A method for evaluating urban air quality based on uncertainty theory

By applying the k-means clustering and KNN classification model of uncertain theory in air quality evaluation, the problem of failure to comprehensively consider all pollutant indexes in the existing technology is solved, and a more accurate air quality evaluation is achieved, which is in line with the public's feelings. And better reflect the overall air quality of the city.

CN115541829BActive Publication Date: 2024-11-26NANJING UNIV OF POSTS & TELECOMM
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202211255325.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-10-13
Publication Date
2024-11-26
Estimated Expiration
2042-10-13

AI Technical Summary

Technical Problem

The existing air quality evaluation method only considers the maximum average concentration of individual pollutants, and fails to comprehensively consider the total pollutant index, resulting in the public's actual feeling of inconsistent with the AQI air quality index.

Method used

Using an uncertain theory method, the pollutant concentration data of each monitoring station in the city is converted into linear uncertain variables, and a k-means clustering and KNN classification model is established to comprehensively consider the impact of various pollutants on air quality.

Benefits of technology

By comprehensively considering various pollutant indexes, more accurate air quality evaluation results are provided, which conforms to the actual public's feelings and better reflects the overall air quality of the city.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115541829B_ABST
    Figure CN115541829B_ABST
Patent Text Reader

Abstract

An urban air quality evaluation method based on uncertainty theory converts the data of each monitoring station in the city into linear uncertain variables, and uses the distance between the linear uncertain concentration data of pollutants to establish k-means clustering and KNN classification models, so as to evaluate the air quality of the day, and the evaluation result is more accurate; by introducing the method of uncertainty theory, all pollutant data of each monitoring station are taken into comprehensive consideration, which can effectively evaluate the air quality and better reflect the overall air quality of the city, providing a new method for air quality evaluation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of environmental monitoring, and in particular to an urban air quality evaluation method based on uncertainty theory. Background Art

[0002] With the rapid development of my country's economy and the improvement of people's living standards, environmental issues have received widespread attention, especially air quality issues. The level of air quality can directly affect human life and health, so the relevant research on air quality has profound significance and value, and the evaluation method of air quality is one of the important contents.

[0003] According to the Technical Regulations for Ambient Air Quality Index (AQI) (Trial) (HJ633-2012), air quality research institutions generally adopt the air quality evaluation standard issued by the state in March 2012. AQI is a quantitative description of air quality, which describes the degree of air cleanliness or pollution and its impact on health. There are 6 pollutants monitored: sulfur dioxide, nitrogen dioxide, PM10, PM2.5, carbon monoxide and ozone.

[0004] According to the AQI standards, the actual concentration values ​​of the six pollutants, PM2.5, PM10, SO2, NO2, CO and O3, are converted into their respective air quality indexes IAQIp according to the values ​​specified in Table 1, and the maximum value of IAQIPM2.5, IAQIPM10, IAQISO2, IAQINO2, IAQICO and IAQIO3 is determined as the AQI value. The atmospheric pollution status is further clarified according to the AQI standards, mainly including the current air quality pollution level, color representation, impact on health and recommended measures.

[0005]

[0006]

[0007] However, the concentration of pollutants in a city is not an exact value. The current monitoring method is mainly to set up several air stations in the city and install multi-parameter automatic monitoring in the station for continuous automatic monitoring. When calculating the AQI index, the pollutant concentrations measured by each air station are simply averaged, and the average value is used to represent the overall pollutant concentration in the city. When calculating the AQI index, only the maximum value of the average concentration of individual pollutants is considered as the highest air quality index, rather than comprehensively considering the impact of all pollutant indices on air quality assessment. Therefore, the actual perception of the public is often inconsistent with the AQI air quality index.

[0008] Therefore, the prior art requires an air quality evaluation method that can comprehensively consider the concentrations of various pollutants measured by various monitoring stations in the city, so as to better reflect the overall air quality of the city. Summary of the invention

[0009] In order to evaluate the inaccurate pollutant concentration, the present invention proposes an air quality evaluation method based on uncertainty theory. Through the uncertainty theory, a class of inaccurately observed data is regarded as an uncertain variable, and the distance between the pollutant concentrations is used to establish a k-means clustering and KNN classification model, so as to evaluate the air quality of the day. The concentrations of various pollutants measured by various monitoring stations in the city can be incorporated into the air quality evaluation method that takes into comprehensive consideration to better reflect the overall air quality of the city.

