A method for identifying high-emission heavy-duty diesel vehicles based on multi-source data

Abstract

The present invention discloses a method for identifying high-emission heavy-duty diesel vehicles based on multi-source data, comprising the following steps: S1, constructing an OBD sample data set; S2, performing data grouping and processing the emission data; S3, dynamically characterizing the emissions of heavy-duty diesel vehicles; S4, determining the emission threshold of high-emission vehicles according to the continuous distribution function established in S3; S5, comparing the NOx / CO2 ratio calculated from the remote sensing data with the emission threshold to determine high-emission vehicles. The present invention realizes high-precision identification and classification of heavy-duty vehicles with high NOx emissions, ensures the real-time and comprehensive nature of supervision, realizes automatic data processing, can dynamically update the screening threshold of high-emission heavy-duty vehicles to make it more accurate, and improves the accuracy of screening.

Description

Technical Field

[0001] The present invention relates to the technical field of motor vehicle emission monitoring, and in particular to a method for identifying high-emission heavy-duty diesel vehicles based on multi-source data. Background Art

[0002] Timely screening and maintenance of high-emission heavy-duty diesel vehicles is the core of motor vehicle pollution prevention and control. Remote sensing detection systems and on-board diagnostic systems (OBD) are used to carry out all-weather and all-round monitoring of nitrogen oxides (NOx) emissions from heavy-duty diesel vehicles because of their low cost and ability to provide real-time data.

[0003] However, these two technologies have some significant limitations when monitoring motor vehicle emissions, which are mainly manifested in the following two aspects: First, the coverage is not comprehensive. Due to the failure of relevant sensors in the OBD system that can realize the detection function or failures in the data transmission process, the OBD data on many heavy-duty diesel vehicles are not available, and high-emission vehicles cannot be effectively monitored; second, the screening is inaccurate. The instantaneous emission data of vehicles collected by remote sensing detection equipment has occasional high NOx concentration records due to the influence of the external environment. Using only a single instantaneous emission data to screen high-emission vehicles may cause misidentification. Summary of the invention

[0004] In order to solve the problems of incomplete coverage and inaccurate screening in existing motor vehicle emission monitoring technologies, the present invention proposes a high-emission heavy-duty diesel vehicle identification method based on multi-source data to solve the above problems.

[0005] The present application discloses a method for identifying high-emission heavy-duty diesel vehicles based on multi-source data, comprising the following steps:

[0006] S1. Construct an OBD sample data set, and determine and screen the OBD data quality to obtain OBD data that meets the requirements;

[0007] S2, obtaining GPS data, grouping data according to the GPS data and the OBD data obtained in S1, and processing the emission data;

[0008] S3, Dynamic characterization of heavy-duty diesel vehicle emissions, establishment of NOx / CO 2 The continuous distribution function of the emission ratio is used to describe the NOx emission characteristics of heavy-duty diesel vehicles;

[0009] S4, determining the emission threshold of high-emission vehicles according to the continuous distribution function established in S3;

[0010] S5. Associate the remote sensing data with the OBD data and calculate the NOx / CO2 The ratio is compared with the emission threshold value obtained by S4 to determine the high emission vehicle.

[0011] Preferably, the S1 comprises the following steps:

[0012] S11, dividing travel events according to OBD data;

[0013] S12. Count the missing rate and abnormal rate of OBD data of key fields in each travel event. If the missing rate or abnormal rate of single field data is greater than 30%, the data of this trip will not be used to characterize NOx emission characteristics;

[0014] Missing rate = number of missing data / total number of data;

[0015] Abnormal rate = number of abnormal data / total number of data;

[0016] The key fields include the NOx volume concentration downstream of the SCR, the instantaneous intake mass flow rate, and the volume fuel flow rate, a total of three indicators;

[0017] S13. Use KNN algorithm to process OBD abnormal data and missing data.

[0018] Preferably, the S11 comprises the following steps:

[0019] S111. Classify heavy vehicles according to their gross mass and purpose;

[0020] S112, sorting the bicycle OBD data under each category in chronological order;

[0021] S113, if the time interval between two adjacent data is greater than 10 minutes, it is considered that a parking event occurs;

[0022] S114. Extract travel data between two adjacent parking events in sequence, and define the travel data between the two adjacent parking events as one travel event.

