Multi-source heterogeneous sensor mobile source fuel consumption dynamic credibility evaluation system and method

Abstract

The invention discloses a multi-source heterogeneous sensor mobile source fuel consumption dynamic credibility evaluation system and method, and aims to solve the problems that a traditional single sensor is large in monitoring error, low in reliability when working conditions are suddenly changed and incapable of meeting the carbon transaction MRV compliance requirement. The system comprises a heterogeneous sensor group (an OBD interface module, a 0.2-level ultrasonic flowmeter and a three-axis vibration sensor), a dynamic credibility evaluation engine, a weight distribution unit and a Kalman adaptive fusion module. A credibility factor model alpha i = 1 / sigma i2 * e (-beta * delta t) is constructed, dynamic weight adjustment is triggered in combination with vibration intensity, and self-adaptive fusion is carried out on multi-source data. Tests show that the average fuel consumption error is smaller than or equal to 0.41%, the rapid acceleration peak error is 0.63%, the maximum error is authenticated to be 0.48% by the Chinese metering institute, and the requirement that the MRV error is smaller than or equal to 1% is met. According to the method, a working condition self-adaptive credibility quantification mechanism is created for the first time, the limitation of traditional static fusion is broken through, the method is deployed on more than 3000 logistics vehicles, more than 2000 tons of vehicle fuel consumption accounting disputes are reduced annually, and the method is suitable for vehicle fuel consumption accounting and mobile source carbon accounting system construction and transaction MRV full scenes.

Description

Technical Field

[0001] This invention belongs to the field of vehicle energy consumption monitoring, carbon metering verification and sensor fusion technology. Specifically, it relates to a mobile source fuel consumption dynamic reliability evaluation system and method based on multi-source heterogeneous sensors, which is particularly suitable for carbon trading MRV (monitorable, reportable and verifiable) scenarios. Background Technology

[0002] With the rapid expansion of the global carbon trading market, mobile sources, as one of the core sources of carbon emissions (accounting for 25%-30% of global emissions), directly determine the credibility and fairness of carbon trading through the accuracy of their carbon accounting. The Ministry of Ecology and Environment's "Guidelines for Mobile Source Carbon Accounting" clearly requires that the fuel consumption measurement error under the carbon trading MRV (Monitorable, Reportable, Verifiable) system must be ≤1%. However, existing monitoring technologies have significant deficiencies and are unable to meet this rigid requirement.

[0003] Traditional OBD fuel consumption monitoring relies on the vehicle's ECU to calculate fuel injection volume. This system is susceptible to errors of ≥5% due to factors such as calibration strategies, fuel quality, and engine wear. Data from a 2023 study published in *Automotive Engineering* shows that the 25th percentile error is 4.8%, and the 75th percentile error reaches 6.1%. Under dynamic conditions such as rapid acceleration and deceleration, the peak error soars to 22.3%. This deficiency is particularly pronounced in the logistics industry, where long-haul vehicles travel over 100,000 kilometers annually. Even small errors can accumulate to several tons of CO2e deviation in carbon asset accounting, leading to frequent disputes between companies and verification agencies, and even policy penalties.

[0004] While a single ultrasonic flow meter can achieve a static accuracy of 0.2 and an average error of only 0.35%, its practical application is limited by complex operating conditions. Factors such as vehicle vibration, fuel bubbles, and pipeline turbulence can cause instantaneous errors exceeding ±1%. Under extreme conditions such as mountain climbing or starting under full load, the error further increases to 1.8%, barely meeting MRV requirements and failing to pass long-term verification stably. More importantly, performance degradation under extreme environments has not been fully considered—at low temperatures (below -20°C), the ultrasonic propagation speed is affected by the viscosity of the medium, increasing data delay by more than 30% and expanding the error to 2%; under high-load conditions, fuel turbulence intensifies, significantly increasing sensor signal distortion.

[0005] In-depth analysis reveals three core bottlenecks in existing technologies: First, when sensors operate independently, the reliability of data lacks quantifiable standards, making it impossible to determine the reliability of real-time data. Second, in dynamic scenarios such as sudden changes in operating conditions and extreme environments, no effective adaptive adjustment mechanism has been established, leading to a sharp drop in data stability. Third, multi-source data fusion often employs fixed weight allocation, ignoring the dynamic changes in sensor performance, resulting in insufficient accuracy of the fusion results. Although some multi-source fusion attempts have been made in the industry, none have broken through the limitations of static weights, resulting in limited accuracy improvements and failing to fundamentally solve the MRV compliance problem. Therefore, developing a fuel consumption monitoring system that can quantify reliability, dynamically adapt to operating conditions, and meet high-precision requirements has become an urgent need for the standardized development of the carbon trading market. Summary of the Invention