[0010] An urban air quality evaluation method based on uncertainty theory comprises the following steps:

[0011] Step 1, obtaining daily pollutant concentration monitoring data of each monitoring station in a period of the city;

[0012] Step 2: convert the daily pollutant concentration monitoring data of each monitoring station in the city into linear uncertainty concentration data;

[0013] Step 3: Establish a k-means clustering model of urban air quality based on uncertainty theory. First, normalize the linear uncertainty concentration data, then define the cluster center and calculate the distance between the data and the cluster center, and then classify the air quality into good and poor categories by distance. In the classification process, update the cluster center and calculate its mean vector and compare them to achieve iteration.

[0014] Step 4: Use the cross-validation method to establish the urban air quality KNN classification model based on uncertainty theory; first calculate the distance between the uncertainty concentration data in the test set and the data in the training set and sort them accordingly, take out the data and count its categories, and use the category with the most counts as the final classification result. Determine the nearest neighbor value of the KNN classification model by comparing the actual classification results;

[0015] Step 5: Input the data to be evaluated and obtain the air quality classification for the day based on the KNN classification model.

[0016] Furthermore, in step 1, when obtaining daily pollutant concentration monitoring data, the daily pollutant concentrations of each station are included in the statistical scope.

[0017] Furthermore, in step 2, according to the daily concentration of each pollutant at each station, the maximum and minimum concentrations of each pollutant measured at each station are selected as the upper and lower bounds of the uncertain linear variable. There are m types of daily pollutants measured, so that the daily pollutant concentration monitoring data of each monitoring station in the city are converted into a sample set of pollutant linear uncertain concentration data. a and b are the minimum and maximum concentrations of a pollutant measured at each pollutant selection station.

[0018] Furthermore, in step 3, a k-means clustering model of urban air quality is established based on uncertainty theory, and the corresponding categories of good and poor air quality are obtained based on the characteristics of the obtained classification data.

[0019] Furthermore, in step 3, the specific k-means clustering modeling method is:

[0020] Step 3-1, normalize the daily linear uncertainty concentration data;

[0021] Step 3-2, initialize K cluster centers, randomly select K days of urban pollutant linear uncertainty concentration data from the pollutant concentration data D = {V1, V2, ... Vn} within a period as the initial cluster center

[0022] Step 3-3, based on the current cluster center, calculate the distance between the daily pollutant linear uncertainty concentration data and the cluster center;

[0023] The distance d between two linear uncertainty concentration data of pollutants ξ and η is calculated by expectation:

[0024] d(ξ,η)=E[|ξ-η|]

[0025] The distance satisfies the following conditions: d(ξ,η)≥0; d(ξ,η)=0 if and only if ξ=η; d(ξ,η)=d(η,ξ); d(ξ,η)≤2d(ξ,τ)+2d(η,τ), τ represents another uncertain concentration data in the inequality;

[0026] Assume that the pollutant concentration data ξ, η are independent linear uncertain variables, and the uncertainty distribution functions are φ, ψ; then the distance between ξ, η is:

[0027]

[0028] In the formula, M represents the reliability, which is the degree of credibility, and is a value between 0 and 1; α represents the integral variable, which is the value of the distribution function, and is between 0 and 1;

[0029] Calculating Linear Uncertain Variables Distance between:

[0030] The inverse distribution function is φ -1 (α) = (1-α) a1 + ab1;

[0031] The inverse distribution function is ψ -1 (α) = (1-α)a2 + ab2;

[0032] So

[0033] Linear uncertainty concentration data of pollutants on day p and the linear uncertainty concentration data of pollutants on day q Then the distance d between the linear uncertainty concentration data of pollutants in these two days is:

[0034]

[0035] According to the above distance calculation method, the distance between the daily pollutant linear uncertainty concentration data and the center of each cluster is calculated;

[0036] Step 3-4, according to the distance between the daily pollutant linear uncertainty concentration data and the center of each cluster, classify it into the category with the closest distance;

[0037] Step 3-5: Update the cluster center of each cluster according to the reclassification results and calculate the mean vector of the cluster center;

[0038] Step 3-6, compare the cluster centers before and after to see if there is any change; if there is no change, terminate the iteration; otherwise, go to step 3-3;

[0039] Step 3-7, the data classification at the end of the iteration is the final classification result; the air quality category of good or poor is obtained based on the characteristics of the obtained classification data.

