Method, device, equipment and readable storage medium for detecting air pollutants

Through the deep reinforcement learning algorithm, the atmospheric pollutant detection methods are optimized, and the problems of unreasonable layout of monitoring sites and time-consuming data processing are solved, and the intelligent monitoring site layout and equipment configuration are realized, which improves the efficiency and accuracy of atmospheric pollution monitoring.

CN118568511BActive Publication Date: 2024-11-19LIJIN ECOLOGICAL ENVIRONMENT MONITORING CENT OF DONGYING MUNICIPAL ECOLOGICAL ENVIRONMENT BUREAU
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202410640790.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-22
Publication Date
2024-11-19
Estimated Expiration
2044-05-22

AI Technical Summary

Technical Problem

Among the existing atmospheric pollutant detection methods, the monitoring site layout is unreasonable, resulting in insufficient monitoring blind spots or coverage, the inability to adjust resource allocation in time, affecting the timely response to pollution incidents, and the data processing takes a long time, making decisions unable to be supported in time.

Method used

A deep reinforcement learning algorithm is adopted, combined with historical data and real-time control information to build an optimization model for detection of atmospheric pollutants. Through preprocessing models and dynamic topological space models, the monitoring site layout and equipment configuration are optimized, and the sampling frequency, location and number of equipment are adjusted.

Benefits of technology

It has realized intelligent monitoring site layout and equipment configuration optimization, improved the efficiency and accuracy of air pollution monitoring, and can respond to pollution incidents in a timely manner and reduce resource waste.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118568511B_ABST
    Figure CN118568511B_ABST
Patent Text Reader

Abstract

The present invention provides an atmospheric pollutant detection method, device, equipment and readable storage medium, which relates to the field of atmospheric pollution detection data processing technology, including obtaining real-time data matrices and historical data matrices of all air quality monitoring stations in the area to be detected, the control actions of the history of the air quality monitoring stations, calculating the comprehensive similarity matrix between the air quality monitoring stations based on a similarity measurement algorithm, using a unified manifold approximation and projection algorithm to map the high-dimensional topological structure in the comprehensive similarity matrix to a low-dimensional space, and obtaining historical low-dimensional topological space features, a dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information; the real-time data matrix is ​​input into the atmospheric pollutant detection optimization model, and the real-time control action is output. The present invention improves the efficiency and accuracy of atmospheric pollution monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of air pollution detection data processing technology, and in particular to an air pollutant detection method, device, equipment and a readable storage medium. Background Art

[0002] The following problems exist in the existing methods for detecting atmospheric pollutants: the layout of traditional monitoring sites is not reasonable enough, resulting in blind spots or insufficient monitoring coverage, and the distribution of atmospheric pollutants cannot be fully and effectively monitored; the detection plan cannot be adjusted according to real-time needs, resulting in waste of resources or poor monitoring results; in traditional methods, the processing and analysis of monitoring data are usually time-consuming and cannot provide timely support for decision-making, affecting the timeliness and effectiveness of responding to pollution incidents. Summary of the invention

[0003] The purpose of the present invention is to provide an atmospheric pollutant detection method, device, equipment and readable storage medium to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0004] In a first aspect, the present application provides a method for detecting air pollutants, comprising:

[0005] Acquire first information and second information, wherein the first information includes a real-time data matrix and a historical data matrix of all air quality monitoring stations in the area to be detected, wherein the rows of the data matrix represent air quality monitoring stations, and the columns of the data matrix represent characteristic data, wherein the characteristic data include pollutant concentrations, meteorological conditions, and geographic locations, and the second information includes historical control actions of the air quality monitoring stations, wherein the control actions include adjusting one or more of the sampling frequency, sampling location, and number of sampling devices of the monitoring stations, and both the first information and the second information are time series data;

[0006] An atmospheric pollutant detection optimization model is constructed, wherein the atmospheric pollutant detection optimization model includes a preprocessing model and a dynamic topological space model, wherein:

[0007] The preprocessing model includes calculating the comprehensive similarity matrix between the air quality monitoring sites based on the historical data matrix and the similarity measurement algorithm, and mapping the high-dimensional topological structure in the comprehensive similarity matrix to the low-dimensional space using the unified manifold approximation and projection algorithm to obtain the historical low-dimensional topological space characteristics;

[0008] The dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information;

[0009] The real-time data matrix is ​​input into the atmospheric pollutant detection optimization model, and the real-time control action is output.