[0023] Preferably, the S13 comprises the following steps:

[0024] S131, using the KNN algorithm to adjust and correct the data that deviates from the normal range by analyzing the normal distribution of adjacent data points;

[0025] S132. Using the KNN algorithm, based on the similarity between data points, the nearest K neighbor data points are selected to achieve effective interpolation of the missing OBD field.

[0026] Preferably, the data grouping in S2 comprises the following steps:

[0027] S21, acceleration completion, finding the OBD data and GPS data with the same timestamp, and obtaining the reading of the speed field of the OBD data accordingly;

[0028] If the time difference between an OBD data and the previous available OBD data is 1s, the acceleration corresponding to the OBD data is:

[0029] a t =v t -v t-1 ;

[0030] Among them, v t is the reading of the speed field in the current OBD data, v t-1 The speed field reading in the previous OBD data;

[0031] If the sampling frequency of the OBD data and its adjacent previous available OBD data is less than 1 Hz, the accurate acceleration cannot be obtained based on the above formula, and the interpolation method based on random forest is used to predict the acceleration;

[0032] S22, operation mode division, select vehicle specific power VSP as a substitute parameter of driving condition, and calculate vehicle specific power VSP using a large number of vehicle operation parameters collected in real time in the OBD system of heavy-duty diesel vehicles. The calculation formula is:

[0033]

[0034] Among them, VSP t is the vehicle specific power at the tth second, a t is the vehicle acceleration at the tth second, v t is the vehicle speed at the tth second, g is the acceleration due to gravity, θ t is the road slope at the tth second, m is the vehicle mass including the load, A is the rolling resistance coefficient, B is the rotational resistance coefficient, and C is the aerodynamic resistance coefficient;

[0035] S23. Divide the speed field in the OBD data into five speed intervals, namely deceleration, idling, low speed, medium speed and high speed. The speed interval division should be considered to cover the entire driving condition and the distribution frequency should be in the same order of magnitude. Divide the VSP at intervals of 2kW / t, and combine the VSP and speed intervals to divide the driving condition range into n microscopic operating modes, namely n VSPbins, and group the OBD data according to the VSP bins.

[0036] Preferably, the processing of the emission data in S2 comprises the following steps:

[0037] Select NOx / CO 2The ratio is used as the unit of measurement for emission limits, and the NOx emission rate and CO are calculated from the basic data such as instantaneous injection volume, intake volume, NOx concentration, fuel density, etc. obtained from the OBD system. 2 Emission rate, then based on NOx / CO 2 The ratio is used to obtain the emission characteristics of heavy-duty diesel vehicles, and the calculation formula is as follows:

[0038]

[0039] q t =q MAF +q FF *ρ;

[0040]

[0041] in, is the instantaneous NOx emission rate, is the NOx volume concentration downstream of SCR, q t is the instantaneous exhaust mass flow rate, q MAF is the instantaneous intake air mass flow rate, q FR is the volume fuel flow rate, ρ is the fuel density, is instantaneous CO 2 Emission rate, β is CO 2 Emission factor.

[0042] Preferably, S3 comprises the following steps:

[0043] The emission characteristics are characterized by using the probability distribution method of discrete random variables. The OBD data are divided into different bins according to their corresponding speed and acceleration. In each VSPbin, the OBD emission data of each detected vehicle is used as a discrete random variable.

[0044] Then the NOx / CO obtained based on S2 2 Ratio analysis of heavy-duty diesel vehicle emission characteristics: Using Gumbel distribution to describe heavy-duty vehicle NOx / CO based on OBD data 2 Emission distribution, the probability density function f(x) and cumulative density function F(x) of the Gumbel distribution are shown below:

[0045]

[0046] Among them, a is the location parameter, representing the emission rate with the highest frequency in the data, and b is the scale parameter, which is used to describe the width of the distribution.