[0040] Furthermore, in step 4, the training set and test set are set for the data with the above classification results by cross-validation method; all sample data are evenly divided into s subsets with the same number of samples, one of which is selected as the test set and the others as the training set, and the process is repeated s times until each sample is used as a test set separately; on the basis of cross-validation, the accuracy of the KNN classification model is calculated for different nearest neighbor values ​​t.

[0041] Furthermore, in step 4, the specific KNN classification modeling method is:

[0042] Step 4-1, calculating the distance between the linear uncertainty concentration data of the daily pollutants to be classified in the test set and the n training set data by using the linear uncertainty variable distance calculation method in uncertainty theory;

[0043] Step 4-2, sort the calculated distances from small to large, and record the training set data corresponding to each distance for later use;

[0044] Step 4-3, take out the first t sorted training data and confirm their category attributes;

[0045] Step 4-4, count the categories of the t nearest neighbor training set data;

[0046] In step 4-5, the category with the largest number is the final classification result of the test sample;

[0047] Step 4-6, compare the actual classification results of the test set and select the t value with the highest model classification accuracy as the nearest neighbor value t finally determined by the KNN classification model.

[0048] Furthermore, in step 5, the KNN classification model inputs the concentration values ​​of various pollutants measured by various monitoring stations in the city to obtain the air quality classification of the day.

[0049] The beneficial effects achieved by the present invention are:

[0050] (1) It avoids the problem of the currently widely used AQI air quality evaluation method that only considers the maximum value of the average concentration of individual pollutants as the highest air quality index, rather than comprehensively considering the impact of all pollutant indices on air quality assessment, thus avoiding the situation where the public's actual perception is inconsistent with the AQI air quality index.

[0051] (2) Based on uncertainty theory, this method converts the data from each monitoring station in the city into linear uncertain variables, and uses the distance between the linear uncertain concentration data of pollutants to establish k-means clustering and KNN classification models, so as to evaluate the air quality of the day. The evaluation results are more accurate and in line with the actual feelings of the public.

[0052] (3) This method introduces uncertainty theory and comprehensively considers all pollutant data from each monitoring station, which can effectively evaluate air quality and better reflect the overall air quality of the city, providing a new method for air quality evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of the urban air quality evaluation method described in an embodiment of the present invention.

[0054] Figure 2Schematic diagram of the accuracy of the KNN classification model described in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0056] In order to evaluate the air quality, the present invention proposes an air quality evaluation method based on uncertainty theory, which uses the distance between pollutant concentrations to establish k-means clustering and KNN classification models, so as to evaluate the air quality of the day.

[0057] Step 1: Obtain all pollutant data from 13 air monitoring stations in Nanjing in June 2022 (Liuhe Xiongzhou, Pukou, Shanxi Road, Xuanwu Lake, Ruijin Road, Jiangning Rainbow Bridge, Olympic Sports Center, Xianlin University Town, Zhonghuamen, Maigaoqiao, Lishui Yongyang, Caochangmen, and Gaochun Old Vocational School).

[0058] Step 2: According to the daily pollutant concentrations at each station, the maximum and minimum concentrations of each pollutant measured at each station are selected as the upper and lower bounds of the uncertain linear variable, thereby converting the daily pollutant concentration monitoring data of each monitoring station in Nanjing in June 2022 into pollutant linear uncertainty concentration data D = {V1, V2, … V30}, a and b are the minimum and maximum concentrations of a pollutant measured at each pollutant selection station.

[0059] The data is shown in the following table.

[0060]

[0061]

[0062] Step 3: Establish a k-means clustering model of urban air quality based on uncertainty theory.

[0063] (1) Daily uncertain pollutant concentration data Perform normalization:

[0064]

[0065] Among them, max is the maximum value of all observed data of the pollutant, and min is the minimum value of all observed data of the pollutant, and the normalized uncertain pollutant concentration data is obtained.

[0066] (2) Initialize three cluster centers and randomly select the linear uncertainty concentration data of urban pollutants on June 3, June 21, and June 25 from D as the initial cluster center C = {V3, V21, V25}.

[0067] (3) Based on the current three cluster centers, calculate the distance between the daily pollutant linear uncertainty concentration data and the cluster center.

[0068] (4) According to the distance between the daily pollutant linear uncertainty concentration data and the center of each cluster, it is classified into the category with the closest distance.

