A method for monitoring and early warning of motor vehicle exhaust emissions based on the Internet of Things

Abstract

This invention discloses a method for monitoring and early warning of motor vehicle exhaust emissions based on the Internet of Things (IoT), specifically relating to the field of motor vehicle exhaust emission monitoring. The invention includes benchmark matrix construction and storage, real-time data acquisition, weighted comprehensive deviation calculation of exhaust emission concentration under vehicle operating conditions, second-level verification, exceedance determination, graded early warning, and iterative optimization of the benchmark matrix. This invention constructs a three-dimensional tensor-specific benchmark matrix for each registered motor vehicle through a cloud system, and combines this with roadside sensor modules to collect exhaust emission multi-component concentration, velocity, and acceleration data. The weighted comprehensive deviation is quantitatively calculated through matrix operations to achieve high-precision exhaust emission exceedance determination. This invention dynamically optimizes the benchmark matrix using the recursive least squares method, avoiding false alarms, missed alarms, and monitoring failures caused by benchmark disconnection, thereby improving environmental supervision efficiency, reducing unnecessary waste of regulatory resources, and ensuring the long-term accurate operation of the monitoring system.

Description

Technical Field

[0001] This invention relates to the field of motor vehicle exhaust emission monitoring, and more specifically, to a method for monitoring and early warning of motor vehicle exhaust emissions based on the Internet of Things. Background Technology

[0002] With the deep integration of IoT technology and environmental regulation, vehicle exhaust emissions have become a key factor affecting air quality. According to statistics, mobile source emissions account for 30%-50% of the total urban air pollutants. Among them, pollutants such as carbon monoxide (CO) and nitrogen oxides (NOx) are the core causes of environmental problems such as photochemical smog and acid rain. Real-time and accurate exhaust monitoring and early warning have become the core requirements for curbing excessive emissions and improving the efficiency of environmental regulation.

[0003] Currently, in the field of motor vehicle exhaust emission monitoring, the main existing technologies include fixed exhaust emission testing stations, portable equipment sampling, and simple monitoring methods based on a single sensor. Among them, fixed testing stations require vehicles to actively enter the station, which has the problems of limited coverage and inability to achieve dynamic monitoring. It can only sample and test passing vehicles, making it difficult to comprehensively grasp the emission status of motor vehicles in the area. Portable equipment sampling relies on manual operation, which has low testing efficiency and high cost, and is greatly affected by factors such as testing time and weather, making it impossible to form routine monitoring. Simple monitoring methods based on a single sensor can only collect data on a single exhaust component and do not consider the impact of vehicle speed, acceleration, and other operating conditions on emissions, resulting in monitoring data lacking scenario adaptability.

[0004] However, in actual use, it still has some shortcomings. First, the existing technology generally lacks a precise benchmark that is adapted to the vehicle's operating conditions. It mostly uses a uniform emission limit standard to judge whether it exceeds the standard. It does not combine the vehicle's emission standard, engine displacement, driving speed and acceleration and other personalized attributes to build a special benchmark, which leads to a high false alarm rate. For example, the normal emissions of the same vehicle under rapid acceleration may be judged as exceeding the standard, while the slight and continuous exceedance of the standard of the old vehicle under low speed is difficult to identify, which cannot provide accurate decision-making basis for supervision.

[0005] Second, existing technologies lack spatiotemporal continuity analysis of monitoring data. They determine emission status based solely on single detection data and cannot distinguish between temporary sensor interference, occasional emission fluctuations caused by temporary vehicle operation, and real and continuous exceedances caused by vehicle malfunctions. For example, abnormal data generated by sensors due to dust or airflow interference may be misjudged as vehicle exceedances, leading to unnecessary regulatory intervention. On the other hand, continuous exceedances caused by vehicle emission system malfunctions may also be missed if historical data trend analysis is not combined, resulting in regulatory lag. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a method for monitoring and warning of motor vehicle exhaust emissions based on the Internet of Things. The present invention provides the following technical solution:

[0007] S1. For each registered motor vehicle, a unique three-dimensional tensor matrix is ​​generated by combining its static attribute vector with historical emission and operating condition data of compliant vehicles of the same type through the cloud system. The matrix is ​​then bound to the vehicle identification code and stored in the Internet of Things database.

[0008] S2. Real-time data collection of vehicle identification code and exhaust gas multi-component concentration data, vehicle speed and acceleration data through roadside sensor modules, forming raw data vectors, and synchronously storing them in the Internet of Things database.

