Method for collaborative evaluation of accuracy of emission perception device based on remote online obd monitoring

Abstract

The application provides a kind of precision collaborative evaluation method of emission sensing device based on remote online OBD monitoring, relating to vehicle testing technical field, comprising: based on the historical detection data of vehicle multi-element emission sensing device, a multi-objective optimization model is constructed with emission species and vehicle attributes as targets to determine the evaluation vehicle sample;Real-time monitoring of vehicle samples is carried out through multi-element emission sensing device, and multi-element pollutant emission data of the sample is obtained, and the multi-element pollutant emission data is grouped according to the vehicle speed and specific power, and the weighted average absolute error, weighted relative error, comprehensive error index based on environmental factor coupling and multivariate regression coefficient under different groups are calculated respectively and cluster analysis is carried out to obtain the precision evaluation result of multi-element emission sensing device.The application improves the accuracy, comprehensiveness and scientificity of the whole vehicle emission detection and evaluation by increasing the collaborative evaluation method of the detection precision of the same pollutant between different test devices.

Description

Technical Field

[0001] This invention relates to the field of vehicle testing technology, and in particular to a collaborative evaluation method for the accuracy of emission sensing devices based on remote online OBD monitoring. Background Technology

[0002] With the acceleration of global industrialization and the rapid development of the transportation industry, the number of heavy-duty vehicles is constantly increasing. Due to their large load capacity and high power requirements, heavy-duty vehicles are usually equipped with large-displacement engines, which emit a large amount of pollutants during operation, posing a serious threat to the environment and human health.

[0003] Heavy-duty vehicle emission sensing technology plays an indispensable role in the identification and reduction of high-emission vehicles. On one hand, accurate emission sensing technology can precisely identify high-emission vehicles, distinguishing which vehicles' pollutant emissions far exceed standards, thus confirming their status as high-emission vehicles. On the other hand, high-precision emission sensing technology can provide a reliable basis for emission reduction measures. Taking the formulation of emission standards for heavy-duty vehicles as an example, accurate detection data can help regulatory authorities determine reasonable emission limits, prompting vehicle manufacturers to improve engine technology and optimize exhaust treatment systems to meet standard requirements. Simultaneously, for heavy-duty vehicles already on the road, high-precision detection technology can conduct regular monitoring to ensure the normal operation of emission reduction equipment. If abnormally high emissions are detected, timely repairs or replacement can be carried out, thereby effectively reducing the overall emissions of heavy-duty vehicles.

[0004] However, existing methods for sensing emissions from heavy-duty vehicles have several problems. Remote online OBD monitoring relies on the vehicle's own on-board diagnostic system, indirectly inferring emissions by reading data from the engine control unit. However, this is affected by the accuracy of vehicle sensors, ECU algorithms, and data transmission stability, and differs between systems from different manufacturers. The accuracy of directly measuring pollutants requires periodic evaluation. Telemetry uses optical detection methods, which can detect emissions quickly but are greatly affected by environmental factors (weather, equipment installation location and angle, road conditions, etc.), and may lack sensitivity for low-concentration pollutants. While following vehicles can directly obtain accurate emissions data under actual driving conditions, it is costly, and factors such as vehicle spacing and speed matching affect the accuracy of the results, making large-scale monitoring difficult. Furthermore, in practice, multiple detection methods are used independently, lacking a collaborative evaluation system to comprehensively assess the detection accuracy of different methods for different pollutants, and also lacking collaborative evaluation of the detection accuracy of different devices for the same pollutants. This hinders accurate understanding of heavy-duty vehicle emissions and the optimization and improvement of detection equipment. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring.

[0006] This application provides the following technical solution: a collaborative evaluation method for the accuracy of emission sensing devices based on remote online OBD monitoring, comprising:

[0007] Based on historical detection data from vehicle multi-emission sensing devices, a multi-objective optimization model is constructed with emission species and vehicle attributes as objectives. After solving the Pareto front through the multi-objective optimization model, the evaluation vehicle samples are determined according to the diversity index. The multi-emission sensing devices include a remote online OBD monitoring system, telemetry equipment, and vehicle-following test equipment.

[0008] The multi-emission sensing device is used to monitor the test vehicle samples in real time to obtain multi-pollutant emission data of the test vehicles, and the multi-pollutant emission data is grouped according to the vehicle speed and specific power.

[0009] Define the first influence weight of different groups on the evaluation accuracy, and calculate the weighted average absolute error and weighted relative error of the remote online OBD monitoring data in the multi-pollutant emission data according to the first influence weight.

[0010] Define a second influence weight for different environmental factors on the evaluation accuracy. Based on the first influence weight and the second influence weight, calculate the comprehensive error index and multivariate regression coefficient based on the coupling of environmental factors for the telemetry data and vehicle-following test data in the multi-pollutant emission data.

[0011] Cluster analysis is performed using the weighted average absolute error and weighted relative error corresponding to each group as feature vectors, and cluster analysis is performed using the comprehensive error index and / or multivariate regression results corresponding to each group as feature vectors to obtain the accuracy evaluation results of the multivariate emission sensing equipment.

[0012] According to one embodiment of this application, a multi-objective optimization model is constructed based on historical detection data from a vehicle multi-emission sensing device, targeting emission species and vehicle attributes, including:

[0013] Based on the set emission data constraints and vehicle type constraints, emission species objective functions and vehicle type objective functions corresponding to the vehicle multi-emission sensing device are constructed respectively. The vehicles to be evaluated are screened according to the emission species objective functions and the vehicle type objective functions to obtain the evaluation vehicle screening results.

[0014] Based on the emission species objective function and the vehicle type objective function, a comprehensive fitness function is defined for each vehicle. A multi-objective optimization model based on the comprehensive fitness function is constructed. The Pareto front is then solved by filtering the evaluation vehicle selection results through the multi-objective optimization model.

[0015] According to one embodiment of this application, the process of screening vehicles to be evaluated based on the emission species objective function and the vehicle type objective function to obtain evaluation vehicle screening results further includes:

[0016] With the goal of minimizing the overall impact of emission species detected by the multi-emission sensing devices, a first weighting coefficient is defined for each emission sensing device. Based on the first weighting coefficient and the set emission data constraints, an emission species objective function corresponding to the vehicle multi-emission sensing device is constructed. Based on the emission species objective function, the emission species objective function value of each vehicle is calculated. The emission species objective function value is compared with a set threshold to screen the vehicles to be evaluated.

[0017] According to one embodiment of this application, the process of screening vehicles to be evaluated based on the emission species objective function and the vehicle type objective function to obtain evaluation vehicle screening results further includes:

[0018] With the goal of minimizing the average emission stage of vehicles, a second weighting coefficient is defined for each different emission stage of the vehicle. Based on the second weighting coefficient and the set vehicle type constraints, a vehicle type objective function is constructed. Based on the vehicle type objective function, the vehicle type objective function value for each vehicle is calculated. The vehicle type objective function value is compared with a set threshold to screen the vehicles to be evaluated. The vehicle type includes the vehicle's emission stage, vehicle age, vehicle weight classification, and vehicle industry characteristic classification.

[0019] According to one embodiment of this application, the method further includes:

[0020] Based on historical detection data from vehicle multi-emission sensing devices, the average pollutant emissions detected by each emission sensing device for each vehicle are calculated. The average pollutant emissions are then compared with the set corresponding pollutant emission level thresholds to obtain the initial screening results of the vehicles to be evaluated.

[0021] The initial screening results of the evaluated vehicles are then further screened based on the emission species objective function and the vehicle type objective function to obtain the final evaluation vehicle screening results.

[0022] According to one embodiment of this application, the method further includes:

[0023] Based on historical detection data or historical data statistical models of vehicle multi-emission sensing devices, the error probability distribution between pollutant emission measurement values ​​and reference values ​​is determined under different groups corresponding to vehicle speed and specific power. The error probability distribution is used as an accuracy evaluation index, and together with the weighted average absolute error and weighted relative error corresponding to each group, they are used as feature vectors for cluster analysis.

[0024] According to one embodiment of this application, the method further includes:

[0025] After grouping the multi-pollutant emission data according to vehicle speed and power-to-weight ratio, the multi-pollutant emission data is further subdivided into subsets according to meteorological and wind speed conditions to obtain the final grouping results.

