Method and system for predicting urban road dust accumulation load based on real-time monitoring of construction site

By monitoring dust data in real time at construction sites, using big data analysis and random forest regression models to predict dust accumulation on surrounding roads, the problems of high cost and low frequency of road dust accumulation monitoring in the existing technology are solved, and the construction sites are realized Real-time monitoring of dust and efficient prediction of road dust accumulation meet the society's continuous and comprehensive needs for dust accumulation monitoring.

CN118940933APending Publication Date: 2024-11-12YANGTZE DELTA REGION INST OF TSINGHUA UNIV ZHEJIANG
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202410593399.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-05-14
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing airbound road dust accumulation detection solutions have high cost, low frequency and narrow application range, which cannot meet the society's continuous and comprehensive needs for dust accumulation monitoring, and are especially unable to effectively monitor the dust pollution sources at construction sites.

Method used

Using a prediction method based on real-time monitoring of construction sites, dust data is collected through TSP equipment, big data analysis and random forest regression model are used to predict dust accumulation on surrounding roads, data matching analysis is performed in combination with Pearson's test method and delay method to screen features Indicators and optimal monitoring time difference, build prediction models to achieve real-time monitoring and prediction.

Benefits of technology

It reduces the cost of monitoring road dust accumulation load, increases the monitoring frequency and application range, and realizes real-time continuous monitoring of dust on construction sites and road dust accumulation, providing a new path for environmental governance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940933A_ABST
    Figure CN118940933A_ABST
Patent Text Reader

Abstract

The invention provides an urban road dust accumulation load prediction method based on real-time monitoring of a construction site. The method comprises the following steps: collecting real-time monitoring data of the construction site and road dust accumulation load data; screening to obtain characteristic indexes and data of the characteristic indexes influencing the internal dust accumulation condition of the construction site; finding the optimal monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust accumulation load data to obtain a data combination; inputting the data combination as a training set into a random forest regression analysis model for model construction to obtain a prediction model; and inputting the characteristic indexes and data thereof into a prediction model, and predicting to obtain an urban road dust accumulation load value around the construction site. The prediction model established based on big data analysis can predict the dust accumulation condition of the surrounding road through the flying dust data monitored by the construction site TSP equipment, effectively solves the problems of high cost, long period, low frequency, narrow application range and the like of traditional road dust accumulation load monitoring, and provides a new path for environmental governance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and in particular to a prediction method and a prediction system for urban road dust load based on real-time monitoring of construction sites. Background Art

[0002] Road dust is an important source of fine particulate matter in urban atmosphere. It is dust that enters the ambient air under certain dynamic conditions (wind, motor vehicle rolling, crowd activities, etc.) and is a major public nuisance in cities. From the study of road dust composition, it can be seen that construction dust is an important source of road dust.

[0003] Existing cruise-type road dust detection schemes (such as the schemes disclosed in the Chinese patent publications with publication numbers CN113514373A, CN114264584A, and CN214894715U) mainly detect the concentration of dust raised by vehicles driving on the road, and then reflect the road dust load of the road section according to a pre-calibrated model. On the one hand, the above-mentioned cruise-type road dust detection scheme is limited by the high collection and operation costs. It can only be collected once a week or a month, and the dust collection frequency is low, and continuous monitoring cannot be performed. On the other hand, the cruise-type road dust detection is limited by road conditions and can only be collected on the road. It cannot analyze and monitor key pollution sources of road dust such as construction sites, and cannot meet the increasing needs of society and the public for continuous and comprehensive dust monitoring.

[0004] Therefore, it is necessary to provide a method and system for joint monitoring and prediction of road dust and construction site dust. Summary of the invention

[0005] The present invention provides a method for predicting the dust load of urban roads based on real-time monitoring of construction sites. The method is based on a prediction model established based on big data analysis. It can predict the dust conditions of surrounding roads through the dust data monitored by TSP equipment on the construction site, effectively solving the problems of traditional road dust load monitoring, such as high cost, long cycle, low frequency, and narrow application scope, and provides a new path for environmental governance.