[0009] S3. The cloud system retrieves the corresponding reference matrix based on the vehicle identification code, extracts speed and acceleration data from the original data vector, maps it to the corresponding working condition unit of the matrix, and extracts the theoretical reference vector and tolerance vector of each unit.

[0010] S4. Calculate the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions by combining the measured data of multi-component concentration of exhaust gas from each unit through the matrix operation module.

[0011] S5. Determine the threshold for the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions. If it exceeds the threshold, trigger the second verification. The second verification module retrieves historical monitoring data of the vehicle to construct a time series matrix and calculates the instantaneous volatility and long-term trend of the current vehicle.

[0012] S6. Based on the fluctuation and trend analysis results, determine whether the exhaust gas concentration actually exceeds the standard through the exceedance judgment module, and output a diagnostic conclusion including the exceedance type, confidence level, related operating conditions and historical comparison results.

[0013] S7. Based on the diagnostic conclusion, generate graded early warning information through the early warning decision module, and record the early warning level and trigger time after confirming that the information is real and continues to exceed the standard.

[0014] S8, the model adaptive module uses the recursive least squares method to integrate newly added compliant data into the benchmark matrix and dynamically optimize the theoretical benchmark value and tolerance range under each working condition.

[0015] The technical effects and advantages of this invention are as follows:

[0016] This invention constructs a three-dimensional tensor-specific reference matrix for each registered motor vehicle based on static attribute vectors and historical data of compliant vehicles of the same type. Combined with real-time speed and acceleration mapping to match corresponding operating condition units, it breaks through the limitations of existing technologies that use uniform emission limits. It achieves a precise reference that is deeply adapted to the vehicle's personalized attributes and real-time operating conditions, effectively reducing false alarms caused by differences in operating conditions and providing a scientific and accurate basis for environmental supervision.

[0017] This invention uses a threshold to determine the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions. If the deviation exceeds the threshold, historical monitoring data of the vehicle is retrieved to construct a time series matrix. This matrix is ​​then used to calculate the instantaneous fluctuations and long-term trends of the current vehicle, forming a dual verification mechanism of single data calculation and historical trend analysis. This mechanism can accurately distinguish between temporary sensor interference, temporary vehicle operation fluctuations, and real and continuous exceedances. It solves the shortcomings of existing technologies that rely solely on single data determination and cannot eliminate interference, significantly improving the accuracy of exceedance determination and avoiding unnecessary regulatory intervention and missed exceedances.

[0018] This invention uses a recursive least squares method through a model adaptive module to dynamically integrate newly added compliant data into the benchmark matrix, continuously optimizing the theoretical benchmark values ​​and tolerance ranges under various operating conditions. This allows the benchmark standards to be iteratively updated as regional vehicle technology upgrades and fuel quality improvements occur. This overcomes the problem of existing technical benchmarks being fixed and becoming out of touch with actual compliant emission levels after long-term use, ensuring that the monitoring system maintains high accuracy over the long term and adapts to the long-term dynamic needs of environmental regulation. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0020] Figure 2 This is a flowchart illustrating the overall structure of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] As attached Figure 1 The method for monitoring and warning of motor vehicle exhaust emissions based on the Internet of Things is characterized by including a cloud system, a roadside sensing module, an Internet of Things database, a matrix operation module, a second verification module, an exceedance judgment module, an early warning decision module, and a model adaptation module.

[0023] The cloud system generates a three-dimensional tensor exclusive reference matrix for each registered motor vehicle and binds it to the VIN storage. After retrieving the matrix, it locates the working condition unit according to the real-time speed and acceleration, and extracts the theoretical reference and tolerance vector.

[0024] The roadside sensing module is used to collect multi-component concentration, velocity and acceleration data of exhaust gas, form a raw data vector and store it in the Internet of Things database, and perform environmental compensation calibration on the gas concentration data.

[0025] The IoT database is used to store the reference matrix, raw and calibration data, and historical vehicle monitoring data for retrieval by various modules.

[0026] The matrix operation module is used to combine measured and benchmark data to calculate the standardized residual vector and the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions, thereby quantifying the degree of emission deviation.

[0027] The second verification module retrieves historical data to construct a time series matrix when the deviation exceeds the threshold, and calculates the instantaneous volatility and long-term trend of the current vehicle.