[0010] In a second aspect, the present application also provides an atmospheric pollutant detection device, comprising:

[0011] An acquisition module, the acquisition module is used to acquire first information and second information, the first information includes a real-time data matrix and a historical data matrix of all air quality monitoring stations in the area to be detected, the rows of the data matrix represent air quality monitoring stations, the columns of the data matrix represent characteristic data, the characteristic data include pollutant concentrations, meteorological conditions and geographical locations, the second information includes historical control actions of the air quality monitoring stations, the control actions include adjusting one or more of the sampling frequency, sampling location and number of sampling devices of the monitoring stations, and the first information and the second information are both time series data;

[0012] A construction module is used to construct an atmospheric pollutant detection optimization model, wherein the atmospheric pollutant detection optimization model includes a preprocessing model and a dynamic topological space model, wherein:

[0013] The preprocessing model includes calculating the comprehensive similarity matrix between the air quality monitoring sites based on the historical data matrix and the similarity measurement algorithm, and mapping the high-dimensional topological structure in the comprehensive similarity matrix to the low-dimensional space using the unified manifold approximation and projection algorithm to obtain the historical low-dimensional topological space characteristics;

[0014] The dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information;

[0015] An input-output module is used to input the real-time data matrix into the atmospheric pollutant detection optimization model and output the real-time control action.

[0016] In a third aspect, the present application also provides an atmospheric pollutant detection device, comprising:

[0017] Memory for storing computer programs;

[0018] A processor is used to implement the steps of the atmospheric pollutant detection method when executing the computer program.

[0019] In a fourth aspect, the present application further provides a readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method based on atmospheric pollutant detection are implemented.

[0020] The beneficial effects of the present invention are as follows: the present invention can intelligently optimize the layout of monitoring sites and equipment configuration, thereby improving the efficiency and accuracy of air pollution monitoring, through a method, device, equipment and readable storage medium for detecting air pollutants, through a deep reinforcement learning algorithm, combined with historical data and real-time control information.

[0021] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by practicing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0023] Figure 1 Schematic diagram of the process of the atmospheric pollutant detection method described in an embodiment of the present invention;

[0024] Figure 2 It is a logical schematic diagram of the atmospheric pollutant detection method described in an embodiment of the present invention;

[0025] Figure 3 Schematic diagram of the structure of the air pollutant detection device described in an embodiment of the present invention.

[0026] Markings in the figure: 800, air pollutant detection equipment; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0029] Embodiment 1:

[0030] This embodiment provides a method for detecting air pollutants.

[0031] See also Figure 1 and Figure 2 , the figure shows that the method includes step S100, step S200, and step S300. Figure 2 A logical schematic diagram of the atmospheric pollutant detection method described in an embodiment of the present invention is shown.

[0032] Step S100: Acquire first information and second information, wherein the first information includes a real-time data matrix and a historical data matrix of all air quality monitoring stations in the area to be detected, wherein the rows of the data matrix represent air quality monitoring stations, and the columns of the data matrix represent characteristic data, wherein the characteristic data include pollutant concentrations, meteorological conditions, and geographic locations, and the second information includes historical control actions of the air quality monitoring stations, wherein the control actions include adjusting one or more of the sampling frequency, sampling location, and number of sampling devices of the monitoring stations, and both the first information and the second information are time series data;

[0033] Pollutant concentration: Pollutant concentration is one of the important factors affecting air quality. Monitoring and predicting the spread and distribution of atmospheric pollutants requires considering the changes in pollutant concentration.

[0034] Meteorological conditions: Meteorological conditions, such as wind speed, wind direction, temperature, and humidity, have an important impact on the spread and diffusion of atmospheric pollutants. Wind direction and wind speed determine the direction and speed of pollutant transmission, and meteorological factors such as temperature and humidity also affect the diffusion and dilution of pollutants. Therefore, incorporating meteorological conditions into the model can improve the accuracy of atmospheric pollutant transmission.

[0035] Geographic location: Geographic location factors include the latitude and longitude of the monitoring site, altitude and other information, which also affect the spread and distribution of air pollutants. Therefore, considering the geographical location factor can help the model better understand and predict the spread of air pollutants.

[0036] Taking into account factors such as pollutant concentration, meteorological conditions and geographical location, a more comprehensive description of the monitoring and propagation of atmospheric pollutants can be obtained, thus improving the prediction accuracy and practicality of the model.