[0047] Preferably, S4 comprises the following steps:

[0048] The 1% of vehicles with the highest emissions are removed from the data set, and the Gumbel distribution is reapplied to the remaining vehicle subset. The goodness of fit between the observed value and the expected value of the Gumbel distribution is calculated, and this process is repeated until only 1% of the vehicles remain. During this process, the vehicle subset when the goodness of fit is maximized is considered to be the vehicle subset with normal emissions, and the eliminated vehicles are considered to be the vehicle subset with high emissions. The emission threshold of high-emission vehicles is determined based on the elimination percentage when the goodness of fit is maximized and the Gumbel distribution function.

[0049] Preferably, S5 comprises the following steps:

[0050] S51. Based on the environmental conditions on the day of data collection, remove abnormal data, that is, remove data that is greatly affected by factors such as air disturbance, wind direction and speed, and humidity, and complete the cleaning and preprocessing of remote sensing data.

[0051] S52, calculating the VSP value according to the instantaneous traffic parameter characteristics (vehicle speed and acceleration) collected by the remote sensing detection system, dividing the instantaneous remote sensing emission data into corresponding VSP bins according to the VSP value and the speed range corresponding to the remote sensing data, and correlating the instantaneous remote sensing emission data with the emission distribution characteristics obtained in S3 in the VSP bin;

[0052] S53, high emission vehicle determination, instantaneous NOx / CO of remote sensing data 2 The ratio is directly derived from the NOx emission concentration and CO 2 The emission concentration is obtained by calculating the NOx / CO ratio based on remote sensing data. 2 The ratio is compared with the emission threshold in the corresponding VSP bin. If the NOx / CO ratio calculated from remote sensing data is 2 If the ratio is higher than the emission threshold, the vehicle collected by remote sensing data is determined to be a high-emission vehicle;

[0053] Calculation of NOx / CO based on remote sensing data 2 The formula is:

[0054]

[0055] in, The NOx emission concentration measured by remote sensing equipment, CO measured by remote sensing equipment 2 Emission concentration, is the molar mass of NOx, For CO 2 The molar mass of .

[0056] Beneficial effects of the present invention:

[0057] (1) Accurate identification and classification: By integrating multiple data sources, the present invention achieves high-precision identification and classification of heavy-duty vehicles with high NOx emissions, which is more accurate than traditional monitoring methods;

[0058] (2) Real-time and comprehensiveness. The present invention can comprehensively utilize the wide coverage capability of the remote sensing detection system and the continuous data sampling and analysis capability of the OBD system to ensure the real-time and comprehensiveness of supervision;

[0059] (3) Data processing automation: Automated data processing methods reduce manual intervention, improve data processing efficiency, and make motor vehicle emission monitoring more feasible;

[0060] (4) Real-time update and feedback: Through the continuous update of OBD data, the present invention can dynamically update the high-emission heavy-duty vehicle screening threshold to make it more accurate, thereby improving the accuracy of screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a flow chart of a method for identifying high-emission heavy-duty diesel vehicles based on multi-source data according to an embodiment of the present invention;

[0062] Figure 2 The figure is a schematic diagram of the automated data processing flow according to an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the objectives, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and examples.

[0064] The present application embodiment discloses a method for identifying high-emission heavy-duty diesel vehicles based on multi-source data, such as Figure 2 and Figure 2 As shown, the following steps are included:

[0065] S1. Construct an OBD sample data set, and determine and screen the OBD data quality to obtain OBD data that meets the requirements.

[0066] S11, dividing travel events according to OBD data;

[0067] S111. Heavy-duty vehicles are classified according to their gross mass and vehicle type: according to gross mass, they can be divided into 12,000kg-18,000kg, 18,000kg-31,000kg, 31,000kg-49,000kg, and >49,000kg. According to different vehicle types, they can be divided into dump trucks, tractors, special-purpose vehicles, trucks, etc.

[0068] S112, sorting the bicycle OBD data under each category in chronological order;

[0069] S113, if the time interval between two adjacent data is greater than 10 minutes, it is considered that a parking event occurs;

[0070] S114. Extract travel data between two adjacent parking events in sequence, and define the travel data between the two adjacent parking events as one travel event.