[0028] The above-mentioned exceedance determination module is used to determine whether the exceedance is real or temporary interference based on the volatility and trend, and output a diagnostic conclusion.

[0029] The early warning decision module is used to determine the early warning level based on the diagnostic conclusion, generate early warning information, and record the early warning level and trigger time.

[0030] The model adaptive module is used to filter newly added compliant data, optimize the benchmark matrix using recursive least squares, and rebind the VIN storage after updating.

[0031] As attached Figure 2 The method for monitoring and issuing early warning of vehicle exhaust emissions based on the Internet of Things (IoT) includes the following steps in its specific implementation:

[0032] S1. For each registered motor vehicle, a unique three-dimensional tensor matrix is ​​generated by combining its static attribute vector with historical emission and operating condition data of compliant vehicles of the same type through the cloud system. The matrix is ​​then bound to the vehicle identification code and stored in the Internet of Things database.

[0033] It should be explained that the static attribute vector Ps is an inherent attribute of the motor vehicle, a core parameter that remains essentially unchanged throughout the vehicle's service life. This parameter needs to be digitally encoded to meet the requirements of matrix operations, specifically including the following dimensions and encoding rules:

[0034] Emission Standards: These correspond to the national emission limits met by the vehicle at the time of manufacture, using direct numerical coding, such as China V = 5, China VI = 6; Engine Displacement: The engine's working volume in liters, coded discretely within fixed ranges, such as 1.0L and below = 1, 1.0-1.6L = 2, 1.6-2.0L = 3, and above 2.0L = 4; Curb Weight: The vehicle's unloaded weight in kg, coded according to weight ranges, such as 1000kg and below = 1, 1000-1500kg = 2, 1500-2000kg = 3, and above 2000kg = 4; Vehicle Classification: Classified according to vehicle purpose and structure, using preset numerical coding, such as private car = 1, taxi = 2, truck = 3, bus = 4, and freight car = 5.

[0035] The specific steps for training and generating a dedicated benchmark matrix in the form of a three-dimensional tensor are as follows:

[0036] A1. Sample Data Screening and Preprocessing: Historical data of compliant vehicles of the same type that highly match the static attribute vector of the target vehicle are selected from the IoT database as sample data. That is, the emission standards and vehicle type classifications must be consistent, and the engine displacement and curb weight must be in the same coding range. The sample data must include exhaust gas multi-component concentration data and operating condition data: speed v and acceleration a. Environmental compensation calibration is performed on the screened sample data to eliminate the interference of temperature, humidity and air pressure on exhaust gas concentration, so as to obtain an effective sample set under standard conditions.

[0037] A2. Operating Condition Dimension Interval Division: The speed and acceleration data in the effective sample set are divided into discrete intervals according to the actual driving conditions of the vehicle: Speed ​​interval i: e.g., 0-20km / h=1, 20-60km / h=2, 60-100km / h=3, 100km / h and above=4; Acceleration interval j: e.g., a≤0m / s²=1, 0<a≤1m / s²=2, 1<a≤3m / s²=3, a>3m / s²=4; forming a (i,j) operating condition unit combination covering the main driving conditions of the vehicle;

[0038] A3. Statistical Analysis of Exhaust Gas Concentration in Each Operating Unit: The exhaust gas multi-component concentration data in the valid sample set are grouped according to operating units (i, j). For each exhaust gas component k under each operating unit, the core parameter is calculated using a statistical algorithm: the mean concentration μ of exhaust gas component k under that operating unit is calculated. k As a theoretical benchmark value; calculate the standard deviation σ of the concentration of exhaust gas component k under this operating condition unit. k This serves as the basic data for the tolerance range; the reasonable concentration parameter (μ) for each (i, j, k) dimension is obtained. k , σ k );

[0039] A4. Assembly and Storage of the Three-Dimensional Tensor Reference Matrix: Using velocity interval i, acceleration interval j, and exhaust gas component k as the coordinate axes of the three-dimensional tensor, the (μ) obtained in A3 is... k , σ k The parameters are filled into the corresponding (i, j, k) positions to form a dedicated reference matrix Bm in the form of a three-dimensional tensor. The matrix element Bm[i, j, k] = [μ k , σ k Finally, the matrix is ​​uniquely bound to the vehicle identification number (VIN) of the target vehicle and stored in the IoT database to provide a benchmark for subsequent real-time monitoring deviation calculation.