[0037] Get the historical data matrix, including:

[0038] Obtaining original monitoring data of all air quality monitoring stations in the area to be tested, wherein the original monitoring data includes pollutant concentration, meteorological conditions and geographical location;

[0039] Performing data enhancement on the original monitoring data according to the trained generative adversarial network to obtain enhanced data;

[0040] Performing data cleaning on the enhanced data to obtain cleaned data;

[0041] The data matrix of history is constructed based on the cleaned data.

[0042] The training steps of the trained generative adversarial network are as follows:

[0043] Constructing a generator network and a discriminator network, wherein the generator network receives a random noise vector as input, and also receives the pollutant concentration, meteorological conditions, and geographic location as condition information;

[0044] Constructing a joint loss function, the joint loss function including an adversarial loss and an additional task loss, the adversarial loss is used to train a discriminator network to distinguish generated data from real data, and the additional task loss is used to ensure that the generated data has similar features to the real data;

[0045] Alternately train the generator network and the discriminator network to optimize the joint loss function. In each iteration, the parameters of the discriminator are updated first, and then the parameters of the generator are updated.

[0046] The random noise vector and the conditional information are randomly perturbed during the training process.

[0047] Step S200: constructing an atmospheric pollutant detection optimization model, wherein the atmospheric pollutant detection optimization model includes a preprocessing model and a dynamic topological space model, wherein:

[0048] The preprocessing model includes calculating the comprehensive similarity matrix between the air quality monitoring sites based on the historical data matrix and the similarity measurement algorithm, and mapping the high-dimensional topological structure in the comprehensive similarity matrix to the low-dimensional space using the unified manifold approximation and projection algorithm to obtain the historical low-dimensional topological space characteristics;

[0049] The similarity matrices corresponding to the geographical location data, meteorological data and pollutant concentration data are S geo , S weather and S pollutant The weights of the three given feature data are w geo 、w weather and wpollutant ,satisfy

[0050] w geo +w weather +w pollutant =1

[0051] Comprehensive similarity matrix S combined The calculation formula is:

[0052] S combined =w geo ·S geo +w weather ·S weather +w pollutant ·S pollutant

[0053] The specific calculation steps are as follows:

[0054] Weighted summation: Perform weighted summation on each similarity matrix. According to the given weight conditions, multiply the similarity matrices of the three feature data by the corresponding weights and perform weighted summation.

[0055] Normalization: The obtained comprehensive similarity matrix is ​​normalized to ensure that the similarity value is in the range of [0, 1]. A linear or nonlinear normalization method can be used, such as dividing the similarity value by the maximum value in the matrix so that the maximum value is 1.

[0056] Through the above steps, we can obtain a comprehensive similarity matrix that takes into account geographic location, meteorological data, and pollutant concentration data for subsequent data analysis and application tasks.

[0057] The expression of the unified manifold approximation and projection algorithm is:

[0058]

[0059] N is the number of air quality monitoring stations; D ij is the comprehensive similarity matrix between air quality monitoring site i and air quality monitoring site j; w ij is the weight between air quality monitoring site i and air quality monitoring site j, which can usually be determined based on the distance; α is a parameter that controls the balance between local structure and global structure.

[0060] Wherein, calculating the similarity matrix between the air quality monitoring sites based on the similarity measurement algorithm includes:

[0061] According to the geographical locations in the historical data matrix, a first similarity matrix between the air quality monitoring sites is calculated using a similarity measurement method based on a spatial network, wherein the geographical locations include geographical coordinates of the air quality monitoring sites;

[0062] The geographic coordinates of the following two air quality monitoring stations are:

[0063] Site A: (40.7128°N, 74.0060°W)

[0064] Site B: (34.0522°N, 118.2437°W)

[0065] We first need to calculate the spatial distance between these sites. The commonly used method to calculate the distance between two points on the earth's surface is to use spherical trigonometry. The most commonly used method is to use the spherical cosine theorem. The formula of the spherical cosine theorem is as follows:

[0066] cos(d)=sin(lat1)·sin(lat2)+cos(lat1)·cos(lat2)·cos(lon2-lon1)

[0067] Where d represents the distance between the stations, lat1, lon1 and lat2, lon2 are the latitude and longitude of the air quality monitoring stations respectively;

[0068] Calculate the spatial distance between sites:

[0069] The distance d between site A and site B AB :

[0070] lat1=40.7128°,lon1=-74.0060°

[0071] lat2=34.0522°, lon2=-118.2437°

[0072] Calculate cos(d AB ) and calculate the arc cosine to get the distance dAB. Since the smaller the distance, the more similar the sites are, we can use the following formula to convert the distance into the first similarity value:

[0073] similarity=e -α·d

[0074] Where α is a parameter that controls the decay speed of similarity, usually a positive number. Here we assume α = 0.1.