[0006] The specific technical solutions are as follows:

[0007] A method for predicting dust load on urban roads based on real-time monitoring of construction sites, comprising:

[0008] (1) Collect real-time monitoring data of urban construction sites and road dust load data on roads around construction sites;

[0009] (2) Pearson test method was used to conduct pairwise correlation analysis on the real-time monitoring data of the construction site. Based on the correlation between each indicator and the particle concentration indicator in the real-time monitoring data of the construction site, characteristic indicators and their data that affect construction site dust were screened out;

[0010] (3) Matching and analyzing the particle concentration data in the real-time monitoring data of the construction site with the road dust load data collected at different times on the roads around the construction site by the delay method, finding the optimal monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust load data, and obtaining the data combination of the real-time monitoring data of the construction site and the road dust load data under the optimal monitoring time difference;

[0011] (4) using the data combination as a training set and inputting it into a random forest regression analysis model for model construction, thereby obtaining a prediction model that can accurately predict the dust load data of roads around the construction site through real-time monitoring data of the construction site;

[0012] (5) Real-time monitoring of characteristic indicators and data of construction sites around the urban roads to be predicted, inputting the characteristic indicators and data into the prediction model of step (4), and predicting the dust load value of the urban roads around the construction sites.

[0013] Furthermore, in step (1), the indicators of the real-time monitoring data of the construction site include PM 2.5 Concentration value, PM 10 Concentration value, temperature, humidity, wind speed, wind direction and noise; real-time monitoring frequency is once per minute;

[0014] The urban road dust load data is a road dust load value collected and calculated by a road dust traveling vehicle, and the collection frequency is once per minute.

[0015] Furthermore, in step (1), the TSP equipment is used to collect real-time monitoring data of the construction site, and the road dust load data is collected by a traveling vehicle;

[0016] In step (2), the construction site real-time monitoring data and road dust load data of step (1) are first cleaned and then subjected to correlation analysis.

[0017] Furthermore, the data cleaning method includes:

[0018] i) Eliminate null value data in the construction site real-time monitoring data and road dust load data;

[0019] ii) Eliminate road dust load values ​​less than 0.1 and greater than 5;

[0020] iii) Eliminate the road dust load data with vehicle speeds less than 20km / h and greater than 60km / h;

[0021] iv) Eliminate the acceleration of the mobile vehicle ≥ 0.7m / s 2 Road dust load data at that time.

[0022] Furthermore, in step (2), the Pearson test method is used to perform pairwise correlation analysis on the indicators of the real-time monitoring data of the construction site.

[0023] Furthermore, in step (2) and step (5), the characteristic index is PM 2.5 Concentration value, PM 10 Concentration value and humidity. The experiment found that PM 2.5 Value, PM 10 Value and humidity can be used as characteristic indicators, and the prediction results have a high accuracy rate.

[0024] Furthermore, in step (3), the specific method for obtaining the optimal monitoring time difference is:

[0025] (A) In units of hours, the particle concentration data in the real-time monitoring data of the construction site at a certain time node is correlated with the road dust load data around the construction site collected m hours after the time node, and the correlation analysis coefficient R corresponding to different delay hours is obtained. m value;

[0026] The particulate matter concentration data is PM 2.5 Value and PM 10 Value; the road dust load data is the road dust load value collected and calculated by the road dust traveling vehicle;

[0027] According to the formulas shown in formula (1) to (3), R m value;

[0028]

[0029] In formula (1), m is the number of hours that the time node is delayed; PM in real-time monitoring data of a construction site at a certain time point 2.5 The correlation between the particle concentration data and the road dust load data around the construction site collected m hours later at that time point; PM in real-time monitoring data of a construction site at a certain time point 10 The correlation between the particle concentration data and the road dust load data around the construction site collected m hours later at that time point;

[0030]

[0031] In formula (2), n is the data collection time (minute level), n = 1, 2, 3...60 (min); X 2.5n is the PM in the construction site at the nth minute 2.5 Concentration value; is the PM in the construction site corresponding to n minutes 2.5 The hourly average value of the concentration value; Y(60*m+n) is the road dust load value around the construction site at the nth minute after a delay of m hours; is the hourly average value of dust load on roads around the construction site corresponding to a delay of m hours;

[0032]

[0033] Where: n is the data collection time (minute level), n = 1, 2, 3...60 (min); X 10 n is the PM in the construction site at the nth minute 10 Concentration value; is the PM in the construction site corresponding to n minutes 10 Hourly average of concentration values;

[0034] Y(60*m+n) is the road dust load value around the construction site at the nth minute after a delay of m hours; is the hourly average value of dust load on roads around the construction site corresponding to a delay of m hours;

[0035] (B) Based on m R m The values ​​are sorted by size to determine the optimal monitoring time difference for the construction site real-time monitoring data that has the highest correlation with the road dust load data.