[0040] It needs further explanation that Bm[i, j, k] stores the average concentration μ of exhaust gas component k under this operating condition. k and standard deviation σ k It is used to define a reasonable concentration range.

[0041] S2. Real-time data collection of vehicle identification code and exhaust gas multi-component concentration data, vehicle speed and acceleration data through roadside sensor modules, forming raw data vectors, and synchronously storing them in the Internet of Things database.

[0042] It should be explained that the multi-component concentration data of exhaust gas refers to the concentration parameters of core pollutants in motor vehicle exhaust that have a significant impact on the environment and human health, including carbon monoxide (CO) concentration, hydrocarbon (HC) concentration, nitrogen oxide (NOx) concentration, and smoke opacity. The units for CO, HC, and NOx concentrations are all volume concentrations in parts per million (ppm). CO is the main product of incomplete combustion of fuel, HC is an important precursor to photochemical smog, and NOx mainly consists of nitric oxide (NO) and nitrogen dioxide (NO2), which are both toxic and environmentally damaging. The unit for smoke opacity is the light absorption coefficient (m). -1 It is used to characterize the particulate matter (PM) content in the exhaust gas of compression-ignition engines such as diesel vehicles, reflecting the degree of turbidity of the exhaust gas.

[0043] Multi-component concentration data is collected using a roadside fixed multi-component gas sensor array. When a vehicle enters the 3-5 meter monitoring range of the sensor, the infrared vehicle detector triggers sampling. The air pump built into the sensor array extracts exhaust gas samples. After filtering out impurities, the electrochemical sensor converts the concentrations of CO, HC, and NOx into electrical signals through a redox reaction. The optical sensor converts smoke opacity into light absorption coefficient through laser transmission. The electrical signals are converted from analog to digital and output as standard unit data, which is then uploaded to the Internet of Things database. Vehicle speed data is collected using a roadside 77GHz millimeter-wave speed radar, which is deployed at an angle of 30° to 45° to the vehicle's direction of travel. The system calculates the difference between the echo and the transmission frequency using the Doppler effect, and converts this into instantaneous speed by combining radar parameters. This instantaneous speed is then linked to exhaust gas data via timestamps. Vehicle acceleration data is acquired using a combination of speed difference calculation and real-time acquisition. The speed-measuring radar continuously acquires two instantaneous speeds, v1 and v2, of the same vehicle within 0.1 seconds. Instantaneous acceleration is calculated using the formula a=(v1-v2)÷0.1. If the acceleration value exceeds the vehicle's physical acceleration limit of ±5m / s², the cloud system will remove the outlier and recalculate. Finally, the acceleration data, along with the speed and exhaust gas concentration data, are incorporated into the original data vector and synchronously stored in the IoT database.

[0044] S3. The cloud system retrieves the corresponding reference matrix based on the vehicle identification code, extracts speed and acceleration data from the original data vector, maps it to the corresponding working condition unit of the matrix, and extracts the theoretical reference vector and tolerance vector of each unit.

[0045] It should be explained that both the theoretical baseline vector and the tolerance vector are generated based on the vehicle-specific baseline matrix Bm retrieved by the cloud system. The theoretical baseline vector Bth is the set of average values ​​of the reasonable concentrations of each exhaust gas component under the current operating condition unit. Specifically, after the cloud system extracts the velocity and acceleration data from the original data vector, it locates the (i, j) operating condition unit matched in the baseline matrix, and then extracts the average concentration μ of each type of exhaust gas component k from that unit. k , such as μ CO μ HC μ NOx With μ Smoke These means are combined in a fixed order to form the theoretical benchmark vector Bth=[μ CO μ HC μ NOx μ Smoke The tolerance vector σ is the set of standard deviations of the allowable range of concentration fluctuations of each exhaust gas component under the same operating condition unit. Similarly, the standard deviation σ of the concentration of each type of exhaust gas component k is extracted from the matched (i, j) operating condition unit. k , such as σ CO σ HC σ NOx and σ SmokeBy combining the components in the same order as the theoretical reference vector, the tolerance vector σ = [σ] is obtained. CO , σ HC , σ NOx , σ Smoke Both serve as benchmarks for judging whether exhaust emissions exceed standards.

[0046] S4. Calculate the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions by combining the measured data of multi-component concentration of exhaust gas from each unit through the matrix operation module.