[0026] According to one embodiment of this application, the method further includes:

[0027] The accuracy assessment data of the multi-emission sensing device is obtained over a set long period of time. Based on the accuracy assessment data, trend analysis and prediction algorithms are used to perform trend analysis on the accuracy of the multi-emission sensing device and issue an early warning.

[0028] According to one embodiment of this application, the method further includes:

[0029] After acquiring the historical detection data of the vehicle's multi-emission sensing device, and after acquiring the multi-pollutant emission data of the evaluation vehicle, data preprocessing is performed. The data preprocessing process includes data cleaning, data standardization, and data integrity check.

[0030] According to one embodiment of this application, the multi-objective optimization model employs the NSGA-II algorithm.

[0031] Compared to traditional methods, the beneficial effects achieved by at least one of the above-mentioned technical solutions adopted in the embodiments of this specification include at least the following: The embodiments of this invention construct a collaborative evaluation framework that integrates data from various emission sensing technologies and devices, such as remote online OBD monitoring, telemetry, and vehicle-following testing. By integrating test vehicles (containing multiple test devices) and following vehicles to collect data on actual roads and utilizing telemetry stations to acquire telemetry data, a comparative analysis method for the detection accuracy of different pollutants in different detection methods is proposed. Simultaneously, a collaborative evaluation method for the detection accuracy of the same pollutant among different test devices is added, along with an accuracy prediction method, thereby upgrading the collaborative evaluation technology method for the data accuracy of multi-emission sensing technology devices for heavy-duty vehicles. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart illustrating the workflow of the collaborative evaluation method for the accuracy of emission sensing devices based on remote online OBD monitoring, according to an embodiment of the present invention.

[0034] Figure 2 This is a schematic diagram of the operating platform for the collaborative evaluation method of emission sensing device accuracy based on remote online OBD monitoring, according to an embodiment of the present invention.

[0035] Figure 3 This is a statistical diagram illustrating vehicle driving conditions for a collaborative evaluation method of emission sensing device accuracy based on remote online OBD monitoring, according to an embodiment of the present invention. Detailed Implementation

[0036] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0037] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] According to the first embodiment of the present invention, such as Figure 1 As shown, this invention claims protection for a collaborative evaluation method for the accuracy of emission sensing devices based on remote online OBD monitoring, comprising:

[0039] S101. Based on historical detection data from vehicle multi-emission sensing devices, a multi-objective optimization model is constructed with emission species and vehicle attributes as objectives. After solving the Pareto front through the multi-objective optimization model, the evaluation vehicle samples are determined according to the diversity index. The multi-emission sensing devices include a remote online OBD monitoring system, telemetry equipment, and vehicle-following test equipment.

[0040] S102. The multi-emission sensing device is used to monitor the test vehicle sample in real time to obtain the multi-emission data of the test vehicle, and the multi-emission data of the multi-emission pollutants is grouped according to the vehicle speed and specific power.

[0041] S103. Define the first influence weight of different groups on the evaluation accuracy. Based on the first influence weight, calculate the weighted average absolute error and weighted relative error of the remote online OBD monitoring data in the multi-pollutant emission data under different groups respectively.

[0042] S104. Define the second influence weight of different environmental factors on the evaluation accuracy. Based on the first influence weight and the second influence weight, calculate the comprehensive error index and multivariate regression coefficient based on the coupling of environmental factors for the telemetry data and the vehicle test data in the multi-pollutant emission data.

[0043] S105. The weighted average absolute error and weighted relative error corresponding to each group are used as feature vectors for cluster analysis, and the comprehensive error index and / or multivariate regression results corresponding to each group are used as feature vectors for cluster analysis to obtain the accuracy evaluation results of the multivariate emission sensing equipment.

[0044] According to some embodiments of the present invention, in S101, historical detection data from the vehicle's multi-emission sensing device is first acquired. Specifically, the collection of historical detection data from the multi-emission sensing device includes the following:

[0045] 1) Data source integration

[0046] Historical data from heavy-duty vehicle emission sensing technology devices were collected from multiple sources, including remote online OBD monitoring systems, telemetry equipment, and vehicle-following testing equipment. This ensured that the data covered vehicle emissions data under different seasons and road conditions.

[0047] The main emission data obtained by the remote online OBD monitoring system is NOx; the main emission data from telemetry monitoring includes CO2, CO, HC, NO, and opacity; the main emission data from vehicle-to-vehicle testing includes CO2, NOx / NO, and PM2.5. 2.5 PN, black carbon BC.

[0048] 2) Preliminary data processing and analysis

[0049] a. Data standardization

[0050] Since different monitoring systems may have different data units and magnitudes, the data first needs to be standardized. For example, for NOx data, whether it comes from a remote online OBD monitoring system or on-vehicle testing, it should be standardized to use volumetric concentration units (such as ppm) or normalized.

[0051] For opacity in telemetry monitoring, it is a dimensionless proportional value and no unit conversion is required. However, it is necessary to ensure that the opacity data measured by different telemetry devices are within the same order of magnitude (for example, by calibrating the equipment to make the measured values ​​comparable).

[0052] b. Data integrity check

[0053] Check the data integrity of each vehicle across all monitoring systems. If a vehicle has a large amount of missing data in a particular monitoring system (e.g., more than 30% of data points are missing), it should be handled with caution or excluded during the screening process.

[0054] According to some embodiments of the present invention, in S101, based on historical detection data from vehicle multi-emission sensing devices, a multi-objective optimization model is constructed with emission species and vehicle attributes as objectives, including:

[0055] Based on the set emission data constraints and vehicle type constraints, emission species objective functions and vehicle type objective functions corresponding to the vehicle multi-emission sensing device are constructed respectively. The vehicles to be evaluated are screened according to the emission species objective functions and the vehicle type objective functions to obtain the evaluation vehicle screening results.

[0056] Based on the emission species objective function and the vehicle type objective function, a comprehensive fitness function is defined for each vehicle. A multi-objective optimization model based on the comprehensive fitness function is constructed. The Pareto front is then solved by filtering the evaluation vehicle selection results through the multi-objective optimization model.

[0057] According to some embodiments of the present invention, the process of selecting vehicles to be evaluated based on the emission species objective function and the vehicle type objective function to obtain evaluation vehicle selection results further includes:

[0058] With the goal of minimizing the overall impact of emission species detected by the multi-emission sensing devices, a first weight coefficient is defined for each emission sensing device. Based on the first weight coefficient and the set emission data constraints, an emission species objective function corresponding to the vehicle multi-emission sensing device is constructed. Based on the emission species objective function, the emission species objective function value of each vehicle is calculated. The emission species objective function value is compared with a set threshold to screen the vehicles to be evaluated.

[0059] With the goal of minimizing the average emission stage of vehicles, a second weighting coefficient is defined for each different emission stage of the vehicle. Based on the second weighting coefficient and the set vehicle type constraints, a vehicle type objective function is constructed. Based on the vehicle type objective function, the vehicle type objective function value for each vehicle is calculated. The vehicle type objective function value is compared with a set threshold to screen the vehicles to be evaluated. The vehicle type includes the vehicle's emission stage, vehicle age, vehicle weight classification, and vehicle industry characteristic classification.

[0060] According to some embodiments of the present invention, the method further includes: calculating the average pollutant emissions detected by each emission sensing device for each vehicle based on historical detection data of the vehicle's multi-emission sensing devices; comparing the average pollutant emissions with a set corresponding pollutant emission level threshold to obtain the initial screening result of the evaluation vehicles; and performing a secondary screening on the initial screening result of the evaluation vehicles according to the emission species objective function and the vehicle type objective function to obtain the final screening result of the evaluation vehicles.

[0061] In practical implementation, this embodiment mainly includes the following processes:

[0062] (1) First, vehicle samples were initially screened and evaluated based on different pollutant emission levels.

[0063] 1) NOx emission level screening (based on remote online OBD monitoring data)

[0064] Calculate the average NOx concentration for each vehicle. NOx concentration data obtained from a remote online OBD monitoring system is the primary basis.

[0065] Vehicles are sorted according to their arithmetic mean NOx concentration, and then classified into different emission levels based on set thresholds. For example, vehicles with an average NOx concentration below 100 ppm are classified as low-NOx emission vehicles, those between 100-300 ppm as medium-NOx emission vehicles, and those above 300 ppm as high-NOx emission vehicles.

[0066] 2) Screening of CO and HC emission levels (based on telemetry monitoring data)

[0067] For CO and HC emission data, the average emission value for each vehicle is also calculated.