[0036] Furthermore, the optimal monitoring time difference of the construction site real-time monitoring data is to delay the time node of the construction site real-time monitoring data by 1 hour.

[0037] Furthermore, the parameters of the random forest regression analysis model are: the number of decision trees is 100; the minimum number of samples for node splitting is 2; the booster type is selected as gbtree; the learning rate is set to 0.1; and the maximum tree depth is set to 6.

[0038] Furthermore, the roads surrounding the construction site are roads within a 1.5km diameter range of the construction site.

[0039] The present invention provides a prediction system for urban road dust load based on real-time monitoring of construction sites, comprising:

[0040] A construction site real-time monitoring unit that collects real-time monitoring data from urban construction sites and transmits the data to the prediction module;

[0041] The road dust navigation unit collects the road dust load data of the roads around the construction site and transmits the data to the prediction module;

[0042] The prediction unit includes: a data receiving module, a data preprocessing module, an analysis module, a model building module and a prediction module;

[0043] A data receiving module receives the real-time monitoring data of the construction site transmitted by the real-time monitoring module of the construction site and the road dust load data transmitted by the road dust navigation module;

[0044] The data preprocessing module obtains the original data of the data receiving unit, performs data preprocessing, and performs pairwise correlation analysis on each indicator in the real-time monitoring data of the construction site. According to the correlation between each indicator and the particle concentration indicator in the real-time monitoring data of the construction site, the characteristic indicators and their data that affect the dust accumulation inside the construction site are screened out;

[0045] The analysis module receives the data processed by the data preprocessing module, performs correlation analysis on the particle concentration data in the real-time monitoring data of the construction site and the road dust load data collected at different times on the roads around the construction site, and finds the best monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust load data;

[0046] The model building module retrieves the data from the data preprocessing module based on the results of the analysis module, forms a data combination of the real-time monitoring data of the construction site and the road dust load data under the optimal monitoring time difference, and inputs the data into the random forest regression analysis model for model building, thereby obtaining a prediction model that can accurately predict the road dust load data around the construction site through the real-time monitoring data of the construction site;

[0047] The prediction module retrieves the characteristic indicators and data of the construction sites around the urban roads to be predicted after being processed by the data preprocessing module, inputs the characteristic indicators and data into the prediction model, and predicts the dust load value of the urban roads around the construction sites.

[0048] The client is a PC that receives the data preprocessed by the data preprocessing module and displays the data visually; or it is a mobile terminal that receives the prediction results of the prediction module and pushes alarm text messages.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] (1) The prediction method and prediction system provided by the present invention are based on a prediction model established based on big data analysis. They can predict the dust accumulation of surrounding roads through the dust data monitored by the TSP equipment on the construction site. This system solves the problems of high cost, long cycle, low frequency and narrow application range of traditional road dust load monitoring, and provides a new path for environmental governance.

[0051] (2) The prediction method provided by the present invention uses a vehicle-mounted rapid cruise monitoring of urban road dust load, and the vehicle monitors the road dust load in real time according to the set monitoring points; the construction site TSP equipment is used to monitor the construction site dust, and the TSP continuously monitors the construction site dust in real time at the set monitoring points; the dust accumulation and dust data collected by the two types of equipment are transmitted to the management platform, and the road dust and construction site dust conditions are calculated through the big data algorithm. In addition, the management platform calculates and organizes the data into reports, which can support export and other services; the historical data of all monitored roads and construction sites can also be viewed through the management platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 This is a schematic diagram of the prediction process of dust load on urban roads based on real-time monitoring of construction sites in Example 1;

[0053] Figure 2 This is the result diagram of using the Pearson test method to screen characteristic indicators and their data in Application Example 1;

[0054] Among them, noise represents noise; wind direction represents wind direction; temp represents temperature; humidity represents humidity; wind speed represents wind speed.

[0055] Figure 3 A trend chart showing the best monitoring time difference found in Application Example 1;

[0056] Figure 4 This is a fitting diagram of the road dust test prediction and true value obtained by using the random forest model in Application Example 1;

[0057] Among them, A is the result of a 1-hour delay in road dust detection time compared with the real-time monitoring data of the construction site; B is the result of a 11-hour delay in road dust detection time compared with the real-time monitoring data of the construction site.