[0047] It needs to be explained that the calculation of the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions needs to be implemented step by step through matrix operations. First, it is necessary to construct the measured vector M of the multi-component exhaust gas concentration. That is, extract the CO, HC, NOx concentration values ​​and smoke opacity value after environmental compensation calibration from the original data vector, and combine them according to the component order of the theoretical reference vector to form the measured vector M = [CO, HC, NOx, Smoke]. Next, the standardized residual vector is calculated using the formula Z = (M - Bth) ⊘ σ, where ⊘ represents the Hadamard product. That is, the measured vector M is subtracted from the theoretical reference vector Bth element by element to obtain the concentration deviation value of each component, and then divided by the corresponding element of the tolerance vector σ to finally generate the standardized residual vector Z = [Z CO Z HC Z NOx Z Smoke In the vector, each element represents the standard deviation of the corresponding exhaust component from its theoretical baseline value; finally, the weighted comprehensive deviation of the vehicle's exhaust concentration under operating conditions, Δmatrix, is calculated using the formula Δmatrix=Z. T •W•Z is implemented, where Z T It is the transpose of the standardized residual vector Z, and W is a diagonal weight matrix. The matrix elements are weight coefficients set according to the degree of harm of each exhaust gas component to the environment and health. The specific form of the weight matrix is ​​W=diag(w CO w HC w NOx w Smoke ), where w represents the weight of each component, and Δmatrix, obtained through matrix multiplication, is a scalar value. The magnitude of the value directly reflects the severity of the vehicle's current exhaust emissions deviating from the baseline. The larger the value, the more significant the emission exceedance.

[0048] S5. Determine the threshold for the weighted comprehensive deviation of vehicle exhaust gas concentration under operating conditions. If it exceeds the threshold, trigger the second verification. The second verification module retrieves historical monitoring data of the vehicle to construct a time series matrix and calculates the instantaneous volatility and long-term trend of the current vehicle.

[0049] It needs to be explained that the threshold determination of the weighted comprehensive deviation of vehicle exhaust concentration under operating conditions refers to setting a comprehensive deviation threshold Δthreshold. If Δmatrix ≤ Δthreshold, the vehicle is qualified and the data is entered into the database; if Δmatrix > Δthreshold, a second verification is triggered. The comprehensive deviation threshold Δthreshold is set based on the weighted comprehensive deviation mean of compliant vehicles of the same type under the corresponding operating conditions plus 2 times the standard deviation.

[0050] When the second verification is triggered, the second verification module will retrieve the historical data of the vehicle from the most recent N monitoring times. This historical data comes from the raw data vectors collected by the roadside sensor module and stored in the IoT database. After preprocessing and calculation, it includes the measured vector Mi of exhaust gas multi-component concentration, vehicle speed vi, acceleration ai, and the corresponding weighted comprehensive deviation of exhaust gas concentration under vehicle operating conditions Δmatrixi at each monitoring time. These data are organized in chronological order of monitoring time to construct a time series matrix Tmatrix. The specific form of the matrix is: Tmatrix=[[M1, V1, a1, Δmatrix1], [M2, V2, a2, Δmatrix2]…[MN, VN, aN, ΔmatrixN]], thus fully presenting the recent emission and operating condition change trajectory of the vehicle.

[0051] Based on the constructed time series matrix, instantaneous volatility and long-term trends are further calculated: When calculating instantaneous volatility, the mean μΔ and standard deviation σΔ of the weighted comprehensive deviation of exhaust gas concentration under all historical vehicle operating conditions from Δmatrix1 to ΔmatrixN in the matrix are first calculated using statistical methods. Then, these values ​​are substituted into the formula ZΔ=(current Δmatrix-μΔ)÷σΔ to obtain the standardized volatility value ZΔ of the current deviation relative to historical data. This value is used to determine whether the current exceedance is a sudden anomaly. When calculating long-term trends, a coordinate system is established with the monitoring time as the horizontal axis and the corresponding historical Δmatrix value as the vertical axis. The least squares method is used to perform linear fitting on these data points. The trend is determined by calculating the slope of the fitted line: if the slope is positive and the absolute value is large, it indicates that the vehicle weighted comprehensive deviation is showing a continuous upward trend, and the emission situation is gradually deteriorating; if the slope is close to 0 or negative, it indicates that the emission situation is relatively stable or has improved, providing a trend-level basis for subsequent determination of the nature of the exceedance.