[0075] Calculating a first similarity matrix: filling the calculated first similarity values ​​into the similarity matrix to obtain a first similarity matrix between monitoring sites.

[0076] According to the pollutant concentrations in the historical data matrix, a second similarity matrix between the air quality monitoring sites is calculated using a similarity measurement method based on Euclidean distance, wherein the pollutant concentrations include pollutant concentrations monitored by the air quality monitoring sites at different times or different locations;

[0077] Take the pollutant concentration data of the following two air quality monitoring stations as an example:

[0078] Site A:C A =(50, 60, 70)

[0079] Site B: C B =(55, 65, 75)

[0080] The data in brackets indicate pollutant concentrations at different time points.

[0081] The formula for calculating the Euclidean distance is as follows:

[0082]

[0083] So we have:

[0084]

[0085] Convert to the second similarity value:

[0086]

[0087] α is a parameter that controls the decay speed of similarity. Here we assume that α = 0.1. Substituting the calculated Euclidean distance into the formula, we can obtain the second similarity value of the pollutant concentration data between air quality monitoring station A and air quality monitoring station B;

[0088] The calculated similarity values ​​are filled into the similarity matrix to obtain a second similarity matrix between the monitoring sites.

[0089] According to the meteorological conditions in the historical data matrix, a third similarity matrix between the air quality monitoring sites is calculated using a similarity measurement method based on Euclidean distance, wherein the meteorological conditions include the temperature, humidity and wind speed monitored by the air quality monitoring sites, and the specific algorithm of the third similarity matrix in this part is the same as the second similarity matrix calculation method;

[0090] Performing weighted summation on the first similarity matrix, the second similarity matrix and the third similarity matrix to obtain the comprehensive similarity matrix;

[0091] Wherein, the high-dimensional topological structure in the comprehensive similarity matrix is ​​mapped to a low-dimensional space using a unified manifold approximation and projection algorithm to obtain low-dimensional space data, including:

[0092] Constructing a neighborhood graph according to the comprehensive similarity matrix, wherein each node in the neighborhood graph represents an air quality monitoring site, and each edge represents the comprehensive similarity matrix between two corresponding air quality monitoring sites;

[0093] Determining neighbor nodes of each of the nodes based on a local neighborhood of a fixed size;

[0094] Constructing a local neighborhood of each node in a high-dimensional space according to all neighboring nodes of each node;

[0095] Optimizing the node positions of the nodes in the low-dimensional space based on a stochastic gradient descent optimization algorithm to obtain low-dimensional topological space features;

[0096] Wherein, based on a local neighborhood of a fixed size, determining the neighbor nodes of each of the nodes comprises:

[0097] Set fixed parameters;

[0098] The calculation formula of the fixed parameter K is as follows:

[0099] K = round(p × N)

[0100] Where N represents the total number of sample points in the data set, p represents the percentage of K in the total number of sample points in the data set, round represents the rounding function, and p is between 5% and 20%. Set the percentage of K value: According to actual needs and experience, set a percentage of K value in the data set size. This percentage can be adjusted according to specific circumstances, and can usually be selected between 5% and 20%.

[0101] For each node, sorting the values ​​of the comprehensive similarity matrix between the node and other nodes to obtain a sorting result;

[0102] In the sorting results, taking nodes with the highest values ​​of the fixed parameter number of the comprehensive similarity matrix as neighbor nodes of the node;

[0103] The dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information.

[0104] The application of the unified manifold approximation and projection algorithm in the atmospheric pollutant detection optimization model has the following benefits:

[0105] Dimensionality reduction: This algorithm can map high-dimensional data to low-dimensional space, thereby achieving dimensionality reduction of data. In atmospheric pollutant monitoring, the characteristic data of monitoring sites contain multiple dimensions, such as pollutant concentration, meteorological conditions, and geographical location. Through dimensionality reduction, the data dimension can be reduced, the complexity of the model can be simplified, and the calculation efficiency can be improved.