[0058] Figure 5 This is the decision path diagram for the random forest prediction simulation of road dust accumulation (delayed by 1 hour) in Application Example 1.

[0059] Figure 6 This is the fitting diagram of the road dust accumulation (delayed by 1 hour) test prediction and true value obtained by using the XGBOOST model in Application Example 1;

[0060] Among them, A is the result of a 1-hour delay in road dust detection time compared with the real-time monitoring data of the construction site; B is the result of a 11-hour delay in road dust detection time compared with the real-time monitoring data of the construction site. DETAILED DESCRIPTION

[0061] The present invention will be further described below in conjunction with specific embodiments. The following are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto.

[0062] Example 1

[0063] This embodiment provides a method for predicting dust load on urban roads based on real-time monitoring of construction sites, and the specific steps are as follows:

[0064] (1) Collect real-time monitoring data of urban construction sites and road dust load data on roads around construction sites;

[0065] The TSP equipment is used to collect real-time monitoring data of the construction site, and the UAV is used to collect road dust load data; the indicators of the real-time monitoring data of the construction site include PM 2.5 Value, PM 10 Value, temperature, humidity, wind speed, wind direction and noise; the real-time monitoring frequency is once per minute;

[0066] Urban road dust load data refers to the road dust load value collected and calculated by a road dust UAV, with a collection frequency of once per minute. The roads surrounding the construction site are roads within a 1.5km diameter range of the construction site.

[0067] (2) First, the data in step (1) is cleaned, and then the Pearson test method is used to perform a pairwise correlation analysis on the real-time monitoring data of the construction site. According to the correlation between each indicator and the particle concentration indicator in the real-time monitoring data of the construction site, the characteristic indicators and their data that affect the construction dust of the construction site are screened;

[0068] Data cleaning methods include:

[0069] i) Eliminate null value data in the construction site real-time monitoring data and road dust load data;

[0070] ii) Eliminate road dust load values ​​less than 0.1 and greater than 5;

[0071] iii) Eliminate the road dust load data with vehicle speeds less than 20km / h and greater than 60km / h;

[0072] iv) Eliminate the acceleration of the mobile vehicle ≥ 0.7m / s 2 Road dust load data at that time.

[0073] (3) Matching and analyzing the particle concentration data in the real-time monitoring data of the construction site with the road dust load data collected at different times on the roads around the construction site by the delay method, finding the optimal monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust load data, and obtaining the data combination of the real-time monitoring data of the construction site and the road dust load data under the optimal monitoring time difference;

[0074] The specific method to obtain the best monitoring time difference is:

[0075] (A) In units of hours, the particle concentration data in the real-time monitoring data of the construction site at a certain time node is correlated with the road dust load data around the construction site collected m hours after the time node, and the correlation analysis coefficient R corresponding to different delay hours is obtained. m value;

[0076] The particulate matter concentration data is PM 2.5 Value and PM 10 Value; the road dust load data is the road dust load value collected and calculated by the road dust traveling vehicle;

[0077] According to the formulas shown in formula (1) to (3), R m value;

[0078]

[0079] In formula (1), m is the number of hours that the time node is delayed; PM in real-time monitoring data of a construction site at a certain time point 2.5 The correlation between the particle concentration data and the road dust load data around the construction site collected m hours later at that time point; PM in real-time monitoring data of a construction site at a certain time point 10 The correlation between the particle concentration data and the road dust load data around the construction site collected m hours later at that time point;

[0080]

[0081] In formula (2), n is the data collection time (minute level), n = 1, 2, 3...60 (min); X 2.5 n is the PM in the construction site at the nth minute 2.5 Concentration value; is the PM in the construction site corresponding to n minutes 2.5 The hourly average value of the concentration value; Y(60*m+n) is the road dust load value around the construction site at the nth minute after a delay of m hours; is the hourly average value of dust load on roads around the construction site corresponding to a delay of m hours;

[0082]

[0083] Where: n is the data collection time (minute level), n = 1, 2, 3...60 (min); X 10n is the PM in the construction site at the nth minute 10 Concentration value; is the PM in the construction site corresponding to n minutes 10 Hourly average of concentration values;

[0084] Y(60*m+n) is the road dust load value around the construction site at the nth minute after a delay of m hours; is the hourly average value of dust load on roads around the construction site corresponding to a delay of m hours;

[0085] (B) Based on m R m The values ​​are sorted by size to determine the optimal monitoring time difference for the construction site real-time monitoring data that has the highest correlation with the road dust load data.