[0052] S6. Based on the fluctuation and trend analysis results, determine whether the exhaust gas concentration actually exceeds the standard through the exceedance judgment module, and output a diagnostic conclusion including the exceedance type, confidence level, related operating conditions and historical comparison results.

[0053] When determining whether an error exceeds the limit, the module uses the ZΔ value calculated from instantaneous fluctuations and the slope of the fitted straight line obtained from long-term trend analysis as the core criteria. If ZΔ is extremely large, regardless of the trend slope, it is determined to be temporary interference. This type of situation often stems from temporary interference to the sensor from the external environment or temporary special operation of the vehicle. In this case, the type of error exceeding the limit in the diagnostic conclusion is marked as temporary interference, and the confidence level is set according to the degree of deviation of ZΔ. The larger ZΔ is, the higher the confidence level of temporary interference. For the associated operating conditions, the current monitored speed interval i and acceleration interval j are extracted. The historical comparison results need to explain the deviation factor of the current ZΔ from the historical average μΔ, clarifying that the current anomaly is an occasional situation. If ZΔ is large and the slope of the fitted straight line is positive, it is determined to be a real and continuous error exceeding the limit. In cases where the emission system may be malfunctioning, the type of exceedance in the diagnostic conclusion should be marked as genuine and persistent. The confidence level should be set by combining the magnitude of ZΔ and the absolute value of the slope. The associated operating conditions should also clearly define the intervals corresponding to the current speed and acceleration. Historical comparison results should show the trend of the change of the Δmatrix over the most recent N times, intuitively reflecting the emission deterioration process. If ZΔ is large but the slope is close to 0 or negative, it is necessary to further check whether there are similar fluctuations in the historical data. If there is only one abnormality and there is no upward trend, it is still judged as a suspected temporary interference, and the confidence level should be appropriately reduced. The diagnostic conclusion should indicate that subsequent monitoring and observation are recommended to ensure that the judgment results are rigorous and consistent with the actual emission situation, providing an accurate basis for subsequent early warning decisions.

[0054] S7. Based on the diagnostic conclusion, generate graded early warning information through the early warning decision module, and record the early warning level and trigger time after confirming that the information is real and continues to exceed the standard.

[0055] The early warning decision module first determines the corresponding early warning level classification standard based on the type of exceedance, confidence level, and historical comparison results in the diagnostic conclusion. If the diagnostic conclusion is a true and continuous exceedance with a confidence level ≥90%, and the historical comparison shows a continuous increase in the Δmatrix over the last N times, it is judged as a Level 1 warning, corresponding to a serious vehicle emission problem and a clear trend of deterioration. If the confidence level for a true and continuous exceedance is between 80% and 90%, and the historical data only shows an upward trend recently, it is judged as a Level 2 warning, indicating that there is a problem with vehicle emissions but the degree of deterioration is relatively slow. If the diagnostic conclusion is a suspected true exceedance, it is judged as a Level 3 warning, serving only as a potential risk warning. After determining the warning level, the early warning decision module will simultaneously record the specific time of the warning trigger and associate it with the key information in the corresponding diagnostic conclusion to ensure that the recorded information is complete and traceable.

[0056] S8, the model adaptive module uses the recursive least squares method to integrate newly added compliant data into the benchmark matrix and dynamically optimize the theoretical benchmark value and tolerance range under each working condition.

[0057] The model adaptation module first filters newly added compliant data from the IoT database. This data must meet two core conditions: first, the corresponding vehicle must be determined to be compliant by the emission violation judgment module; second, the data must contain complete exhaust gas multi-component concentration data, vehicle speed and acceleration data, and vehicle static attribute information, and temperature, humidity, and air pressure compensation calibrations must be completed to ensure that the data quality meets the requirements of the baseline matrix optimization. The newly added compliant data after filtering will be classified according to speed range i and acceleration range j, corresponding one-to-one with the existing (i, j) operating condition units in the baseline matrix Bm. For example, the speed 20-60km / h and acceleration 0<a≤1m / s² will be classified. 2 The newly added compliant CO concentration data under the operating conditions are classified under the operating condition unit i=2, j=2 in Bm.