[0106] Topological structure preservation: The unified manifold approximation and projection algorithm can maintain the topological structure between data samples and preserve the relative distance relationship between data as much as possible during the mapping process. This is very important for maintaining the local characteristics of data and reducing information loss, and helps maintain the spatial relationship and similarity between monitoring sites in atmospheric pollutant detection.

[0107] Feature extraction: The key features of the data can be better extracted by unifying the low-dimensional spatial features mapped by the manifold approximation and projection algorithm. These features can be used for subsequent model training and optimization, helping to capture important patterns and changing trends in air pollutant monitoring.

[0108] Among them, the dynamic topological space model is constructed and trained based on the deep Q network, including:

[0109] Constructing a historical state representation according to the historical low-dimensional topological space characteristics;

[0110] Constructing historical actions according to the historical control actions;

[0111] Constructing a reward function, wherein the reward function is the difference between the monitoring effect score and the monitoring cost score after taking the current action under the current state representation;

[0112] Use the deep Q network to build the model architecture and obtain the initial model;

[0113] The calculation process of the deep Q network is as follows:

[0114] At each time step, an action is selected based on the current state and receives rewards from the environment. The strategy is then updated to maximize the cumulative reward by interacting with the environment for a certain number of training iterations, where the update formula is:

[0115] Q(s,a)←Q(s,a)+α(r+γmax a′ Q(s′, a′)-Q(s, a))

[0116] Q(s,a) is the Q value of selecting action a in state s; α is the learning rate; r is the immediate reward given by the environment; γ is the discount factor used to balance the importance of current rewards and future rewards; s′ is the next state obtained after performing action a; a′ is the action selected in the next state s′;

[0117] The above reward value represents the immediate feedback obtained by performing a specific action in a specific state. In an environment, an action is selected according to the current state, and then the environment returns a reward value, which represents the immediate reward obtained after the agent performs the action.

[0118] The Q value is used to measure the estimated value of the long-term reward of performing an action in a certain state. Specifically, for a given state and action, the Q value represents the expected value of the cumulative reward that may be obtained in the future, starting from the next state, after performing the action in the current state. It is used to represent the value of each state-action pair, that is, the expected return of taking a specific action in a specific state.

[0119] The initial model is trained according to the historical state representation and the historical action to obtain the dynamic topological space model.

[0120] Wherein, optimizing the position of the node in the low-dimensional space based on the stochastic gradient descent optimization algorithm includes:

[0121] Initializing a node position of each of the nodes in the low-dimensional space;

[0122] A preset loss function is used to measure the difference between the neighborhood structure in the high-dimensional space and the neighborhood structure in the low-dimensional space, and the loss function includes a topology preservation loss and a reconstruction loss;

[0123] Derivative the loss function to obtain the gradient of each node position;

[0124] According to the calculated gradient of each node position, the node position of each node is updated according to the gradient information of the loss function using the stochastic gradient descent method, and the node position is updated multiple times in an iterative manner until the loss function converges or reaches a preset number of iterations.

[0125] The formula of the reward function is:

[0126] R(s,a)=Monitoring effect score(s)-cost score(s)

[0127]

[0128]

[0129] Among them, s is the state representation, a is the action, α is the contribution weight of the monitoring error to the monitoring effect, β is the contribution weight of the monitoring site coverage to the monitoring effect, γ is the contribution weight of the energy consumption to the cost, and δ is the contribution weight of the equipment maintenance cost to the cost. The monitoring error represents the difference between the predicted value and the actual value output by the atmospheric pollutant detection optimization model, the ratio between the number of all the air quality monitoring sites in the area to be detected and the total number of air quality monitoring sites, the energy consumption is the energy consumed by all the air quality monitoring sites in the area to be detected, and the equipment maintenance cost is the equipment maintenance and repair cost of all the air quality monitoring sites in the area to be detected.

[0130] The predicted value and actual value mentioned here are the predicted value of the low-dimensional topological space feature and the actual value of the low-dimensional topological space feature. When calculating the monitoring error, the difference between the predicted value of the low-dimensional topological space feature and the actual value of the low-dimensional topological space feature can be directly calculated, and finally converted into the predicted value of the pollutant concentration. By establishing a mapping relationship between the topological space feature and the pollutant concentration, and combining it with the actual measurement data of the monitoring station for calibration and verification, the corresponding pollutant concentration trend can be evaluated through the predicted low-dimensional topological space feature.