[0071] S12. Count the missing rate and abnormal rate of OBD data of key fields in each travel event. If the missing rate or abnormal rate of single field data is greater than 30%, the data of this trip will not be used to characterize the NOx emission characteristics. Among them, the data missing rate refers to the ratio of the amount of data with a time interval greater than 1s between two adjacent data to the total amount of data for this travel event. The data abnormality rate refers to the ratio of the amount of data readings of the required key fields that exceed the value range specified in Appendix Q of the "Emission Limits and Measurement Methods of Pollutants for Heavy-Duty Diesel Vehicles (China Phase VI)" to the total amount of data for this travel event. In addition, if any key field does not change in 10 consecutive records, it will also be considered abnormal. The key fields include three indicators: the volume concentration of NOx downstream of the SCR (ppm), the instantaneous intake mass flow rate (kg / h), and the volume fuel flow rate (L / h).

[0072] The missing rate and anomaly rate are calculated as follows:

[0073] Missing rate = number of missing data / total number of data;

[0074] Abnormal rate = number of abnormal data / total number of data.

[0075] S13. In view of the multi-dimensional characteristics of OBD data and potential outlier problems, the KNN algorithm (K nearest neighbor method) is used in this embodiment to process OBD abnormal data and missing data.

[0076] S131. Using the KNN algorithm, by analyzing the normal distribution of adjacent data points, the data that deviates from the normal range is adjusted and corrected, thereby ensuring the accuracy and consistency of the data;

[0077] S132. Using the KNN algorithm, based on the similarity between data points, the nearest K neighbor data points are selected to achieve effective interpolation of the missing OBD field.

[0078] S2, obtain GPS data from the heavy-duty vehicle satellite navigation precision positioning system, group the data according to the GPS data and the OBD data obtained in S1, and process the emission data.

[0079] Data grouping includes the following steps:

[0080] S21, acceleration completion, finding the OBD data and GPS data with the same timestamp, and obtaining the reading of the speed field of the OBD data accordingly;

[0081] If the time difference between an OBD data and the previous available OBD data is 1s, the acceleration corresponding to the OBD data is:

[0082] a t =v t -v t-1 ;

[0083] Among them, v t is the reading of the speed field in the current OBD data, v t-1 is the reading of the speed field in the previous OBD data, v t and v t-1 The unit is m / s, a t The unit is m / s 2 ;

[0084] If the sampling frequency of the OBD data and the previous available OBD data adjacent to it is less than 1 Hz, the accurate acceleration cannot be obtained based on the above formula. In this embodiment, the interpolation method based on random forest is used to predict the acceleration;

[0085] S22, operation mode division, select vehicle specific power (VSP) as a substitute parameter for driving conditions, and use a large number of vehicle operating parameters (speed, acceleration) collected in real time in the OBD system of heavy-duty diesel vehicles to calculate the vehicle specific power VSP. The calculation formula is:

[0086]

[0087] Among them, VSP t is the vehicle specific power at the tth second (kW / ton), a t is the vehicle acceleration at the tth second (m / s 2 ), v t is the vehicle speed at the tth second (m / s), g is the acceleration due to gravity, which is about 9.8m / s 2 ,θ t is the road slope at the tth second (%), m is the vehicle mass including the load (ton), A is the rolling resistance coefficient (kWs / m), B is the rotational resistance coefficient (kWs 2 / m 2 ), C is the aerodynamic drag coefficient (kWs 3 / m 3 ); The road load factor of the heavy-duty diesel vehicle used in this embodiment, that is, and They are 0.0875, 0 and 0.000331 respectively.

[0088] S23. Divide the speed field in the OBD data into five speed intervals, namely, deceleration, idling, low speed, medium speed and high speed. The speed interval division should be considered to cover the entire driving condition and the distribution frequency is in the same order of magnitude. Divide the VSP at intervals of 2kW / t, and divide the driving condition range into n microscopic operating modes, i.e., n VSPbins, by combining the VSP and speed intervals. Group the OBD data according to the VSP bins. In one embodiment, the VSP bin division example is shown in Table 1:

[0089] Table 1 VSP bin division example

[0090]

[0091] Processing of emissions data includes the following steps:

[0092] Select NOx / CO 2 The ratio (quantitative rigid quantity) is used as the unit of measurement for emission limits. The NOx emission rate and CO emission rate need to be calculated based on the basic data such as instantaneous fuel injection volume, intake volume, NOx concentration, fuel density, etc. obtained from the OBD system. 2 Emission rate, then based on NOx / CO 2 The ratio is used to obtain the emission characteristics of heavy-duty diesel vehicles, and the calculation formula is as follows:

[0093]

[0094]

[0095] q t =q MAF +q FR *ρ;

[0096]

[0097] in, is the instantaneous NOx emission rate (g / s), is the NOx volume concentration downstream of SCR (ppm), q t is the instantaneous exhaust mass flow rate (kg / h), q MAF is the instantaneous intake air mass flow rate (kg / h), q FR is the volumetric fuel flow rate (L / h), ρ is the fuel density (kg / L), and the recommended value is 0.84kg / L for diesel. is instantaneous CO 2 Emission rate (g / s), β is CO 2 Emission coefficient, that is, the amount of CO emitted per kg of fuel burned2 , using the recommended value, diesel is 20684kg / L.

[0098] S3. Establish NOx / CO in different micro-operation modes (i.e. VSP bins) 2 The continuous distribution function of the emission ratio is used to finely describe the NOx emission characteristics of heavy-duty diesel vehicles. Dynamic characterization of heavy-duty diesel vehicle emissions, establishment of NOx / CO 2 The continuous distribution function of the emission ratio is used to finely describe the NOx emission characteristics of heavy-duty diesel vehicles.

[0099] The emission characteristics are characterized by using the probability distribution method of discrete random variables. The OBD data is divided into different bins according to its corresponding speed and acceleration. In each VSP bin, each detected vehicle OBD emission data is used as a discrete random variable.

[0100] Then the NOx / CO obtained based on S2 2 Ratio analysis of heavy-duty diesel vehicle emission characteristics: Using Gumbel distribution to describe heavy-duty vehicle NOx / CO based on OBD data 2 Emission distribution, the probability density function f(x) and cumulative density function F(x) of the Gumbel distribution are shown below:

[0101]

[0102] Among them, a is the location parameter, representing the emission rate with the highest frequency in the data, and b is the scale parameter, which is used to describe the width of the distribution.

[0103] S4. Determine the emission threshold of high-emission vehicles according to the continuous distribution function established in S3.

[0104] The 1% of vehicles with the highest emissions are gradually removed from the dataset, and the Gumbel distribution is reapplied to the remaining subset of vehicles. The goodness of fit (R 2 The process is repeated until only 1% of the vehicles remain. In this process, the goodness of fit (R 2 The subset of vehicles when the goodness of fit (value) reaches the maximum is considered to be the subset of vehicles with normal emissions (i.e., "conforming to Gumbel distribution"), while the excluded vehicles (i.e., "not conforming to Gumbel distribution") are considered to be the subset of vehicles with high emissions. The emission threshold of high-emission vehicles is determined based on the exclusion percentage when the goodness of fit is maximum and the Gumbel distribution function.

[0105] S41. Apply Gumbel distribution to the entire data set F 100 ;

[0106] S42. Calculate data set F100 The goodness of fit (R 2 value);

[0107] S43. Cut the entire dataset into 100 parts at each integer percentile, remove the top 1% of vehicles with the highest emissions each time, and apply the Gumbel distribution to the remaining subsets F i , i=99,98,…,1;

[0108] S44. Calculate data set F i (F 99 ,F 98 ,…,F 1 ) and its corresponding Gumbel distribution expected value. 2 value;

[0109] S45, repeat steps 3-4 until there are no vehicles left in the data set;

[0110] S46. Creating Cut Percentiles with R 2 Point plot of values;

[0111] S47, goodness of fit (R 2 The subset when the value of ( ) reaches the maximum is regarded as the normal emission vehicle subset (i.e., “conforming to the Gumbel distribution”), while the excluded vehicles (i.e., “not conforming to the Gumbel distribution”) are regarded as the high emission vehicle subset.

[0112] S5. Associate the remote sensing data with the OBD data and calculate the NOx / CO 2 The ratio is compared with the emission threshold value obtained by S4 to determine the high emission vehicle.