[0069] (5) Based on the reclassification results, the cluster center of each cluster is updated and the mean vector of the cluster center is calculated. The calculation here is to find the mean, that is, to find the average value of the upper and lower bounds of the linear uncertain pollutant concentration for each pollutant in each cluster data, and obtain the linear uncertain pollutant concentration data of the new cluster center.

[0070] (6) Compare the cluster centers before and after to see if there is any change. If there is no change, terminate the iteration; otherwise, go to (3) to continue with the subsequent steps.

[0071] (7) The data classification at the end of the iteration is the final classification result.

[0072] Finally, the air quality classification results in Nanjing in June 2022 are as follows.

[0073]

[0074]

[0075] Comparing the concentration characteristics of classification labels 1, 2, and 3, the pollutant concentration of label 3 is significantly lower than that of labels 1 and 2, and the pollutant concentration of label 1 is significantly lower than that of label 2. In addition, a new set of pollutant concentration data is assigned to the minimum value of the corresponding pollutant concentration data, and the cluster label is 3; a new set of pollutant concentrations is assigned to the maximum value of the corresponding pollutant concentration data, and the cluster label is 2. Therefore, the attribute of label 3 is defined as excellent air quality, the attribute of label 1 is defined as good air quality, and the attribute of label 2 is defined as poor air quality.

[0076] Step 4: Use the cross-validation method to establish the urban air quality KNN classification model based on uncertainty theory.

[0077] For the Nanjing air quality data in June 2022 with the above-mentioned classification results, the cross-validation method is used to set the training set and test set; all the sample data are evenly divided into 10 subsets with the same number of samples, one of which is selected as the test set, and the others are used as training sets. This process is repeated 10 times until each sample is used as a test set separately.

[0078] On the basis of cross-validation, the accuracy of the KNN classification model is calculated for different t values. The specific KNN classification modeling method is:

[0079] (1) The distance between the linear uncertainty concentration data of the daily pollutants to be classified in the test set and the n training set data is calculated by using the linear uncertainty variable distance calculation method in uncertainty theory.

[0080] (2) Sort the calculated distances from small to large, and record the training set data corresponding to each distance for later use.

[0081] (3) Take out the first t sorted training data and confirm their category attributes.

[0082] (4) Count the categories of the t nearest neighbor training set data.

[0083] (5) The category with the largest number is the final classification result of the test sample.

[0084] (6) Compare the actual classification results of the test set and select the t value with the highest classification accuracy as the final neighbor value t determined by the KNN classification model.

[0085] By calculating different t values ​​from 1 to 15, we can get the corresponding KNN classification model accuracy as follows: Figure 2 .

[0086] Therefore, t=3 is selected as the number of nearest neighbors of the model.

[0087] Step 5: Input the data to be evaluated and obtain the air quality classification for the day based on the KNN classification model.

[0088] Input the concentration data of each pollutant measured by each monitoring station in Nanjing from August 18 to August 20, 2022. According to the KNN classification model, the air quality classification labels for the three days are obtained as shown in the following table.

[0089]

[0090]

[0091] According to the above classification results, the air quality in Nanjing was poor on August 18, good on August 19, and good on August 20.

[0092] The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiment. Any equivalent modifications or changes made by ordinary technicians in this field based on the contents disclosed by the present invention should be included in the protection scope recorded in the claims.