[0131] From this, we can use the reciprocal of the monitoring error and the weighted sum of the air quality monitoring site coverage as components of the monitoring effectiveness score, the reciprocal of energy consumption and the reciprocal of equipment maintenance cost as components of the cost score, and then calculate the reward value based on these scores.

[0132] Step S300: input the real-time data matrix into the atmospheric pollutant detection optimization model, and output the real-time control action.

[0133] Through the output control actions, the sampling frequency, sampling location and number of sampling devices at the air quality monitoring station can be directly guided to be adjusted.

[0134] The advantages of the present invention compared to the prior art are:

[0135] Comprehensive use of multiple data information: The algorithm combines real-time and historical data from air quality monitoring sites, including multiple characteristic information such as pollutant concentrations, meteorological conditions, and geographical locations, which can more comprehensively reflect the complexity and changes of the monitoring environment.

[0136] Dynamic topological space model: By constructing a dynamic topological space model and using deep Q-network for modeling and training, it is possible to achieve intelligent monitoring optimization based on historical low-dimensional topological space characteristics and control actions, thereby improving monitoring efficiency and accuracy.

[0137] Data preprocessing and feature extraction: Using similarity measurement algorithm and unified manifold approximation and projection algorithm to preprocess data and extract features can reduce data dimension and improve the computational efficiency and generalization ability of the model.

[0138] Generate adversarial network to enhance data: By training a generative adversarial network to enhance data, we can expand the dataset and improve the robustness and generalization ability of the model, thereby improving detection effect and accuracy.

[0139] Dynamic optimization model: This algorithm can realize dynamic optimization based on real-time monitoring data and control actions, and timely adjust the sampling strategy and parameter configuration of the monitoring site, so that the monitoring system can adapt to the ever-changing monitoring environment and needs.

[0140] Comprehensive consideration of cost and effect: By designing a reward function and comprehensively considering the trade-off between monitoring effect and cost, efficient resource utilization and performance optimization can be achieved during the operation of the monitoring system.

[0141] Embodiment 2:

[0142] This embodiment provides an atmospheric pollutant detection device, the device comprising:

[0143] An acquisition module, the acquisition module is used to acquire first information and second information, the first information includes a real-time data matrix and a historical data matrix of all air quality monitoring stations in the area to be detected, the rows of the data matrix represent air quality monitoring stations, the columns of the data matrix represent characteristic data, the characteristic data include pollutant concentrations, meteorological conditions and geographical locations, the second information includes historical control actions of the air quality monitoring stations, the control actions include adjusting one or more of the sampling frequency, sampling location and number of sampling devices of the monitoring stations, and the first information and the second information are both time series data;

[0144] A construction module is used to construct an atmospheric pollutant detection optimization model, wherein the atmospheric pollutant detection optimization model includes a preprocessing model and a dynamic topological space model, wherein:

[0145] The preprocessing model includes calculating the comprehensive similarity matrix between the air quality monitoring sites based on the historical data matrix and the similarity measurement algorithm, and mapping the high-dimensional topological structure in the comprehensive similarity matrix to the low-dimensional space using the unified manifold approximation and projection algorithm to obtain the historical low-dimensional topological space characteristics;

[0146] The dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information;

[0147] An input-output module is used to input the real-time data matrix into the atmospheric pollutant detection optimization model and output the real-time control action.

[0148] It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0149] Embodiment 3:

[0150] Corresponding to the above method embodiment, this embodiment further provides an atmospheric pollutant detection device, and the atmospheric pollutant detection device described below and the atmospheric pollutant detection method described above can refer to each other.

[0151] Figure 3 FIG. 8 is a block diagram of an air pollutant detection device 800 according to an exemplary embodiment. Figure 3 As shown, the air pollutant detection device 800 may include: a processor 801 and a memory 802. The air pollutant detection device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0152] Among them, the processor 801 is used to control the overall operation of the air pollutant detection device 800 to complete all or part of the steps in the above-mentioned air pollutant detection method. The memory 802 is used to store various types of data to support the operation of the air pollutant detection device 800, which may include instructions for any application or method used to operate on the air pollutant detection device 800, and application-related data, such as contact data, messages sent and received, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input an audio signal. For example, the audio component may include a microphone, which is used to receive an external audio signal. The received audio signal may be further stored in the memory 802 or sent through the communication component 805. The audio component also includes at least one speaker for outputting an audio signal. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules may be a keyboard, a mouse, a button, etc. These buttons may be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the atmospheric pollutant detection device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G or 4G, or a combination of one or more thereof, so the corresponding communication component 805 may include: Wi-Fi module, Bluetooth module, NFC module.