[0113] S51. According to the environmental conditions on the day of data collection, abnormal data is eliminated, that is, data greatly affected by factors such as air disturbance, wind direction and speed, and humidity are eliminated, and the cleaning and preprocessing of remote sensing data are completed;

[0114] S52, calculating the VSP value according to the instantaneous traffic parameter characteristics (vehicle speed and acceleration) collected by the remote sensing detection system, dividing the instantaneous remote sensing emission data into corresponding VSP bins according to the VSP value and the speed range corresponding to the remote sensing data, and correlating the instantaneous remote sensing emission data with the emission distribution characteristics obtained in S3 in the VSP bin;

[0115] S53, high emission vehicle determination, instantaneous NOx / CO of remote sensing data 2 The ratio is directly derived from the NOx emission concentration and CO 2 The emission concentration is obtained by calculating the NOx / CO2 The ratio is compared with the emission threshold in the corresponding VSP bin. If the NOx / CO ratio calculated from remote sensing data is 2 If the ratio is higher than the emission threshold, the vehicle collected by remote sensing data is determined to be a high-emission vehicle;

[0116] Calculation of NOx / CO based on remote sensing data 2 The formula is:

[0117]

[0118] in, NOx emission concentration (ppm) measured by remote sensing equipment, CO measured by remote sensing equipment 2 Emission concentration (%), is the molar mass of NOx, due to the conversion of NO to NO 2 The rate is very fast, so NO 2 The molar mass of is 46 g / Mol. For CO 2 The molar mass of is 44 g / Mol.

[0119] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for identifying high-emission heavy-duty diesel vehicles based on multi-source data, characterized in that: The following steps are involved: S1. Construct an OBD sample data set, and determine and screen the OBD data quality to obtain OBD data that meets the requirements; S2, obtaining GPS data, grouping data according to the GPS data and the OBD data obtained in S1, and processing the emission data; The steps of processing the emission data include: selecting the NOx / CO2 ratio as the unit of measurement for the emission limit, and the NOx / CO2 ratio calculation formula is as follows: q t =q MAF +q FR *ρ; in, is the instantaneous NOx emission rate, is the NOx volume concentration downstream of the SCR, qt is the instantaneous exhaust mass flow rate, q MAF is the instantaneous intake mass flow rate, q FR is the volume fuel flow rate, ρ is the fuel density, is the instantaneous CO2 emission rate, β is the CO2 emission coefficient; S3. Dynamic characterization of heavy-duty diesel vehicle emissions, establishing a continuous distribution function of the NOx / CO2 emission ratio to describe the NOx emission characteristics of heavy-duty diesel vehicles; S4, determining the emission threshold of high-emission vehicles according to the continuous distribution function established in S3; S5. Correlate the remote sensing data with the OBD data, and compare the NOx / CO2 ratio calculated based on the remote sensing data with the emission threshold obtained in S4 to determine the high-emission vehicle.

2. The high-emission heavy-duty diesel vehicle identification method based on multi-source data according to claim 1 is characterized in that: The S1 comprises the following steps: S11, dividing travel events according to OBD data; S12. Count the missing rate and abnormal rate of OBD data of key fields in each travel event. If the missing rate or abnormal rate of single field data is greater than 30%, the data of this trip will not be used to characterize NOx emission characteristics; S13. Use KNN algorithm to process OBD abnormal data and missing data.

3. The high-emission heavy-duty diesel vehicle identification method based on multi-source data according to claim 2 is characterized in that: The S11 comprises the following steps: S111. Classify heavy vehicles according to their gross mass and purpose; S112, sorting the bicycle OBD data under each category in chronological order; S113, if the time interval between two adjacent data is greater than 10 minutes, it is considered that a parking event occurs; S114. Extract travel data between two adjacent parking events in sequence, and define the travel data between the two adjacent parking events as one travel event.

4. The method for identifying high-emission heavy-duty diesel vehicles based on multi-source data according to claim 3 is characterized in that: The S13 comprises the following steps: S131, using the KNN algorithm to adjust and correct the data that deviates from the normal range by analyzing the normal distribution of adjacent data points; S132. Using the KNN algorithm, based on the similarity between data points, the nearest K neighbor data points are selected to achieve effective interpolation of the missing OBD field.