[0058] The module then calls the recursive least squares method to calculate the theoretical baseline value μ for each working condition unit. k With tolerance range σ k Dynamic optimization is performed: For a certain (i, j, k) dimension, the historical compliance data and the newly added compliance data under that dimension are first integrated. The newly added data are then given a higher weight using the recursive least squares method. The updated concentration mean μ is then calculated through algorithm iteration. k ', replacing μ in the original Bm[i,j,k] k This makes the theoretical benchmark value closer to the actual emission levels of current compliant vehicles; at the same time, it recalculates the standard deviation σ of the integrated data in this dimension. k Update the tolerance range to ensure that the tolerance can cover normal emission fluctuations of currently compliant vehicles.

[0059] After optimization, the model adaptation module rebinds the updated baseline matrix Bm' with the corresponding vehicle identification number (VIN), overwriting the old baseline matrix in the IoT database.

[0060] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0061] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring and early warning of motor vehicle exhaust based on Internet of Things, characterized in that, The method comprises a cloud system, a roadside sensing module, an Internet of Things database, a matrix operation module, a second re-checking module, an over-standard judgment module, an early warning decision module and a model self-adaptive module, and the specific steps are as follows: S1, for each registered motor vehicle, the cloud system combines the static attribute vector with the historical emission and working condition data of the same type of compliant vehicle to train and generate a three-dimensional tensor form of an exclusive reference matrix, and the matrix is stored in the Internet of Things database in combination with the vehicle identification code; S2, the roadside sensing module collects motor vehicle identification codes and tail gas multi-component concentration data, vehicle speed and acceleration data in real time, forms an original data vector, and synchronously stores it in the Internet of Things database; S3, the cloud system retrieves the corresponding reference matrix according to the vehicle identification code, extracts the speed and acceleration data from the original data vector, maps them to the corresponding working condition unit of the matrix, and extracts the theoretical reference vector and tolerance vector of each unit; S4, the matrix operation module combines the measured data of the tail gas multi-component concentration of each unit to calculate the weighted comprehensive deviation degree of the vehicle working condition tail gas concentration; S5, the threshold value of the vehicle working condition tail gas concentration weighted comprehensive deviation degree is determined, if it exceeds the threshold value, the second re-checking is triggered, the historical monitoring data of the vehicle is retrieved by the second re-checking module to construct a time series matrix, and the instantaneous volatility and long-term trend of the current vehicle are calculated; S6, according to the analysis results of volatility and trend, the over-standard judgment module is used to judge whether the tail gas concentration is truly over-standard, and the diagnostic conclusion containing the over-standard type, confidence, associated working condition and historical comparison result is outputted; S7, based on the diagnostic conclusion, the early warning decision module generates graded early warning information, and records the early warning level and trigger time after confirming that the over-standard is real and continuous; S8, the model self-adaptive module uses the recursive least squares method to integrate the new compliant data into the reference matrix, and dynamically optimizes the theoretical reference value and tolerance range under each working condition.

2. The method for monitoring and warning of motor vehicle exhaust based on Internet of Things according to claim 1, characterized in that: The dimension and coding rule of the static attribute vector Ps specifically include: Emission standard: corresponding to the national tail gas emission limit standard met by the vehicle at the time of leaving the factory, directly coded by numbers; engine displacement: engine working volume in liters, discretely coded according to fixed intervals; kerb mass: the mass of the vehicle when it is empty, kg, coded according to weight intervals; vehicle type classification is based on vehicle purpose and structure, and preset digital coding is used.

3. The method of claim 1, wherein the method further comprises: The specific steps of training and generating an exclusive reference matrix in the form of a three-dimensional tensor are as follows: A1, sample data screening and pretreatment: screening sample data from the Internet of Things database, and performing environmental compensation calibration on the screened sample data to obtain an effective sample set; A2, working condition dimension interval division: dividing the speed and acceleration data in the effective sample set according to the characteristics of the actual driving working conditions of the vehicle to form discrete intervals: speed interval i and acceleration interval j, forming a combination of (i, j) working condition units covering the main driving working conditions of the vehicle; A3、Each operating unit tail gas concentration statistical analysis: for the tail gas multi-component concentration data in the effective sample set, grouping according to (i, j) operating unit, for each type of tail gas component k under each operating unit, calculate the concentration mean μ of tail gas component k under the operating unit as the theoretical reference value; k , as the theoretical reference value; Calculate the concentration standard deviation σ of the exhaust component k under the working condition unit k , as the basic data of tolerance range; get the reasonable concentration parameters (μ k , σ k ) under each (i, j, k) dimension A4, Three-dimensional tensor benchmark matrix assembly and storage: take the speed interval i, the acceleration interval j and the exhaust component k as the coordinate axes of the three-dimensional tensor, fill the (μ k , σ k ) parameters into the corresponding (i, j, k) positions to form a three-dimensional tensor form of the exclusive benchmark matrix Bm.