[0068] Taking HC as an example, vehicles with HC concentrations below 10 ppm are classified as low-HC emission vehicles, those between 10-20 ppm as medium-HC emission vehicles, and those above 30 ppm as high-HC emission vehicles. A similar classification method is used for CO, with appropriate thresholds set according to relevant emission standards or research objectives.

[0069] 3) Screening of PM2.5, PN, and BC emission levels (based on vehicle follow-up test data)

[0070] Calculate the average emissions for each vehicle based on the PM2.5, PN, and BC data from the vehicle-following test.

[0071] Taking BC (Brown-Coated Vehicle) as an example, vehicles with an average BC emission value below 0.01 g / kg fuel are classified as low-BC emission vehicles, those between 0.01 and 0.05 g / kg fuel are classified as medium-BC emission vehicles, and those above 0.05 g / kg fuel are classified as high-BC emission vehicles. For PM2.5 and PN (Neural Element), appropriate thresholds are also set based on their characteristics and relevant standards to classify emission levels.

[0072] (2) Then, based on the multi-objective optimization model, vehicles are screened and evaluated a second time.

[0073] 1) Determine the constraints

[0074] a. Emissions data constraints

[0075] For each emission species, reasonable upper and lower limits are set. For example, according to environmental emission standards, the lower limit of x1 (NOx) is 0 (theoretically), and the upper limit is the NOx emission standard value stipulated locally; the lower limit of x2 (HC) is 0, and the upper limit may vary depending on the vehicle type and driving conditions (e.g., for heavy vehicles, the upper limit may be 50 ppm); x3 (BC) also has corresponding upper and lower limit constraints.

[0076] b. Vehicle type constraints

[0077] For emission stages, the range of values ​​is constrained to the existing emission stage standards (e.g., China III to China VI; some China V and China VI vehicles have remote OBD online monitoring; when describing remote OBD online monitoring, models without remote online monitoring can be ignored). Vehicle age constraints can be set based on the actual age range of the monitored vehicles; for example, a lower limit of 0 years (new car) and an upper limit of 20 years (hypothetical). Vehicle weight classification is a discrete value (1, 2, 3), and vehicle industry characteristic classification is also a discrete value (1-6, etc.).

[0078] 2) Determine the objective function

[0079] a. Preliminary screening based on emission species targets

[0080] For each vehicle, an emission species objective function is determined, and its value is calculated. Let the three emission species be x1 (NOx), x2 (HC), and x3 (BC), representing remote online OBD monitoring, telemetry, and on-board testing, respectively. The objective is to minimize the combined impact of these three emission species and find sample vehicles representing different pollutant concentration levels. The objective function can be defined as:

[0081]

[0082] w1, w2, and w3 are weighting coefficients, representing the relative importance of different emitting species.

[0083] These weighting coefficients can be determined based on factors such as environmental policies and health impacts. For example, if NOx has a significant impact on local air quality, w1 can be set relatively large.

[0084] Set an initial threshold T1 (which can be determined based on historical data or preliminary research). If the Z1 value is less than T1, the vehicle is initially selected as a possible sample vehicle.

[0085] b. Preliminary screening based on vehicle type-related objectives

[0086] Objective functions can also be constructed for factors such as emission stage, vehicle age, vehicle weight, and vehicle industry characteristics. For example, for emission stage, an objective function can be defined to minimize the average emission stage of the vehicle (assuming that a higher emission stage corresponds to lower emissions). For vehicle i, the objective function value related to its vehicle type is calculated. Let the emission stage be y. 1i The vehicle's age is y 2i Vehicle weight is classified as y 3i (Light-duty = 1, Medium-duty = 2, Heavy-duty = 3), vehicle industry characteristic classification is y 4i (Bus = 1, Freight truck = 2, Construction vehicle = 3, Dump truck = 4, Sanitation vehicle = 5, Passenger bus = 6, etc.)

[0087] Objective functions can be constructed as follows:

[0088]

[0089] Calculate the Z2 value for each vehicle; where v1, v2, v3, and v4 are the corresponding weighting coefficients. These coefficients can be determined based on the research focus; for example, if the research focus is on the impact of emission stages on emissions, v1 can be set to a larger value.

[0090] Set an initial threshold T2 (which can be determined based on the research focus and existing data). If the Z2 value meets certain conditions (e.g., less than T2, or within a pre-defined interval), then the vehicle initially meets the requirements for vehicle type-related targets.

[0091] 3) Model Solving

[0092] a. Fitness calculation adjustment

[0093] The multi-objective optimization model described in this embodiment employs the NSGA-II algorithm. When using the NSGA-II algorithm, a fitness value is calculated for each vehicle. The calculation of the fitness value requires comprehensive consideration of both the emission species objective function Z1 and the vehicle type-related objective function Z2.

[0094] Define a comprehensive fitness function:

[0095]

[0096] Here, α is a weighting coefficient (0 < α < 1) used to balance the importance of emission species targets and vehicle type-related targets in fitness calculation.

[0097] Vehicles are ranked based on their overall fitness value, which serves as the basis for selecting operations in the NSGA-II algorithm.

[0098] b. Sample selection considerations in crossover and mutation operations

[0099] In the crossover operation of the NSGA-II algorithm, when selecting parent vehicles to generate offspring, in addition to considering conventional methods such as random selection and fitness ratio selection, the diversity of vehicle types should also be considered.

[0100] For example, a rule could be set such that, within a certain number of cross-operations, a certain proportion of cross-operations must be between vehicles of different emission stages, vehicle ages, vehicle weights, and vehicle industry characteristics.

[0101] In mutation operations, when a certain feature of a vehicle (such as emission stage, vehicle age, or other features that can be represented by coding) is mutated, it must be ensured that the mutated vehicle still meets all the constraints, and that the diversity of the mutated vehicle in the overall sample increases or at least does not decrease.

[0102] c. Algorithm for finalizing the evaluation vehicle sample

[0103] Screening based on multi-objective optimization results:

[0104] After obtaining the Pareto front solution using the NSGA-II algorithm, the vehicles on the Pareto front are further screened according to the research objectives.

[0105] If the research focuses on the impact of new technologies on emissions, vehicles on the Pareto frontier are classified according to whether they employ specific emission control technologies, and vehicles that adopt new control technologies and have good emission performance are selected as typical evaluation vehicle samples.

[0106] Final determination under diversity maintenance:

[0107] When selecting typical test vehicle samples, a diversity index was established to ensure coverage of different emission stages, vehicle ages, vehicle weights, and vehicle industry characteristics.

[0108] For example, for emission stage diversity, the ratio of the number of emission stages covered by the selected vehicle to the total number of emission stages (e.g., China III to China VI, a total of 4 stages) can be calculated; similar diversity indicators can also be established for vehicle age, vehicle weight, and vehicle industry characteristics.

[0109] Based on these diversity indicators, the initially selected typical test vehicles are adjusted. If a certain diversity indicator is lower than the preset threshold, vehicles of the corresponding type are selected to supplement the selection until all diversity indicators meet the requirements.

[0110] According to some embodiments of the present invention, in S102, before the process of real-time monitoring of the evaluation vehicle sample by the multi-emission sensing device, the detection equipment is first integrated, which mainly includes the following:

[0111] (1) Equipment integration and installation of the vehicle under test

[0112] 1) OBD online monitoring data acquisition equipment

[0113] It connects directly to the OBD interface of the vehicle under test to collect operating and emission data such as engine speed, vehicle speed, and NOx concentration, and indirectly obtains exhaust emission-related data using the vehicle's own diagnostic system.

[0114] 2) PEMS testing equipment

[0115] This device accurately measures the emissions of carbon dioxide (CO2), carbon monoxide (CO), hydrocarbons (HC), nitrogen oxides (NOx), particulate matter number (PN), and particulate matter mass (PM) from the exhaust of the tested vehicle. It can dynamically monitor emissions during actual vehicle operation, adapting to different driving conditions such as acceleration, deceleration, and idling, thus comprehensively and accurately reflecting the emission characteristics of the vehicle in real-world operation.

[0116] 3) Opaque smoke meter

[0117] For diesel vehicles, the smoke opacity is determined by measuring the degree to which the exhaust gases are opaque to light, thus reflecting the particulate matter emissions from diesel vehicle exhaust. This detection method is direct and effective, enabling a rapid assessment of the particulate matter emission levels of diesel vehicles.