[0086] (4) The data combination is used as a training set and input into a random forest regression analysis model for model construction, thereby obtaining a prediction model that can accurately predict the dust load data of roads around the construction site through real-time monitoring data of the construction site;

[0087] The parameters of the random forest regression analysis model are: the number of decision trees is 100; the minimum number of samples for node splitting is 2; the booster type is selected as gbtree; the learning rate is set to 0.1; and the maximum tree depth is set to 6.

[0088] (5) Real-time monitoring of characteristic indicators and data of construction sites around the urban roads to be predicted, inputting the characteristic indicators and data into the prediction model of step (4), and predicting the dust load value of the urban roads around the construction sites.

[0089] This embodiment also provides a prediction system for urban road dust load based on real-time monitoring of construction sites, specifically:

[0090] A construction site real-time monitoring unit that collects real-time monitoring data from urban construction sites and transmits the data to the prediction module;

[0091] The road dust navigation unit collects the road dust load data of the roads around the construction site and transmits the data to the prediction module;

[0092] The prediction unit includes: a data receiving module, a data preprocessing module, an analysis module, a model building module and a prediction module;

[0093] A data receiving module receives the real-time monitoring data of the construction site transmitted by the real-time monitoring module of the construction site and the road dust load data transmitted by the road dust navigation module;

[0094] The data preprocessing module obtains the original data of the data receiving unit, performs data preprocessing, and performs pairwise correlation analysis on each indicator in the real-time monitoring data of the construction site. According to the correlation between each indicator and the particle concentration indicator in the real-time monitoring data of the construction site, the characteristic indicators and their data that affect the dust accumulation inside the construction site are screened out;

[0095] The analysis module receives the data processed by the data preprocessing module, performs correlation analysis on the particle concentration data in the real-time monitoring data of the construction site and the road dust load data collected at different times on the roads around the construction site, and finds the best monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust load data;

[0096] The model building module retrieves the data from the data preprocessing module based on the results of the analysis module, forms a data combination of the real-time monitoring data of the construction site and the road dust load data under the optimal monitoring time difference, and inputs the data into the random forest regression analysis model for model building, thereby obtaining a prediction model that can accurately predict the road dust load data around the construction site through the real-time monitoring data of the construction site;

[0097] The prediction module retrieves the characteristic indicators and data of the construction sites around the urban roads to be predicted after being processed by the data preprocessing module, inputs the characteristic indicators and data into the prediction model, and predicts the dust load value of the urban roads around the construction sites.

[0098] The client is a PC that receives the data preprocessed by the data preprocessing module and displays the data visually; or it is a mobile terminal that receives the prediction results of the prediction module and pushes alarm text messages.

[0099] Application Example 1

[0100] Based on the prediction method and prediction system provided in Example 1, this application example provides the following cases:

[0101] 1. Data Source and Data Volume

[0102] In this case, monitoring data from three different regions (Beijing Mingguang Village, Beijing Shuangjing and Zhejiang Jiaxing) are provided for data analysis. The construction site monitoring equipment is used to obtain real-time construction site monitoring data, including PM 2.5 Value, PM 10 Value, temperature, humidity, wind speed, wind direction and noise; the monitoring frequency is once per minute; the road dust data of urban roads around the construction site are collected by a UAV, and the road dust load value is calculated, and the collection frequency is once per minute.

[0103] If the monitoring frequency of real-time construction site monitoring data is changed to once every five minutes, the subsequent modeling results will be poor.

[0104] The amount of data is as follows:

[0105] (A) Beijing Mingguang Village: 22,069 data items were collected by TSP equipment at the construction site, and 21,750 data items were collected by the UAV. The period was from August 20 to August 28.

[0106] (B) Beijing Shuangjing: 26,322 data items were collected by the TSP equipment on site, and 15,678 data items were collected by the UAV, from August 20 to September 15;

[0107] (C) Jiaxing, Zhejiang: 10,561 pieces of data were collected by TSP equipment on the construction site, and 12,683 pieces of data were collected by the UAV. The period was from August 14 to August 28.