5. The method for identifying high-emission heavy-duty diesel vehicles based on multi-source data according to claim 4 is characterized in that: The data grouping in S2 includes the following steps: S21, acceleration completion, finding the OBD data and GPS data with the same timestamp, and obtaining the reading of the speed field of the OBD data accordingly; If the time difference between an OBD data and the previous available OBD data is 1s, the acceleration corresponding to the OBD data is: and t =in t -v t-1 ; Among them, v t is the reading of the speed field in the current OBD data, v t-1 The speed field reading in the previous OBD data; If the sampling frequency of the OBD data and its adjacent previous available OBD data is less than 1 Hz, the interpolation method based on random forest is used for acceleration prediction; S22, operation mode division, using a large number of vehicle operation parameters collected in real time in the heavy-duty diesel vehicle OBD system to calculate the vehicle specific power VSP, the calculation formula is: Among them, VSP t is the vehicle specific power at the tth second, a t is the vehicle acceleration at the tth second, v t is the vehicle speed at the tth second, g is the acceleration due to gravity, θ t is the road slope at the tth second, m is the vehicle mass including the load, A is the rolling resistance coefficient, B is the rotational resistance coefficient, and C is the aerodynamic resistance coefficient; S23, dividing the speed field in the OBD data into five speed intervals, namely, deceleration, idling, low speed section, medium speed section and high speed section, and dividing the driving condition range into n microscopic operating modes, namely, n VSP bins, by combining the VSP and the speed intervals, and grouping the OBD data according to the VSP bins.

6. The method for identifying high-emission heavy-duty diesel vehicles based on multi-source data according to claim 5 is characterized in that: The S3 comprises the following steps: The OBD data is divided into different bins according to its corresponding speed and acceleration. In each VSP bin, each detected vehicle OBD emission data is treated as a discrete random variable; Then, the emission characteristics of heavy-duty diesel vehicles are analyzed based on the NOx / CO2 ratio obtained by S2: Gumbel distribution is used to describe the NOx / CO2 emission distribution of heavy-duty vehicles based on OBD data. The probability density function f(x) and cumulative density function F(x) of the Gumbel distribution are as follows: Among them, a is the location parameter, representing the emission rate with the highest frequency in the data, and b is the scale parameter, which is used to describe the width of the distribution.

7. The method for identifying high-emission heavy-duty diesel vehicles based on multi-source data according to claim 6 is characterized in that: The S4 comprises the following steps: The 1% of vehicles with the highest emissions are removed from the data set, and the Gumbel distribution is reapplied to the remaining vehicle subset. The goodness of fit between the observed value and the expected value of the Gumbel distribution is calculated, and this process is repeated until only 1% of the vehicles remain. During this process, the vehicle subset when the goodness of fit is maximized is considered to be the vehicle subset with normal emissions, and the eliminated vehicles are considered to be the vehicle subset with high emissions. The emission threshold of high-emission vehicles is determined based on the elimination percentage when the goodness of fit is maximized and the Gumbel distribution function.

8. The method for identifying high-emission heavy-duty diesel vehicles based on multi-source data according to claim 7 is characterized in that: The S5 comprises the following steps: S51. According to the environmental conditions on the day of data collection, abnormal data is eliminated to complete the cleaning and preprocessing of remote sensing data; S52, calculating the VSP value according to the instantaneous traffic parameter characteristics collected by the remote sensing detection system, dividing the instantaneous remote sensing emission data into corresponding VSP bins according to the VSP value and the speed range corresponding to the remote sensing data, and associating the instantaneous remote sensing emission data with the emission distribution characteristics obtained in S3 in the VSP bin; S53, high emission vehicle determination, comparing the NOx / CO2 ratio calculated based on the remote sensing data with the emission threshold in the corresponding VSP bin, if the NOx / CO2 ratio calculated based on the remote sensing data is higher than the emission threshold, the vehicle collected by the remote sensing data is determined to be a high emission vehicle; The formula for calculating NOx / CO2 based on remote sensing data is: in, The NOx emission concentration measured by remote sensing equipment, The CO2 emission concentration measured by remote sensing equipment, is the molar mass of NOx, is the molar mass of CO2.

Images