4. The method of claim 1, wherein the method further comprises: The tail gas multi-component concentration data specifically include: carbon monoxide CO concentration, hydrocarbon HC concentration, nitrogen oxide NOx concentration and smoke Smoke.

5. The method for monitoring and warning of motor vehicle exhaust according to claim 1, wherein: The theoretical benchmark vector Bth is specifically: the mean set of reasonable concentrations of each exhaust component under the current operating unit. After the cloud system extracts the speed and acceleration data from the original data vector, it will locate the matching (i, j) operating unit in the benchmark matrix, and then extract the concentration mean μ of each type of exhaust component k under this unit k These means are combined in a fixed order, i.e. to form the theoretical benchmark vector Bth; The tolerance vector σ is specifically: the standard deviation set of the allowable range of fluctuation of the concentration of each exhaust component under the same operating unit, and the concentration standard deviation σ of each type of exhaust component k is also extracted from the matched (i, j) operating unit k , combined in the order of the components consistent with the theoretical reference vector, to obtain the tolerance vector σ.

6. The method of claim 1, wherein the method further comprises: The calculation of the vehicle working condition exhaust concentration weighted comprehensive deviation degree is specifically: constructing an exhaust multi-component concentration measured vector M, combining according to the component order of a theoretical benchmark vector to form the measured vector M; calculating a standardized residual vector Z, which is realized by the formula Z=(M-Bth)⊘σ, wherein represents Hadamard product, that is, the corresponding elements of the measured vector M and the theoretical benchmark vector Bth are subtracted to obtain the concentration deviation value of each component, and then the corresponding elements of the tolerance vector σ are divided to finally generate the standardized residual vector Z; and finally calculating a weighted comprehensive deviation degree matrix, the formula being: Δmatrix=WZ, wherein Z T is the transpose of the standardized residual vector Z, and W is a diagonal weight matrix, the matrix elements being weight coefficients set according to the harm degree of each exhaust component to the environment and health, and the specific form of the weight matrix being W=diag(wCO, wHC, wNOx, wSmoke), wherein w represents the weight of each component. T ​ 7. The method of claim 1, wherein the method further comprises: The threshold determination method for the vehicle working condition exhaust concentration weighted comprehensive deviation degree is specifically: setting a comprehensive deviation degree threshold Δthreshold, if Δmatrix≤Δthreshold, the vehicle is qualified, and the data is stored in the database; if Δmatrix>Δthreshold, a second re-checking is triggered; the comprehensive deviation degree threshold Δthreshold is set based on the weighted comprehensive deviation degree mean value of the same type of compliant vehicle under the corresponding working condition plus 2 times the standard deviation.

8. The method of claim 1, wherein the method further comprises: The time series matrix is specifically: arranging the exhaust multi-component concentration measured vector Mi, vehicle speed vi, acceleration ai and the corresponding vehicle working condition exhaust concentration weighted comprehensive deviation degree Δmatrixi at each monitoring time in the order of monitoring time, and constructing a time series matrix Tmatrix, and the specific form of the matrix is: Tmatrix=[[M1, V1, a1, Δmatrix1], [M2, V2, a2, Δmatrix2]…[MN, VN, aN, ΔmatrixN]], N represents the historical data of the vehicle monitored for N times.

9. The method of claim 1, wherein the method further comprises: The calculation of the instantaneous fluctuation of the current vehicle is specifically: calculating the mean value μΔ and the standard deviation σΔ of all historical vehicle working condition exhaust concentration weighted comprehensive deviation degrees Δmatrix1 to ΔmatrixN in the matrix by a statistical method, and then substituting the formula ZΔ=(current Δmatrix-μΔ)÷σΔ to obtain the standardized fluctuation value ZΔ of the current deviation degree relative to the historical data.

10. The method of claim 1, wherein the method further comprises: The long-term trend is specifically: establishing a coordinate system with the monitoring time as the horizontal axis and the corresponding historical Δmatrix value as the vertical axis, linear fitting is performed on the data points by using the least square method, and the trend is determined by calculating the slope of the fitted straight line.

Images