[0118] 4) Black Carbon Analyzer

[0119] Optical detection technology is used to specifically detect the black carbon content in the exhaust of tested vehicles, providing crucial data for accurately assessing the impact of vehicle emissions on air quality. Black carbon is a significant pollutant produced by the incomplete combustion of fuels in vehicles, and the detection of black carbon by a black carbon analyzer helps in-depth research into the environmental impact of vehicle emissions.

[0120] The sampling tubes of the aforementioned PEMS, opacity meter, and black carbon meter are connected to the diesel vehicle exhaust pipe. Typically, a sampling cylinder is installed to fix the sampling head of different devices.

[0121] (2) Accompanying vehicles and equipment

[0122] A separate vehicle will be assigned to conduct follow-up tests on the vehicle being tested. This follow-up vehicle will be equipped with a gas analyzer (CO2, NOx / NO), a black carbon analyzer, a particulate matter number (PN) meter, and a PM2.5 meter. 2.5 Testing equipment, along with related sampling and dilution systems, etc. Vehicle-following testing is used to simultaneously detect the vehicle under test while it is in motion. Vehicle-following testing can be conducted in various driving scenarios, such as urban roads and highways, ensuring that the acquired test data is comprehensive and representative.

[0123] (3) Principles for selecting telemetry sites and other factors

[0124] To obtain as much comparative data as possible between telemetry stations and other detection equipment, and to reflect relevant influencing factors, the selection of telemetry stations should, in principle, ensure that the routes include different types of roads such as suburban roads, urban arterial roads, secondary arterial roads, and highways (if applicable), to cover telemetry stations under different traffic flows, speed limits, and road conditions. In terms of regional coverage, it involves different areas such as city centers, urban-rural fringe areas, and suburbs.

[0125] Meanwhile, since telemetry and vehicle-following tests are greatly affected by the environment (weather conditions, road traffic, etc.), this embodiment proposes a principle based on the diversity of influencing factors:

[0126] 1) Vehicle speed

[0127] Based on vehicle speed data, roads are divided into different speed ranges (e.g., low speed range 0-30 km / h, medium speed range 30-60 km / h, high speed range above 60 km / h). During route search, road segments containing different speed ranges are prioritized. For each speed range, a telemetry station is selected on the road within that range. For example, telemetry stations are set up during vehicle acceleration (the transition from low to medium speed), during stable medium-speed driving, and during high-speed driving.

[0128] 2) Ambient temperature

[0129] In route planning, roads that reflect the impact of high temperatures on telemetry stations during sunny, high-temperature periods are selected, such as sections of roads in cities with many asphalt surfaces and little greenery to provide shade. For cloudy days with relatively stable temperatures, a road located in the city's central business district is selected to study the working status of telemetry stations under stable temperatures. On rainy days, a road with different drainage conditions (such as good drainage and easy water accumulation) is selected to study the combined impact of rainwater and temperature changes on telemetry stations.

[0130] 3) Wind conditions

[0131] Based on wind data, the city is divided into different wind zones (e.g., strong wind zone, moderate wind zone, weak wind zone). Route planning should aim to traverse different wind zones. Within each wind zone, telemetry stations are set up on representative road sections with relatively stable wind directions. For example, in a strong wind zone, open suburban road sections with a single wind direction are selected; in a moderate wind zone, road sections between tall buildings in the city are chosen.

[0132] After obtaining multi-pollutant emission data from the test vehicle, such as PEMS test unit (including CO2, CO, HC, NOx, PN, PM), remote online monitoring (NOx), telemetry monitoring (CO2, CO, HC, NO, opacity), and vehicle-to-vehicle testing (CO2, NOx / NO, PM), the system will perform tests. 2.5 The system collects pollutant data (NOx, PN, and black carbon BC) and performs data preprocessing. Remote online OBD monitoring primarily acquires internal vehicle operating parameters such as engine speed, coolant temperature, and fuel system status; these data are related to the vehicle's electronic control unit (ECU). The data preprocessing stage mainly involves data cleaning and standardization. For example, invalid NOx data points (such as outliers due to sensor malfunctions) are removed, and data from different formats is converted to a unified format for subsequent analysis.

[0133] In addition to acquiring basic vehicle driving parameters (such as speed and acceleration), telemetry and vehicle-following monitoring also require recording meteorological and wind speed data. For telemetry, meteorological data (such as temperature, humidity, and wind speed) can be obtained by setting up meteorological sensors near the monitoring point; for vehicle-following monitoring, portable meteorological instruments can be installed on the test vehicle to obtain real-time meteorological and wind speed data.

[0134] In the data preprocessing stage, in addition to the routine cleaning of driving parameters (such as removing outliers in vehicle speed), it is also necessary to process meteorological and wind speed data. For example, wind speed data may need to be classified according to wind direction (such as tailwind, headwind, crosswind), and meteorological data (such as temperature and humidity) may need to be normalized so that they can be reasonably weighted and calculated with other data in subsequent accuracy evaluation.

[0135] After acquiring data from various detection methods (PEMS testing, remote OBD online monitoring, telemetry, and vehicle-following testing), the vehicle speed (v) and specific power (VSP) data are analyzed. Vehicle speed data can be obtained from the vehicle's speed sensor, and specific power (VSP) can be calculated based on the vehicle's speed and acceleration using the following formula:

[0136]

[0137] Where: a is the vehicle acceleration, and θ is the road gradient.

[0138] PEMS testing, online OBD data, vehicle-to-vehicle testing, and telemetry data are integrated into a single dataset. Since these data sources differ, and the tested vehicles and road conditions may vary, it's crucial to ensure that each data point includes basic information such as vehicle identification, test time, and test location (road position) for subsequent correlation and analysis.

[0139] Check the format of data from different sources. For example, emission data may be expressed in different units. The unit of NOx in the OBD monitoring system may be different from the unit in the vehicle-following test. All emission data need to be unified to the same unit or normalized.

[0140] Next, in S102, the emission data of the multiple pollutants are grouped according to the vehicle speed and specific power.

[0141] v-VSP-based grouping and data classification: The collected data are grouped according to vehicle speed v and specific power VSP. For remote online OBD monitoring data, v-VSP grouping is performed according to Table 1.

[0142] Table 1: v-VSP Operating Condition Grouping

[0143]

[0144] Furthermore, for vehicle-following monitoring and telemetry data, in addition to grouping according to v-VSP, meteorological and wind speed conditions also need to be considered. For example, under each v-VSP group, the data subsets are further subdivided according to meteorological conditions (such as humidity range: low humidity, relative humidity between 0% and 30%; medium humidity, relative humidity between 30% and 60%; high humidity, relative humidity between 60% and 100%) and wind speed range (such as low wind speed <2m / s, medium wind speed 2~5m / s, high wind speed >5m / s) to obtain the final grouping results.

[0145] According to some embodiments of the present invention, in S103 and S104, the method for calculating the accuracy evaluation index mainly includes the following:

[0146] (1) Remote online OBD monitoring

[0147] 1) Weighted average absolute error (WMAE) and weighted relative error (WRE)

[0148] Based on the calculation of mean absolute error (MAE) and relative error (RE), the influence weights of different v-VSP groupings on overall accuracy are considered. For example, the weights are determined according to the probability distribution of actual vehicle driving in different v-VSP intervals. Let w i Let x be the weight of the i-th v-VSP group. ijFor remote online OBD monitoring of the pollutant measurement value at the j-th data point under the i-th v-VSP group, y ij For comparative testing (such as PEMS testing), the reference value is the contaminant value at the j-th data point in the i-th v-VSP group, n i Let be the number of data points in the i-th group.

[0149] Weighted average absolute error (WMAE):

[0150]

[0151] Weighted relative error (WRE):

[0152]

[0153] 2) Error interval assessment based on probability distribution

[0154] To further improve the accuracy of the assessment, this embodiment determines the error probability distribution between the pollutant emission measurement value and the reference value under different groups corresponding to vehicle speed and specific power based on the historical detection data or historical data statistical model of the vehicle multi-emission sensing device. The error probability distribution is used as the accuracy assessment index, and together with the weighted average absolute error and weighted relative error corresponding to each group, they are used as feature vectors for cluster analysis.

[0155] For example, suppose that under a certain v-VSP grouping, the error follows a normal distribution N(μ,σ). 2 ), where μ is the mean error and σ is the standard deviation. The acceptable error interval is defined as [μ-kσ, μ+kσ], where k is a coefficient determined according to actual needs (e.g., k=2 represents a 95% confidence interval). By calculating whether the error of each data point is within this interval, the proportion of data points within the acceptable interval is statistically analyzed, serving as a supplementary indicator for accuracy assessment.