[0108] 2. Data Preprocessing

[0109] In order to ensure that the collected data is accurate and usable, data cleaning and processing are required. The rules for cleaning and processing are as follows:

[0110] i) Eliminate the null value data in the measurement data of the road dust traveling vehicle and the construction site monitoring equipment; such data are abnormal conditions in the detection process operation, such as null values, limit values ​​(>5000), etc.

[0111] ii) Eliminate the data with road dust load values ​​less than 0.1 and greater than 5 in the road dust UAV measurement data; such data are a small amount of special value data caused by abnormal environment.

[0112] iii) Eliminate the data with vehicle speed less than 20km / h and greater than 60km / h in the measurement data of the road dust traveling vehicle; such data are eliminated according to the use requirements of the traveling vehicle detection equipment.

[0113] iv) Eliminate acceleration ≥ 0.7m / s in the measurement data of the road dust vehicle 2 The data at that time.

[0114] 3. Finding characteristic indicators

[0115] In order to reduce the number of independent variables, reduce the difficulty of modeling and the amount of calculation, the Pearson test method can be used to screen key characteristic indicators; the data in the real-time construction site monitoring data can be input into the SPSS (Statistical Product and Service Solutions) software to generate a correlation coefficient matrix diagram.

[0116] Specific steps: Input the minute-level data of 7 factors into SPSS software, select the correlation analysis function, then select bivariate analysis, and then set the specific parameters to Pearson (N) correlation and the significance test to two-tailed. Output the correlation coefficient matrix (such as Figure 2 shown).

[0117] Based on the correlation coefficient matrix, the noise, wind direction, wind speed and other factors measured by the construction site monitoring equipment have an impact on PM 2.5 、PM 10 The two indicators have little impact. Therefore, only PM 2.5 、PM 10 And humidity three indicators as independent variables.

[0118] 4. Best monitoring time difference

[0119] In order to improve the accuracy of the correlation between the site monitoring data and the UAV data and increase the accuracy of the next step of modeling, the delayed method is used to correlate the site monitoring data PM 2.5 、PM 10 The correlation coefficient between the value and road dust determines how to conduct correlation analysis between construction site monitoring data and UAV data to best meet the actual situation.

[0120] The specific method is: enter the PM of the construction site equipment in the horizontal axis of the Excel software 2.5 Hourly data (minute data are summed and averaged), enter road dust data on the ordinate (road dust data for the navigation vehicle is entered later than 1 hour), select the scatter plot, then select Add Trend Line, and then display R 2 , this value is the correlation coefficient delayed by 1 hour. Similarly, the vertical axis is input later than the data of 0-12 hours, and the correlation coefficient of 0-12 hours is obtained respectively. Then, the absolute value trend chart of the correlation coefficient is made based on the correlation coefficient of 0-12 hours, such as Figure 3 The results show that the two scenarios of 1 hour delay and 11 hours delay have the best correlation.

[0121] 5. Model Exploration

[0122] 1. The construction site monitoring data and the UAV data are delayed by 1 hour and then the model is built

[0123] Modeling method 1: Random forest regression analysis model

[0124] The data was split into training and test sets in a ratio of 8:2, and a random forest regression analysis model was used (model source: The SPSSAU project (2023). SPSSAU. (Version 23.0) [Online Application Software]. Retrieved from https: / / www.spssau.com), and PM was entered. 2.5 、PM 10 and humidity as independent variables, and the dust load of the UAV (delayed by 1 hour) as the dependent variable.

[0125] Mean square error (MSE): 0.005534355158120109; Coefficient of determination (R^2): 0.9423985795408614.

[0126] The model setting parameters are as follows:

[0127]

[0128]

[0129] Modeling method 2: XGBOOST analysis model

[0130] The data was split into training and test sets in a ratio of 8:2, and the XGBOOST analysis model was used (model source: TheSPSSAU project (2023). SPSSAU. (Version 23.0) [Online Application Software]. Retrieved from https: / / www.spssau.com), and PM was entered. 2.5 、PM 10 and humidity as independent variables, and the dust load of the UAV (delayed by 1 hour) as the dependent variable.

[0131] Root mean square error: 0.10155553783655108; Coefficient of determination (R^2): 0.8926570840774398.