[0153] In an exemplary embodiment, the atmospheric pollutant detection device 800 can be implemented by one or more application specific integrated circuits (Application Specific Integrated Circuit, referred to as ASIC), digital signal processors (Digital Signal Processor, referred to as DSP), digital signal processing devices (Digital Signal Processing Device, referred to as DSPD), programmable logic devices (Programmable Logic Device, referred to as PLD), field programmable gate arrays (Field Programmable Gate Array, referred to as FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned atmospheric pollutant detection method.

[0154] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned air pollutant detection method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the air pollutant detection device 800 to complete the above-mentioned air pollutant detection method.

[0155] Embodiment 4:

[0156] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the atmospheric pollutant detection method described above can refer to each other.

[0157] A readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the atmospheric pollutant detection method of the above method embodiment.

[0158] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0159] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0160] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for detecting air pollutants, characterized in that: include: Acquire first information and second information, wherein the first information includes a real-time data matrix and a historical data matrix of all air quality monitoring stations in the area to be detected, wherein the rows of the data matrix represent air quality monitoring stations, and the columns of the data matrix represent characteristic data, wherein the characteristic data include pollutant concentrations, meteorological conditions, and geographic locations, and the second information includes historical control actions of the air quality monitoring stations, wherein the control actions include adjusting one or more of the sampling frequency, sampling location, and number of sampling devices of the monitoring stations, and both the first information and the second information are time series data; An atmospheric pollutant detection optimization model is constructed, wherein the atmospheric pollutant detection optimization model includes a preprocessing model and a dynamic topological space model, wherein: The preprocessing model includes calculating the comprehensive similarity matrix between the air quality monitoring sites based on the historical data matrix and the similarity measurement algorithm, and mapping the high-dimensional topological structure in the comprehensive similarity matrix to the low-dimensional space using the unified manifold approximation and projection algorithm to obtain the historical low-dimensional topological space characteristics; The dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information; The real-time data matrix is ​​input into the atmospheric pollutant detection optimization model, and the real-time control action is output.

2. The method for detecting air pollutants according to claim 1, characterized in that ,According to the historical data matrix, the similarity matrix between the air quality monitoring sites is calculated based on a similarity measurement algorithm, including: According to the geographical locations in the historical data matrix, a first similarity matrix between the air quality monitoring sites is calculated using a similarity measurement method based on a spatial network, wherein the geographical locations include geographical coordinates of the air quality monitoring sites; According to the pollutant concentrations in the historical data matrix, a second similarity matrix between the air quality monitoring sites is calculated using a similarity measurement method based on Euclidean distance, wherein the pollutant concentrations include pollutant concentrations monitored by the air quality monitoring sites at different times or different locations; Calculating a third similarity matrix between the air quality monitoring sites using a similarity measurement method based on Euclidean distance according to the meteorological conditions in the historical data matrix, wherein the meteorological conditions include temperature, humidity and wind speed monitored by the air quality monitoring sites; The first similarity matrix, the second similarity matrix and the third similarity matrix are weightedly summed to obtain the comprehensive similarity matrix.

3. The method for detecting air pollutants according to claim 1, characterized in that ,Using unified manifold approximation and projection algorithm, the high-dimensional topological structure in the comprehensive similarity matrix is ​​mapped to a low-dimensional space to obtain low-dimensional space data, including: Constructing a neighborhood graph according to the comprehensive similarity matrix, wherein each node in the neighborhood graph represents an air quality monitoring site, and each edge represents the comprehensive similarity matrix between two corresponding air quality monitoring sites; Determining neighbor nodes of each of the nodes based on a local neighborhood of a fixed size; Constructing a local neighborhood of each node in a high-dimensional space according to all neighboring nodes of each node; The node positions of the nodes in the low-dimensional space are optimized based on a stochastic gradient descent optimization algorithm to obtain low-dimensional topological space features.