[0156] (2) Vehicle monitoring and telemetry

[0157] 1) Comprehensive Error Index (CEI) of Environmental Factor Coupling

[0158] Since vehicle-mounted monitoring and telemetry need to consider the coupling relationship between v-VSP, weather, and wind speed, a comprehensive error index (CEI) is defined. Let x ijkl For vehicle-mounted monitoring or remote sensing, the pollutant measurement value at the l-th data point under the i-th v-VSP group, the j-th meteorological condition (e.g., humidity range classification), and the k-th wind speed condition (e.g., wind speed range classification) is y. ijkl To compare the pollutant value of the reference value at the l-th data point under the corresponding conditions, m ijkl The number of data points in each subdivided data subset.

[0159] First, calculate the mean absolute error (MAE) in the subset:

[0160]

[0161] Then, the weights are determined based on the degree of influence of different environmental factors on the error. For example, suppose that through comparative test data analysis, it is found that the weight of v-VSP on the error in the monitoring of a certain pollutant is w. v-VSP =0.4, the influence weight of meteorological factors is w hu =0.3, the influence weight of wind speed factor is w win =0.3.

[0162] Comprehensive Error Index (CEI):

[0163]

[0164] 2) Multivariate regression analysis to assess the trend of accuracy

[0165] The measured value x ijkl As the dependent variable, v-VSP, meteorological factors (such as humidity), and wind speed are used as independent variables in a multivariate regression analysis. For example, for carbon monoxide (CO) measurements in telemetry data, a regression model is established:

[0166]

[0167] Where β0, β1, β2, β3, and β4 are regression coefficients, and ε is the error term.

[0168] By analyzing the magnitude and significance of the regression coefficients, the influence trend of different factors on measurement accuracy can be assessed. If β1 (the coefficient corresponding to vehicle speed) is large and significant, it indicates that vehicle speed has a significant impact on CO measurement accuracy; if β3 (the coefficient corresponding to temperature) is not significant, it indicates that temperature has a relatively small impact on CO measurement accuracy.

[0169] According to some embodiments of the present invention, in S105, the data accuracy classification and evaluation process based on cluster analysis mainly includes the following:

[0170] (1) Data clustering

[0171] 1) Clustering of remote online OBD monitoring data:

[0172] When performing data clustering, the first step is to determine appropriate feature vectors. For remote online OBD monitoring, select accuracy metrics (such as WMAE and WRE) under different v-VSP groupings. For example, vehicles can be finely grouped according to speed (v) and power-to-weight ratio (VSP) (as shown in Table 1 above).

[0173] Then, the WMAE and WRE values ​​calculated for each group are used as elements of the feature vector. Clustering algorithms (such as K-Means clustering) are then used to cluster these data points. Assuming K=3 is chosen, the data is clustered into three classes: high-precision, medium-precision, and low-precision. The basic principle of the K-Means clustering algorithm is to iteratively minimize the sum of the distances from each data point to its cluster center. In this process, the algorithm assigns data points to different clusters based on their feature vectors.

[0174] When analyzing the accuracy categories of remote online OBD monitoring, the characteristics of the corresponding v-VSP ranges for high, medium, and low accuracy categories obtained by clustering are studied in depth. For example, in high-accuracy clustering, it may be found that most data points come from the medium vehicle speed and medium power ratio group. This may be because under this operating condition, the vehicle engine runs relatively stably, and the OBD monitoring equipment can measure pollutant emission values ​​more accurately.

[0175] In low-precision clustering, a large number of data points may be found in the high-vehicle-speed, high-specific-power group. Further analysis may reveal that at high vehicle speeds and high specific power, the engine combustion process is more complex, resulting in greater variations in the types and concentrations of pollutants produced. Additionally, under these conditions, the sensors in OBD monitoring devices may face issues such as insufficient response time and limited measurement range, leading to decreased accuracy.

[0176] 2) Clustering of vehicle-following monitoring and telemetry data:

[0177] For vehicle-mounted monitoring and telemetry, the Comprehensive Error Index (CEI) or multivariate regression analysis results under different v-VSP-meteorological-wind speed subdivisions are used as feature vectors. For example, for nitrogen oxide (NOx) monitoring in vehicle-mounted monitoring, the CEI values ​​calculated under different meteorological conditions (e.g., humidity range classification: low humidity, medium humidity, high humidity), wind speed conditions (e.g., low wind speed, medium wind speed, high wind speed), and v-VSP groupings are used as part of the feature vector. If multivariate regression analysis results are also considered, such as regression coefficients, they can also be added to the feature vector.

[0178] Clustering algorithms (such as K-Means clustering) are also used to cluster these data points. The purpose of clustering is to group data points with similar precision characteristics into one class for subsequent analysis.

[0179] Analysis of vehicle-following monitoring and telemetry accuracy categories:

[0180] For the clustering results of vehicle-following monitoring and telemetry, the characteristics of the v-VSP-weather-wind speed combination corresponding to each accuracy category are analyzed. For example, in low-accuracy clustering, a large proportion of data points may be found in the high temperature, high wind speed, high vehicle speed, and high specific power group. This is because under this combination, the vehicle-following monitoring equipment may be affected by the high temperature environment, causing changes in sensor performance; high wind speed may interfere with the sampling process; and the complexity of vehicle emissions increases under high vehicle speed and high specific power. These factors combined lead to a decrease in accuracy.

[0181] (2) Dynamic accuracy assessment, trend analysis and prediction

[0182] 1) Real-time data updates and evaluation

[0183] In remote online OBD monitoring, new v-VSP data points are constantly generated during vehicle operation. For example, new values ​​for vehicle speed and specific power are generated every second. Accuracy metrics (such as WMAE and WRE) need to be calculated in real time for these newly generated data points.

[0184] During the calculation, the number of data points (n) in the v-VSP group to which the new data point belongs is updated. i ) and measured values ​​(x) ij ) and reference value (y) ij The sum of the values ​​is then calculated in real time according to the formulas for weighted average absolute error (WMAE) and weighted relative error (WRE).

[0185] Suppose that the vehicle enters a new v-VSP group during its operation. Previously, this group had 0 data points. As the vehicle operates under this condition, the number of data points gradually increases, and the accuracy indicators will be continuously updated based on the new data.

[0186] For vehicle-following monitoring and telemetry, new data points under new v-VSP-weather-wind speed combinations are constantly generated as the vehicle moves. For example, the vehicle may encounter different weather conditions (such as changes in humidity and wind speed) on different road sections, while its speed and specific power also change.

[0187] For these new data points, accuracy metrics such as the Comprehensive Error Index (CEI) need to be updated in real time. When calculating the CEI, the number of data points (m) in each subset is updated based on the subdivision criteria to which the new data point belongs. ijkl ) and measured values ​​(x) ijkl ) and reference value (y) ijkl The sum of ) is then calculated in real time according to the CEI calculation formula.

[0188] For example, when a vehicle monitoring another vehicle travels from a high humidity, low wind speed area to a low humidity, high wind speed area, the relevant accuracy indicators will be updated in real time as weather and vehicle speed conditions change.

[0189] According to some embodiments of the present invention, the method of the present invention further includes: acquiring accuracy assessment data of the multi-emission sensing device over a set long period of time, and based on the accuracy assessment data, performing trend analysis and prediction algorithms to analyze the accuracy of the multi-emission sensing device and issue an early warning.

[0190] In practice, remote online OBD monitoring is taken as an example.

[0191] Collect long-term accuracy assessment data for remote online OBD monitoring, such as data from multiple comparative tests over one or more years, and perform data preprocessing, including data cleaning and smoothing. For different v-VSP groups, analyze the long-term trends of their accuracy metrics (such as WMAE and WRE).

[0192] Trends can be visually observed by plotting time series graphs. For example, for the low vehicle speed, low specific power group, observe the changes in WMAE and WRE values ​​each month. If an upward trend in WMAE values ​​is observed in winter, it may be due to unstable engine operation caused by factors such as cold starts in winter, thus affecting the accuracy of OBD monitoring equipment. Based on long-term trend analysis results, potential future accuracy problems can be predicted. For example, if the accuracy index of a certain v-VSP group shows a downward trend year by year over the past few years, it can be predicted that the accuracy may further decline in the future without intervention, thus allowing for early maintenance of the monitoring equipment or optimization of the algorithm.