[0132] The model setting parameters are as follows:

[0133] Model Summary Table

[0134]

[0135] 2. The construction site monitoring data and the UAV data are linked after a delay of 11 hours and then modeled

[0136] Modeling method 1: Random forest regression analysis model (model calling and parameter setting are the same as the method with a 1-hour delay)

[0137] The data was split into training and test sets in a ratio of 8:2, and a random forest regression analysis model was used (model source: The SPSSAU project (2023). SPSSAU. (Version 23.0) [Online Application Software]. Retrieved from https: / / www.spssau.com), and PM was entered. 2.5 、PM 10and humidity as independent variables, and the dust load of the UAV (delayed by 11 hours) as the dependent variable.

[0138] Root mean square error: 0.10911855323750522; Determination coefficient (R^2): 0.876073733530974

[0139] Modeling method 2: XGBOOST analysis model (model calling and parameter setting are the same as the method with a 1-hour delay)

[0140] The data was split into training and test sets in a ratio of 8:2, and the XGBOOST analysis model was used (model source: TheSPSSAU project (2023). SPSSAU. (Version 23.0) [Online Application Software]. Retrieved from https: / / www.spssau.com), and PM was entered. 2.5 、PM 10 and humidity as independent variables, and the dust load of the UAV (delayed by 11 hours) as the dependent variable.

[0141] Root mean square error: 0.16588701918365792; Coefficient of determination (R^2): 0.5456768979548839

[0142] Conclusion: Pearson correlation analysis was used for correlation analysis, and 3 characteristic factors were selected from 7 characteristics to support model construction; according to the results of correlation analysis, the road dust load values ​​with a delay of 1 hour and 11 hours had the highest correlation with the PM2.5 and PM10 values ​​of the construction site detection equipment; the multivariate nonlinear regression analysis used two model algorithms, random forest and XGBOOST, based on decision tree ensemble learning; the random forest model with a delay of 1 hour had the highest accuracy of 94%.

Claims

1. A method for predicting dust load on urban roads based on real-time monitoring of construction sites, characterized in that: include: (1) Collect real-time monitoring data of urban construction sites and road dust load data on roads around construction sites; (2) Pearson test method was used to conduct pairwise correlation analysis on the real-time monitoring data of the construction site. Based on the correlation between each indicator and the particle concentration indicator in the real-time monitoring data of the construction site, characteristic indicators and their data that affect construction site dust were screened out; (3) Matching and analyzing the particle concentration data in the real-time monitoring data of the construction site with the road dust load data collected at different times on the roads around the construction site by the delay method, finding the optimal monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust load data, and obtaining the data combination of the real-time monitoring data of the construction site and the road dust load data under the optimal monitoring time difference; (4) using the data combination as a training set and inputting it into a random forest regression analysis model for model construction, thereby obtaining a prediction model that can accurately predict the dust load data of roads around the construction site through real-time monitoring data of the construction site; (5) Real-time monitoring of characteristic indicators and data of construction sites around the urban roads to be predicted, inputting the characteristic indicators and data into the prediction model of step (4), and predicting the dust load value of the urban roads around the construction sites.

2. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 1, characterized in that: In step (1), the indicators of the real-time monitoring data of the construction site include PM 2.5 Concentration value, PM 10 Concentration value, temperature, humidity, wind speed, wind direction and noise; real-time monitoring frequency is once per minute; The urban road dust load data is a road dust load value collected and calculated by a road dust traveling vehicle, and the collection frequency is once per minute.

3. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 1, characterized in that: In step (1), the TSP equipment is used to collect real-time monitoring data of the construction site, and the road dust load data is collected by a road dust traveling vehicle; In step (2), the construction site real-time monitoring data and road dust load data of step (1) are first cleaned and then subjected to correlation analysis.

4. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 3, characterized in that: The data cleaning method includes: i) Eliminate null value data in the construction site real-time monitoring data and road dust load data; ii) Eliminate road dust load values ​​less than 0.1 and greater than 5; iii) Eliminate the road dust load data with vehicle speeds less than 20km / h and greater than 60km / h; iv) Eliminate the acceleration of the mobile vehicle ≥ 0.7m / s 2 Road dust load data at that time.

5. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 1, characterized in that: In step (2) and step (5), the characteristic index is PM 2.5 Concentration value, PM 10 Concentration values ​​and humidity.

6. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 1, characterized in that: In step (3), the specific method for obtaining the optimal monitoring time difference is: (A) In units of hours, the particle concentration data in the real-time monitoring data of the construction site at a certain time node is correlated with the road dust load data around the construction site collected m hours after the time node, and the correlation analysis coefficient R corresponding to different delay hours is obtained. m value; The particulate matter concentration data is PM 2.5 Value and PM 10 Value; the road dust load data is the road dust load value collected and calculated by the road dust traveling vehicle; According to the formulas shown in formula (1) to (3), R m value; In formula (1), m is the number of hours that the time node is delayed; PM in real-time monitoring data of a construction site at a certain time point 2.5 The correlation between the particle concentration data and the road dust load data around the construction site collected m hours later at that time point; PM in real-time monitoring data of a construction site at a certain time point 10 The correlation between the particle concentration data and the road dust load data around the construction site collected m hours later at that time point; In formula (2), n is the data collection time (minute level), n = 1, 2, 3...60 (min); X 2.5 n is the PM in the construction site at the nth minute 2.5 Concentration value; is the PM in the construction site corresponding to n minutes 2.5 The hourly average value of the concentration value; Y(60*m+n) is the road dust load value around the construction site at the nth minute after a delay of m hours; is the hourly average value of dust load on roads around the construction site corresponding to a delay of m hours; Where: n is the data collection time (minute level), n = 1, 2, 3...60 (min); X 10 n is the PM in the construction site at the nth minute 10 Concentration value; is the PM in the construction site corresponding to n minutes 10 Hourly average of concentration values; Y(60*m+n) is the road dust load value around the construction site at the nth minute after a delay of m hours; is the hourly average value of dust load on roads around the construction site corresponding to a delay of m hours; (B) Based on m R m The values ​​are sorted by size to determine the optimal monitoring time difference for the construction site real-time monitoring data that has the highest correlation with the road dust load data.

7. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 6, characterized in that: m=1。 8. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 7, characterized in that: The parameters of the random forest regression analysis model are: the number of decision trees is 100; the minimum number of samples for node splitting is 2; the booster type is selected as gbtree; the learning rate is set to 0.1; and the maximum tree depth is set to 6.

9. The method for predicting dust load on urban roads based on real-time monitoring of construction sites according to claim 1, characterized in that: The roads surrounding the construction site are roads within a 1.5km diameter range of the construction site.

10. A prediction system for urban road dust load based on real-time monitoring of construction sites, characterized in that: include: A construction site real-time monitoring unit that collects real-time monitoring data from urban construction sites and transmits the data to the prediction module; The road dust navigation unit collects the road dust load data of the roads around the construction site and transmits the data to the prediction module; The prediction unit includes: a data receiving module, a data preprocessing module, an analysis module, a model building module and a prediction module; A data receiving module receives the real-time monitoring data of the construction site transmitted by the real-time monitoring module of the construction site and the road dust load data transmitted by the road dust navigation module; The data preprocessing module obtains the original data of the data receiving unit, performs data preprocessing, and performs pairwise correlation analysis on each indicator in the real-time monitoring data of the construction site. According to the correlation between each indicator and the particle concentration indicator in the real-time monitoring data of the construction site, the characteristic indicators and their data that affect the dust accumulation inside the construction site are screened out; The analysis module receives the data processed by the data preprocessing module, performs correlation analysis on the particle concentration data in the real-time monitoring data of the construction site and the road dust load data collected at different times on the roads around the construction site, and finds the best monitoring time difference of the real-time monitoring data of the construction site with the highest correlation with the road dust load data; The model building module retrieves the data from the data preprocessing module based on the results of the analysis module, forms a data combination of the real-time monitoring data of the construction site and the road dust load data under the optimal monitoring time difference, and inputs the data into the random forest regression analysis model for model building, thereby obtaining a prediction model that can accurately predict the road dust load data around the construction site through the real-time monitoring data of the construction site; The prediction module retrieves the characteristic indicators and data of the construction sites around the urban roads to be predicted after being processed by the data preprocessing module, inputs the characteristic indicators and data into the prediction model, and predicts the dust load value of the urban roads around the construction sites; The client is a PC that receives the data preprocessed by the data preprocessing module and displays the data visually; or it is a mobile terminal that receives the prediction results of the prediction module and pushes alarm text messages.

Citation Information

Patent Citations

  • System and method for evaluating road cleanliness

    CN113514373A

  • Vehicle-mounted mobile monitoring system and monitoring method

    CN114264584A

  • Vehicle-mounted urban road dust accumulation load rapid underway monitoring system

    CN214894715U