4. The method for detecting air pollutants according to claim 1, characterized in that According to the historical low-dimensional topological space characteristics and the second information, a dynamic topological space model is constructed and trained based on a deep Q network, including: Constructing a historical state representation according to the historical low-dimensional topological space characteristics; Constructing historical actions according to the historical control actions; Constructing a reward function, wherein the reward function is the difference between the monitoring effect score and the monitoring cost score after taking the current action under the current state representation; Use the deep Q network to build the model architecture and obtain the initial model; The initial model is trained according to the historical state representation and the historical action to obtain the dynamic topological space model.

5. The method for detecting air pollutants according to claim 4, characterized in that ,The formula of the reward function is: R(s,a)=Monitoring effect score(s)-cost score(s) Among them, s is the state representation, a is the action, α is the contribution weight of the monitoring error to the monitoring effect, β is the contribution weight of the monitoring site coverage to the monitoring effect, γ is the contribution weight of the energy consumption to the cost, and δ is the contribution weight of the equipment maintenance cost to the cost. The monitoring error represents the difference between the predicted value and the actual value output by the atmospheric pollutant detection optimization model, the ratio between the number of all the air quality monitoring sites in the area to be detected and the total number of air quality monitoring sites, the energy consumption is the energy consumed by all the air quality monitoring sites in the area to be detected, and the equipment maintenance cost is the equipment maintenance and repair cost of all the air quality monitoring sites in the area to be detected.

6. The method for detecting air pollutants according to claim 1, characterized in that , get the historical data matrix, including: Obtaining original monitoring data of all air quality monitoring stations in the area to be tested, wherein the original monitoring data includes pollutant concentration, meteorological conditions and geographical location; Performing data enhancement on the original monitoring data according to the trained generative adversarial network to obtain enhanced data; Performing data cleaning on the enhanced data to obtain cleaned data; The data matrix of history is constructed based on the cleaned data.

7. The method for detecting air pollutants according to claim 6, characterized in that ,The training steps of the trained generative adversarial network are as follows: Constructing a generator network and a discriminator network, wherein the generator network receives a random noise vector as input, and also receives the pollutant concentration, meteorological conditions, and geographic location as condition information; Constructing a joint loss function, the joint loss function including an adversarial loss and an additional task loss, the adversarial loss is used to train a discriminator network to distinguish generated data from real data, and the additional task loss is used to ensure that the generated data has similar features to the real data; Alternately train the generator network and the discriminator network to optimize the joint loss function. In each iteration, the parameters of the discriminator are updated first, and then the parameters of the generator are updated. The random noise vector and the conditional information are randomly perturbed during the training process.

8. An atmospheric pollutant detection device, characterized in that: include: An acquisition module, the acquisition module is used to acquire first information and second information, the first information includes a real-time data matrix and a historical data matrix of all air quality monitoring stations in the area to be detected, the rows of the data matrix represent air quality monitoring stations, the columns of the data matrix represent characteristic data, the characteristic data include pollutant concentrations, meteorological conditions and geographical locations, the second information includes historical control actions of the air quality monitoring stations, the control actions include adjusting one or more of the sampling frequency, sampling location and number of sampling devices of the monitoring stations, and the first information and the second information are both time series data; A construction module is used to construct an atmospheric pollutant detection optimization model, wherein the atmospheric pollutant detection optimization model includes a preprocessing model and a dynamic topological space model, wherein: The preprocessing model includes calculating the comprehensive similarity matrix between the air quality monitoring sites based on the historical data matrix and the similarity measurement algorithm, and mapping the high-dimensional topological structure in the comprehensive similarity matrix to the low-dimensional space using the unified manifold approximation and projection algorithm to obtain the historical low-dimensional topological space characteristics; The dynamic topological space model is a model constructed and trained based on a deep Q network according to the historical low-dimensional topological space features and the second information; An input-output module is used to input the real-time data matrix into the atmospheric pollutant detection optimization model and output the real-time control action.

9. An atmospheric pollutant detection device, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of the atmospheric pollutant detection method as claimed in any one of claims 1 to 7 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the atmospheric pollutant detection method according to any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Method and device for establishing gas concentration prediction model

    CN112986492A

  • Atmospheric environmental pollutant prediction model based on dynamic space-time attention mechanism

    CN113326981A

  • Granular pollution prediction method and device based on generative adversarial network

    CN116913409A

  • Air pollution detecting system based on neural network

    CN208255169U

  • Crowdsourced indoor air quality rating system

    US20230408477A1