[0193] 1) Trend Analysis Algorithm

[0194] Trend analysis algorithms include linear regression analysis:

[0195] Perform linear regression analysis on the smoothed data, with time as the independent variable t and a precision index (such as WMAE or WRE) as the dependent variable y. Assume the linear regression model is as follows:

[0196]

[0197] Where: β0 is the intercept, β1 is the slope, and ε is the error term.

[0198] The values ​​of β0 and β1 are estimated using the least squares method.

[0199]

[0200]

[0201] in ,

[0202] .

[0203] The trend is determined by the value of β1. If β1 > 0, it indicates that the accuracy index is increasing over time (accuracy is decreasing); if β1 < 0, it indicates that the accuracy index is decreasing over time (accuracy is increasing).

[0204] 2) Prediction Algorithm

[0205] Prediction based on linear regression:

[0206] If the model obtained from linear regression analysis is:

[0207]

[0208] For a future time point t n+1 The value of the accuracy index can be predicted as follows:

[0209]

[0210] 3) Long-term trend analysis of vehicle-following monitoring and telemetry

[0211] For vehicle-following monitoring and telemetry, long-term series accuracy assessment data are also collected, including the comprehensive error index (CEI) under different v-VSP-weather-wind speed combinations, multivariate regression analysis results, etc.

[0212] Analyze the variation patterns of these indicators across different seasons and years. For example, for telemetry equipment, analyze the impact of changes in meteorological conditions (temperature, humidity, air pressure) on accuracy indicators under different seasons. If it is found that the CEI value of a certain pollutant is consistently high during the high temperatures of summer, it may be because high temperatures have a significant impact on the sensor performance of the telemetry equipment.

[0213] Based on the above method, the accuracy of the multi-emission sensing technology equipment data for heavy-duty vehicles is evaluated collaboratively, and the impact on accuracy is analyzed.

[0214] like Figure 2 The diagram shows the operational platform of the emission sensing device accuracy collaborative evaluation method based on remote online OBD monitoring, according to an embodiment of the present invention. The sample vehicle is equipped with a remote OBD monitoring terminal and transmits the remote OBD monitoring data to the monitoring platform. A PESM testing system is also synchronously installed on the sample vehicle. A following platform is integrated into a small car and synchronously follows the sample vehicle. The sample vehicle normally passes by a roadside telemetry system. Through the integration of the above sensing systems, emission data of the vehicle is collected synchronously, and the accuracy of different sensing technologies is evaluated according to the method of this embodiment of the invention.

[0215] This invention constructs a collaborative evaluation framework, integrating data from multiple emission sensing technologies and devices to upgrade the collaborative evaluation method for the accuracy of data from heavy-duty vehicle emission sensing technologies and devices. Its test vehicle sample selection mechanism collects historical data from multiple emission sensing technologies and selects vehicles based on a multi-objective optimization model, ensuring that the samples represent different pollutant concentrations and vehicle type characteristics. Regarding the integration of detection equipment, the tested vehicle integrates multiple devices to collect different emission data, while accompanying vehicles are equipped with relevant devices for synchronous detection, comprehensively and accurately reflecting vehicle emissions. The selection of telemetry stations follows multiple principles, considering the diversity of different road types, regions, and influencing factors, enabling the acquisition of more representative comparative data. The design of differences in data acquisition and preprocessing helps ensure the accuracy and completeness of the data. The design of the accuracy evaluation index calculation method and the data accuracy assessment method helps to accurately evaluate the accuracy of device data. These aspects combined improve the accuracy, comprehensiveness, and scientific rigor of the entire vehicle emission detection and evaluation process.

[0216] The embodiments of the present invention are illustrated below using specific actual test data:

[0217] Based on environmental policy and health impact studies, we assume that the weighting coefficients for the emitting species are determined as follows: w1 = 0.4 (weight of NOx), w2 = 0.3 (weight of HC), and w3 = 0.3 (weight of BC).

[0218] For the weighting coefficients of vehicle type-related targets, assuming that the research focuses on emission stage and vehicle age, we set v1=0.5, v2=0.3, v3=0.1, and v4=0.1.

[0219] Assume there are 10 cars, and the relevant data is shown in Table 2:

[0220] Table 2 Historical Emissions Test Data of Test Vehicle

[0221]

[0222] For vehicle 1, according to the objective function:

[0223] ;

[0224] Calculation yields:

[0225] ;

[0226] Similarly, the Z1 values ​​for other vehicles are calculated: the values ​​for vehicles 2 to 10 are 0.53, 0.91, 0.06, 0.36, 0.70, 0.15, 0.94, 0.00, and 0.59, respectively.

[0227] For vehicle 1, calculate Z2, and let y = vehicle i. 1i y 2i y 3i y 4i These categories are the vehicle's emission stage, age, weight, and industry characteristics.

[0228] ;

[0229] By coding the emission stages (China III = 3, China IV = 4, China V = 5, China VI = 6), the following calculations can be performed:

[0230] ;

[0231] Similarly, calculate the Z2 values ​​for other vehicles: the values ​​for vehicles 2 to 10 are 4.40, 6.10, 3.70, 5.90, 5.10, 5.00, 5.80, 4.00, and 5.10, respectively.

[0232] Then, Z2 is normalized to obtain Z2. ’ Z2 of vehicles 1-10 ’ The values ​​are, in order: 0.83, 0.29, 1.00, 0.00, 0.92, 0.58, 0.54, 0.88, 0.13, 0.58.

[0233] Emissions data constraint check:

[0234] For vehicle 1, x 1i =100, x 2i =20, x 3i =10.

[0235] For vehicle 4, x 1i =50, x 2i =15, x 3i =8, vehicle 4 meets emission data constraints.

[0236] For vehicle 7, x 1i =80, x 2i =18, x 3i =9, vehicle 7 meets emission data constraints.

[0237] Vehicle type constraint check:

[0238] For vehicle 1, the emission standard is "China IV", and the vehicle age is y. 2i =10, vehicle weight classification y 3i =1, Vehicle industry characteristic classification y 4i =6, so vehicle 1 satisfies the vehicle type constraint.

[0239] For vehicle 4, the emission standard is "China VI", and the vehicle age is y.2i =1, vehicle weight classification y 3i =1, Vehicle industry characteristic classification y 4i =3, vehicle 4 satisfies the vehicle type constraint.

[0240] For vehicle 7, the emission standard is "China IV", and the vehicle age is y. 2i =9, vehicle weight classification y 3i =1, Vehicle industry characteristic classification y 4i =2, vehicle 7 satisfies the vehicle type constraint.

[0241] Refinement of vehicle sample selection in multi-objective optimization algorithms:

[0242] Fitness calculation adjustment:

[0243] Assuming α = 0.6, for vehicle 1, the overall fitness function is:

[0244] F = αZ1 + (1-α)Z2 ’ ;

[0245] The calculated F-values ​​for vehicles 1, 4, and 7 are: 0.46, 0.04, and 0.30, respectively.

[0246] The vehicles are ranked according to their overall fitness scores as follows: Vehicle 4, Vehicle 7, and Vehicle 1.

[0247] When considering sample selection in crossover and mutation operations:

[0248] Assuming a crossover operation is performed, according to the rules, since vehicles 4 and 7 have the same weight classification (both are light vehicles), vehicles 4 and 1 are selected for crossover operation first to increase the diversity of vehicle types.

[0249] In the mutation operation, assuming the age of vehicle 1 is mutated, the age y is... 2i The mutation changed from 10 to 11. After the mutation, vehicle 1 still satisfies all constraints (emission data constraints and vehicle type constraints) and the diversity in the overall sample is not reduced.

[0250] Assuming the research focuses on the characteristics of low-emission vehicles, we select the vehicle with the lowest combined emission species value from the Pareto frontier (referred to here as vehicle 4, vehicle 7, and vehicle 1), namely vehicle 4.

[0251] Regarding the diversity of emission stages, vehicle 4, which is currently selected as "China VI", will be selected as "China IV" to increase diversity.

[0252] Regarding vehicle age diversity, vehicle 4 has an age of 1, and vehicle 1 has an age of 10, thus meeting certain diversity requirements.

[0253] Regarding the diversity of vehicle weight classification, both vehicle 4 and vehicle 1 are light vehicles, but since there are a large number of light vehicles in the overall sample and the research focuses on emission characteristics, no adjustment will be made for the time being.

