequency. Greater than or equal to If so, maintain the current detection frequency. Less than If so, the candidate detection frequency is set to the current detection frequency; Calculate the difference between the current detection frequency and the candidate detection frequency in the optimized objective function value. ; Set a threshold for the change of the objective function, if | If the value is greater than or equal to the threshold of change in the objective function, then continue generating candidate detection frequencies. | If the value is less than the threshold of the objective function change, it indicates that the frequency has converged to the optimal frequency, and the output is... This is the optimal detection frequency.

6. The urban sanitation operation optimization method based on big data as described in claim 5, characterized in that: During the sanitation operation, the LSTM prediction model is used to predict the failure probability of sanitation equipment, screen out high-risk equipment that does not participate in route scheduling, and automatically generate maintenance orders to record equipment information once per hour, outputting the state sequence X of the equipment at the most recent n moments; Based on the device's state sequence, an LSTM model is used to predict the device's future state. Probability of failure within a certain time period A fault probability threshold is set. If the predicted fault probability is greater than or equal to the fault probability threshold, the equipment is identified as high-risk equipment, will not participate in the route planning of sanitation work, and a maintenance order will be generated directly for equipment maintenance.

7. The urban sanitation operation optimization method based on big data as described in claim 6, characterized in that: The cost-efficiency function method is used to calculate the critical point for deployment scale, and the deployment of sanitation equipment is based on the calculation results to determine the investment. Cost function of a device Construct an efficiency function Calculate unit efficiency cost ( ),right ( Find the derivative and the minimum point where the derivative is zero as the optimal deployment. , using calculations The optimal number of devices to deploy.

8. The urban sanitation operation optimization method based on big data as described in claim 7, characterized in that: The process of collecting and preprocessing urban sanitation data, determining the pollution status label of each area based on the preprocessed data, and constructing a pollution area model based on the determination results refers to deploying sensing devices on fixed monitoring stations and mobile detection equipment to collect environmental data in real time, including geographical location, particulate matter size, environmental images, abnormal gas concentration, and equipment number. The data is then preprocessed and uploaded to the database. The pollution status label includes pollution status, pollution type, and risk level. Thresholds are set for air particle size, abnormal gas concentration, and solid waste area. Data in the database is compared with the set thresholds. If the particle size is large, the abnormal gas concentration is high, or the solid waste area identified by the image is greater than or equal to the set threshold, the pollution status of the area is marked as "polluted". The three pollution types are recorded as particulate matter type, odor type, and stockpile type, respectively. Otherwise, the pollution status of the area is marked as "unpolluted". The pollution index value A is calculated using normalization, and the pollution risk level value of the i-th region is calculated based on the pollution index value and the duration of pollution. ; Set a low-risk threshold and high risk threshold : like Less than If the risk level of the i-th region is set to level one, then the risk level of the i-th region will be set to level one. like Greater than or equal to and less than or equal to If so, then the risk level of the i-th region is set to level two; like Greater than If so, the risk level of the i-th region is set to level three; The city is divided into a grid model of size m, where nodes represent regions and edges represent adjacency relationships. The pollution status label of each region is embedded into the grid model to form a pollution region model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the big data-based urban sanitation operation optimization method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the big data-based urban sanitation operation optimization method as described in any one of claims 1 to 8.

Images