Claims

1. A method for evaluating urban air quality based on uncertainty theory, characterized in that: The following steps are involved: Step 1, obtaining daily pollutant concentration monitoring data of each monitoring station in a period of the city; Step 2: convert the daily pollutant concentration monitoring data of each monitoring station in the city into linear uncertainty concentration data; In step 2, according to the daily pollutant concentrations of each station, the maximum and minimum concentrations of each pollutant measured at each station are selected as the upper and lower bounds of the uncertain linear variable. There are m types of daily pollutants measured, so the daily pollutant concentration monitoring data of each monitoring station in the city are converted into a sample set of pollutant linear uncertain concentration data a and b are the minimum and maximum concentrations of pollutants measured at each station, respectively; Step 3: Establish a k-means clustering model of urban air quality based on uncertainty theory. First, normalize the linear uncertainty concentration data, then define the cluster center and calculate the distance between the data and the cluster center, and then classify the air quality into good and poor categories by distance. In the classification process, update the cluster center and calculate its mean vector and compare them to achieve iteration. In step 3, a k-means clustering model of urban air quality is established based on uncertainty theory, and the corresponding categories of good and poor air quality are obtained based on the characteristics of the obtained classification data; In step 3, the specific k-means clustering modeling method is: Step 3-1, normalize the daily linear uncertainty concentration data; Step 3-2, initialize K cluster centers, randomly select K days of urban pollutant linear uncertainty concentration data from the pollutant concentration data D = {V1, V2, ... Vn} within a period as the initial cluster center Step 3-3, based on the current cluster center, calculate the distance between the daily pollutant linear uncertainty concentration data and the cluster center; Define two pollutant linear uncertainty concentration data ξ, η, and the distance d is calculated by expectation d(ξ,η)=E[|ξ-η|] The distance satisfies the following conditions: d(ξ,η)≥0; d(ξ,η)=0 if and only if ξ=η; d(ξ,η)=d(η,ξ); d(ξ,η)≤2d(ξ,τ)+2d(η,τ), τ represents another uncertain concentration data in the inequality; Assume that the pollutant concentration data ξ, η are independent linear uncertain variables, and the uncertainty distribution functions are φ, ψ; then the distance between ξ, η is: In the formula, M represents the reliability, that is, the degree of credibility, which is a value between 0 and 1; α represents the integral variable, which is the value of the distribution function and is between 0 and 1; Calculating Linear Uncertain Variables Distance between: The inverse distribution function is ψ -1 (α) = (1-α)a1 + αb1; The inverse distribution function is ψ -1 (α) = (1-α)a2 + αb2; So Linear uncertainty concentration data of pollutants on day p and the linear uncertainty concentration data of pollutants on day q Then the distance d between the linear uncertainty concentration data of pollutants in these two days is: According to the above distance calculation method, the distance between the daily pollutant linear uncertainty concentration data and the center of each cluster is calculated; Step 3-4, according to the distance between the daily pollutant linear uncertainty concentration data and the center of each cluster, classify it into the category with the closest distance; Step 3-5: Update the cluster center of each cluster according to the reclassification results and calculate the mean vector of the cluster center; Step 3-6, compare the cluster centers before and after to see if there is any change; if there is no change, terminate the iteration; otherwise, go to step 3-3; Step 3-7, the data classification at the end of the iteration is the final classification result; the air quality category of good or bad is obtained according to the characteristics of the obtained classification data; Step 4: Use the cross-validation method to establish the urban air quality KNN classification model based on uncertainty theory; first calculate the distance between the uncertainty concentration data in the test set and the data in the training set and sort them accordingly, take out the data and count its categories, and use the category with the most counts as the final classification result. Determine the nearest neighbor value of the KNN classification model by comparing the actual classification results; In step 4, a cross-validation method is used to set the training set and the test set for the data with the above classification results; Divide all sample data into s subsets with the same number of samples, select one of them as the test set and the others as the training set, repeat this process s times until each sample is used as a test set; on the basis of cross-validation, calculate the accuracy of the KNN classification model for different nearest neighbor values ​​t; In step 4, the specific KNN classification modeling method is: Step 4-1, calculating the distance between the linear uncertainty concentration data of the daily pollutants to be classified in the test set and the n training set data by using the linear uncertainty variable distance calculation method in uncertainty theory; Step 4-2, sort the calculated distances from small to large, and record the training set data corresponding to each distance for later use; Step 4-3, take out the first t sorted training data and confirm their category attributes; Step 4-4, count the categories of the t nearest neighbor training set data; In step 4-5, the category with the largest number is the final classification result of the test sample; Step 4-6, compare the actual classification results of the test set and select the t value with the highest model classification accuracy as the nearest neighbor value t finally determined by the KNN classification model; Step 5: Input the data to be evaluated and obtain the air quality classification for the day based on the KNN classification model.

2. The urban air quality evaluation method based on uncertainty theory according to claim 1 is characterized by: In step 1, when obtaining daily pollutant concentration monitoring data, the daily pollutant concentrations of each station are included in the statistical scope.

3. The urban air quality evaluation method based on uncertainty theory according to claim 1 is characterized by: In step 5, the KNN classification model inputs the concentration values ​​of various pollutants measured by each monitoring station in the city to obtain the air quality classification for the day.

Citation Information

Patent Citations

  • Air quality forecasting method and system

    CN112465243A

  • Air quality missing data filling method and system based on multivariate stepwise regression

    CN114240719A

  • GNSS data quality comprehensive evaluation method fusing K-means and KNN

    CN114594507A

  • Method for the dynamic evaluation of the air quality.

    MX2015007322A