[0254] Considering the diversity of vehicle industry characteristics, Vehicle 4 is an engineering vehicle, and Vehicle 1 is a passenger vehicle, thus meeting the diversity requirement. Therefore, the final determined typical vehicle types are Vehicle 4 and Vehicle 1.

[0255] Equipment integration and installation for the vehicle under test:

[0256] OBD online monitoring data acquisition equipment can acquire vehicle driving and NOx emission data through its built-in remote OBD online monitoring, or it can connect to the OBD port of the vehicle under test to obtain real-time data.

[0257] The PEMS testing equipment includes modules for collecting carbon dioxide (CO2), carbon monoxide (CO), hydrocarbons (HC), nitrogen oxides (NOx), particulate matter number (PN), and particulate matter mass (PM). It is also equipped with an exhaust gas sampling tube. A sampling fixing tube is installed at the vehicle's exhaust outlet to fix the PEMS exhaust gas sampling tube, the opacity meter's sampling tube, and the black carbon meter's sampling tube. The installation of the fixing tube does not change the original direction of the exhaust pipe, while minimizing the length of the fixing tube and ensuring it does not exceed the vehicle body boundary.

[0258] Install an opacity meter to monitor smoke opacity data.

[0259] Install a black carbon meter to monitor the black carbon content in vehicle exhaust.

[0260] Accompanying vehicles and equipment:

[0261] A separate vehicle will be assigned to conduct follow-up tests on the vehicle being tested. This follow-up vehicle will be equipped with a gas analyzer (CO2, NOx / NO), a black carbon analyzer, a particulate matter number (PN) meter, and a PM2.5 meter. 2.5 Testing equipment, as well as related sampling and dilution systems. Vehicle-following testing is used to perform synchronous testing on the vehicle under test while it is in motion.

[0262] To obtain as much comparative data as possible between telemetry stations and other detection equipment, and to reflect relevant influencing factors, the selection of telemetry stations should, in principle, ensure that the routes include different types of roads such as suburban roads, urban arterial roads, secondary arterial roads, and highways (if applicable), to cover telemetry stations under different traffic flows, speed limits, and road conditions. In terms of regional coverage, it involves different areas such as city centers, urban-rural fringe areas, and suburbs.

[0263] Meanwhile, since telemetry and vehicle-following tests are greatly affected by the environment (meteorological conditions, road traffic, etc.), this embodiment proposes a principle based on the diversity of influencing factors, including the vehicle speed, ambient temperature and wind conditions mentioned above.

[0264] The tests were conducted using the sample vehicles 4 and 1 identified above, where vehicle 4 is an engineering vehicle and vehicle 1 is a passenger vehicle.

[0265] Acquire pollutant emission data under different detection methods, such as PEMS test unit (including CO2, CO, HC, NOx, PN, PM), remote online monitoring (NOx), telemetry monitoring (CO2, CO, HC, NO, opacity), and vehicle-mounted testing (CO2, NOx / NO, PM). 2.5 Pollutant data in PN and black carbon (BC).

[0266] After acquiring data from various detection methods (PEMS testing, remote OBD online monitoring, telemetry, and vehicle-following testing), the vehicle speed (v) and power-specific performance (VSP) data are analyzed. Vehicle speed data can be obtained from the vehicle's speed sensor, while power-specific performance (VSP) can be calculated based on the vehicle's speed and acceleration using the following formula: VSP = v × (1.1 × a + 9.81 × sin(θ) + 0.132) + 0.000302 × v 3 , where a is the vehicle acceleration and θ is the road gradient.

[0267] Grouping and data classification based on v-VSP:

[0268] The collected data are grouped according to vehicle speed (v) and power specificity (VSP). Remote online OBD monitoring data is grouped using the v-VSP method. For vehicle-following monitoring and telemetry data, in addition to grouping by v-VSP, weather and wind speed conditions must also be considered.

[0269] Based on the calculation of mean absolute error (MAE) and relative error (RE), the influence weights of different v-VSP groupings on overall accuracy are considered. For example, the weights are determined according to the probability distribution of actual vehicle driving in different v-VSP intervals. Let w i Let y be the weight of the i-th v-VSP group, xij be the pollutant measurement value of the j-th data point under the i-th v-VSP group by remote online OBD monitoring, and y be the weight of the i-th v-VSP group. ij For comparative testing (such as PEMS testing), the reference value is the contaminant value at the j-th data point in the i-th v-VSP group, n i Let be the number of data points in the i-th group.

[0270] For each v-VSP group, refer to Figure 3The number of times vehicle driving condition data falls within a specific group is counted. The distribution and frequency of data points in Bin1 to Bin9 among the 6080 collected vehicle driving data points are shown in the figure.

[0271] In the v-VSP group, the NOx measurement value of remote online OBD monitoring is x. 1j PEMS test value y 1j Calculate the NOx deviation of OBD and PEMS tests within each VSP Bin.

[0272] Taking the data from the following 5 operating points as an example (NOx unit is g / km), calculate the mean absolute error (MAE) for this group:

[0273] ;

[0274] Then calculate the weighted average absolute error (WMAE):

[0275] Based on the previously calculated percentage of Bin as 7%, this group contributes 0.14 × 7% = 0.0098 to WMAE (unit: g / km).

[0276] Calculate the relative error (RE):

[0277] For the first data point,

[0278] ;

[0279] Similarly, calculate the relative errors for other data points. Then calculate the average relative error for this group: ;

[0280] Calculate the weighted relative error (WRE):

[0281] Based on a weight w1=7%, this group contributes 0.1266*7% = 0.0089 to WRE.

[0282] In a specific v-VSP group, NOx emissions were measured. Error data was obtained by analyzing 100 sets of remote online OBD monitoring and PEMS test data. The calculated mean of these error data was μ = 0.02, and the standard deviation was σ = 0.08. The acceptable error range was defined as [μ - 2σ, μ + 2σ] = [0.02 - 2 × 0.08, 0.02 + 2 × 0.08] = [-0.14, 0.18]. The number of data points within this range was 85. Therefore, the proportion of data points within the acceptable range is 85 / 100 = 0.85. This proportion can be used as a supplementary indicator for accuracy assessment.

[0283] For vehicle-mounted monitoring and telemetry, the comprehensive error index (CEI) coupling environmental factors is as follows:

[0284] Assuming that analysis of 50 comparative tests of vehicle-following monitoring and telemetry shows that in monitoring particulate matter (PM):

[0285] When considering the effects of v-VSP, meteorological factors (humidity), and wind speed on the error, the weight of v-VSP on the error is found to be w. v-VSP =0.4, the influence weight of meteorological factors (humidity) is w hu =0.3, the influence weight of wind speed factor is w win =0.3.

[0286] Assume that in vehicle-following monitoring or telemetry, there are 6 data points for measuring pollutant NOx under v-VSP group Bin2, with meteorological conditions of humidity range [50%, 60%) and wind speed conditions of [2m / s, 3m / s].

[0287] Measured value x ijkl They are respectively:

[0288] [0.5,0.6,0.4,0.55,0.45,0.5]

[0289] PEMS Comparison Test Reference Value y ijkl They are respectively:

[0290] [0.4,0.5,0.3,0.45,0.35,0.4].

[0291] First, calculate the mean absolute error (MAE) of the subset, which is 0.1. The MAE of this subset is: ;

[0292] Then calculate the Comprehensive Error Index (CEI) based on the previously determined weights w. v-VSP =0.4, w hu =0.3, w win =0.3.

[0293] Assuming no other special adjustment factors exist under these specific environmental conditions, then the CEI is: .

[0294] Suppose we establish a regression model for NOx measurements in telemetry data:

[0295] ;

[0296] Analysis of 50 sets of telemetry data yielded the following regression coefficient estimates: β0=0.1, β1=0.2, β2=0.05, β3=0.03, and the standard deviation of the error term ε was 0.02.

[0297] Since β1 = 0.2 is relatively large and passes the significance test (the hypothesis test results show that it is significantly different from 0), it indicates that vehicle speed (v-VSP) has a significant impact on NOx measurement accuracy. On the other hand, β3 = 0.03 is relatively small, and the significance test shows that its impact on measurement accuracy is not significant, indicating that wind speed has a small impact on NOx measurement accuracy.

[0298] Taking remote online OBD monitoring data clustering as an example:

[0299] When performing data clustering, the first step is to determine appropriate feature vectors. For remote online OBD monitoring, select accuracy metrics (such as WMAE, WRE) under different v-VSP groupings. For example, vehicles can be finely grouped according to speed (v) and power-to-weight ratio (VSP).

[0300] Accuracy index calculation (WMAE and WRE):

[0301] For each v-VSP group, we assume there are 10 data points used to calculate the accuracy metrics. For example, in the v-VSP group Bin, after measurement and comparison with the reference value, the calculated WMAE (weighted average absolute error) values ​​are [0.1, 0.12, 0.08, 0.11, 0.09, 0.13, 0.1, 0.09, 0.12, 0.11], and the WRE (weighted relative error) values ​​are [0.05, 0.06, 0.04, 0.055, 0.045, 0.065, 0.05, 0.045, 0.06, 0.055].

[0302] Similar calculations were performed for each group to obtain the WMAE and WRE values ​​for each group.

[0303] Constructing feature vectors:

[0304] Taking the v-VSP group Bin1 as an example, its feature vector is [0.1, 0.05] (here, the average of WMAE and WRE is used as an example). Such feature vectors are constructed for each group, resulting in 9 feature vectors.

[0305] K-Means clustering (K = 3)

[0306] Suppose that three cluster centers are initially selected randomly, for example:

[0307] Cluster center 1: [0.05, 0.025];

[0308] Cluster center 2: [0.15, 0.075];

[0309] Cluster center 3: [0.25, 0.125];

[0310] Calculate the distance from each data point (feature vector) to the three cluster centers. The distance formula can be the Euclidean distance formula. For a feature vector x=(x1,x2) and a cluster center y=(y1,y2), the distance is: ;

[0311] For example, for the feature vector [0.1, 0.05] of the v-VSP group Bin1, calculate the distances to the three cluster centers: d1, d2, d3.

[0312] Based on distance, the data points in the v-VSP group Bin1 are assigned to the cluster corresponding to the cluster center with the smallest distance. Let's assume they are assigned to the cluster corresponding to cluster center 1.

[0313] Repeat this process to divide all 9 feature vectors.

[0314] After multiple iterations (e.g., 5 iterations), the sum of the distances from each data point to its respective cluster center is minimized, resulting in 3 clusters representing high, medium, and low precision data. For example, cluster 1 (high precision) may contain Bin1 groups 1, 4, and 7; cluster 2 (medium precision) may contain Bin1 groups 2, 5, and 8; and cluster 3 (low precision) may contain Bin1 groups 3, 6, and 9.

[0315] For the high, medium, and low precision clusters obtained, we will conduct in-depth research on the characteristics of their corresponding v-VSP ranges. For example, in high precision clustering, we may find that most data points come from the medium vehicle speed and medium power ratio group. This may be because, under this operating condition, the vehicle engine runs relatively stably, and the OBD monitoring equipment can more accurately measure pollutant emission values.

[0316] In low-precision clustering, a large number of data points may be found in the high-vehicle-speed, high-specific-power group. Further analysis may reveal that at high vehicle speeds and high specific power, the engine combustion process is more complex, resulting in greater variations in the types and concentrations of pollutants produced. Additionally, under these conditions, the sensors in OBD monitoring devices may face issues such as insufficient response time and limited measurement range, leading to decreased accuracy.

[0317] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A collaborative evaluation method for the accuracy of emission sensing devices based on remote online OBD monitoring, characterized in that, include: Based on historical detection data from vehicle multi-emission sensing devices, a multi-objective optimization model is constructed with emission species and vehicle attributes as objectives. After solving the Pareto front through the multi-objective optimization model, the evaluation vehicle samples are determined according to the diversity index. The multi-emission sensing devices include a remote online OBD monitoring system, telemetry equipment, and vehicle-following test equipment. The multi-emission sensing device is used to monitor the test vehicle samples in real time to obtain multi-pollutant emission data of the test vehicles, and the multi-pollutant emission data is grouped according to the vehicle speed and specific power. Define the first influence weight of different groups on the evaluation accuracy, and calculate the weighted average absolute error and weighted relative error of the remote online OBD monitoring data in the multi-pollutant emission data according to the first influence weight. Define a second influence weight for different environmental factors on the evaluation accuracy. Based on the first influence weight and the second influence weight, calculate the comprehensive error index and multivariate regression coefficient based on the coupling of environmental factors for the telemetry data and vehicle-following test data in the multi-pollutant emission data. The weighted average absolute error and weighted relative error corresponding to each group are used as feature vectors for cluster analysis, and the comprehensive error index and / or multivariate regression results corresponding to each group are used as feature vectors for cluster analysis to obtain the accuracy evaluation results of the multivariate emission sensing equipment. Among them, based on historical detection data from vehicle multi-emission sensing devices, a multi-objective optimization model is constructed with emission species and vehicle attributes as the objectives, including: Based on the set emission data constraints and vehicle type constraints, emission species objective functions and vehicle type objective functions corresponding to the vehicle multi-emission sensing device are constructed respectively. The vehicles to be evaluated are screened according to the emission species objective functions and the vehicle type objective functions to obtain the evaluation vehicle screening results. Based on the emission species objective function and the vehicle type objective function, a comprehensive fitness function is defined for each vehicle. A multi-objective optimization model based on the comprehensive fitness function is constructed. The multi-objective optimization model is used to filter the evaluation vehicle selection results and solve the Pareto front. The method further includes: After grouping the multi-pollutant emission data according to vehicle speed and power-to-weight ratio, the multi-pollutant emission data is further subdivided into subsets according to meteorological and wind speed conditions to obtain the final grouping results.

2. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 1, characterized in that, Based on the emission species objective function and the vehicle type objective function, the vehicles to be evaluated are screened to obtain the vehicle screening results, which also includes: With the goal of minimizing the overall impact of emission species detected by the multi-emission sensing devices, a first weighting coefficient is defined for each emission sensing device. Based on the first weighting coefficient and the set emission data constraints, an emission species objective function corresponding to the vehicle multi-emission sensing device is constructed. Based on the emission species objective function, the emission species objective function value of each vehicle is calculated. The emission species objective function value is compared with a set threshold to screen the vehicles to be evaluated.

3. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 2, characterized in that, Based on the emission species objective function and the vehicle type objective function, the vehicles to be evaluated are screened to obtain the vehicle screening results, which also includes: With the goal of minimizing the average emission stage of vehicles, a second weighting coefficient is defined for different emission stages of vehicles. Based on the second weighting coefficient and the set vehicle type constraints, a vehicle type objective function is constructed. Based on the vehicle type objective function, the vehicle type objective function value of each vehicle is calculated. The vehicle type objective function value is compared with a set threshold to screen the vehicles to be evaluated. The vehicle types include vehicle emission stage, vehicle age, vehicle weight classification, and vehicle industry characteristic classification.

4. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 1, characterized in that, The method further includes: Based on historical detection data from vehicle multi-emission sensing devices, the average pollutant emissions detected by each emission sensing device for each vehicle are calculated. The average pollutant emissions are then compared with the set corresponding pollutant emission level thresholds to obtain the initial screening results of the vehicles to be evaluated. The initial screening results of the evaluated vehicles are then further screened based on the emission species objective function and the vehicle type objective function to obtain the final evaluation vehicle screening results.

5. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 1, characterized in that, The method further includes: Based on historical detection data or historical data statistical models of vehicle multi-emission sensing devices, the error probability distribution between pollutant emission measurement values ​​and reference values ​​is determined under different groups corresponding to vehicle speed and specific power. The error probability distribution is used as an accuracy evaluation index, and together with the weighted average absolute error and weighted relative error corresponding to each group, they are used as feature vectors for cluster analysis.

6. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 1, characterized in that, The method further includes: The accuracy assessment data of the multi-emission sensing device is obtained over a set long period of time. Based on the accuracy assessment data, trend analysis and prediction algorithms are used to perform trend analysis on the accuracy of the multi-emission sensing device and issue an early warning.

7. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 1, characterized in that, The method further includes: After acquiring the historical detection data of the vehicle's multi-emission sensing device, and after acquiring the multi-pollutant emission data of the evaluation vehicle, data preprocessing is performed. The data preprocessing process includes data cleaning, data standardization, and data integrity check.

8. The method for collaborative evaluation of the accuracy of emission sensing devices based on remote online OBD monitoring according to claim 1, characterized in that, The multi-objective optimization model adopts the NSGA-II algorithm.

Images