[0006] This invention aims to solve the technical problems in existing mobile source fuel consumption monitoring technologies, such as large errors of single sensors, low reliability during sudden changes in operating conditions, inability to quantify credibility, and static fusion not being suitable for dynamic operating conditions, which lead to the inability to meet the MRV error requirement of ≤1% in carbon trading.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A dynamic reliability evaluation system for fuel consumption of mobile sources using multi-source heterogeneous sensors, including:

[0009] The heterogeneous sensor group includes an OBD interface module, an ultrasonic flow meter with an accuracy of 0.2, and a triaxial vibration sensor, which are used to collect OBD fuel data, ultrasonic flow data, and vibration acceleration data.

[0010] The dynamic reliability evaluation engine connects its input to the output of a heterogeneous sensor array and is used to evaluate the historical error standard deviation σ of the sensors. i Data time delay Δt and vibration intensity calculation real-time reliability factor α i ;

[0011] The weight allocation unit, whose input is connected to the output of the dynamic credibility evaluation engine, is used to dynamically adjust and output the normalized fusion weights [α_obd, α_ultra] based on the vibration intensity and data delay.

[0012] The Kalman adaptive fusion module, with its input connected to the output of the weight allocation unit, is used to combine the observed values ​​of operating condition disturbances to output fuel consumption data that meets the MRV requirements.

[0013] Furthermore, the credibility factor α i The calculation model is as follows:

[0014] α i =1 / σ i 2 ×e^(-β·|Δt|);

[0015] Where σ i σ1 represents the standard deviation of the sensor's historical error, σ1 = 5% for the OBD interface module, and σ2 = 0.3% for the ultrasonic flow meter; Δt represents the data time delay (ms); β represents the attenuation coefficient, and β is positively correlated with the vibration intensity collected by the vibration sensor.

[0016] A dynamic weight allocation method includes the following steps:

[0017] S1: Obtain the acceleration value collected by the triaxial vibration sensor;

[0018] S2: If the acceleration value is >3.0g, trigger the OBD weight reduction by 40% and the ultrasonic flow meter weight reduction by 20%;

[0019] S3: Calculate the OBD data delay compensation coefficient e^(-0.01·Δt) and the ultrasonic flow meter data delay compensation coefficient e^(-0.005·Δt), and apply them to the corresponding reliability factors respectively;

[0020] S4: Normalize the adjusted credibility factor and output the fusion weights [α_obd, α_ultra] to the Kalman adaptive fusion module.

[0021] A Kalman adaptive fusion method, with the fusion equation as follows:

[0022]

[0023] Where y_obd is the OBD fuel data, y_ultra is the ultrasonic flow data, and K... k z is a gain matrix that dynamically adjusts with confidence level. k H represents the observation values ​​of the operating condition disturbance provided by the vibration sensor, and H is the observation matrix.

[0024] Furthermore, the system described in any of claims 1-2, the dynamic weight allocation method described in claim 3, or the Kalman adaptive fusion method described in claim 4 are applied in carbon trading MRV.

[0025] Beneficial effects of this invention:

[0026] 1. Breakthrough in Quantitative Reliability Measurement: The first reliability factor model based on historical error, time delay and vibration intensity is used to accurately quantify the reliability of sensor data and solve the problem that traditional technologies cannot assess reliability.

[0027] 2. Significantly improved accuracy: Through dynamic weight allocation and Kalman adaptive fusion, the average fuel consumption error is ≤0.41%, the peak error during rapid acceleration is 0.63%, and the maximum error is 0.48% (certified by the National Institute of Metrology, China CTI-2024-0217), which is 10 times more accurate than traditional OBD and fully meets the requirement of MRV error ≤1%.

[0028] 3. Strong adaptability to operating conditions: As a source of observation of operating condition disturbances, the vibration sensor enables adaptive adjustment of weights when operating conditions change abruptly, solving the problem of sudden drop in data reliability in scenarios such as rapid acceleration and deceleration.

[0029] 4. Significant application value: It has been deployed on more than 3,000 logistics vehicles, reducing vehicle fuel consumption by more than 2,000 tons per year, and providing stable and reliable fuel consumption measurement support for Mobile Source Carbon Trading (MRV). Attached Figure Description

[0030] Figure 1 This is a system architecture diagram of the present invention;

[0031] Figure 2 The curve showing the dynamic change of weights under rapid acceleration conditions;

[0032] Figure 3 A bar chart showing the error comparison before and after fusion. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0034] The present invention proposes a multi-source heterogeneous sensor mobile source fuel consumption dynamic reliability evaluation system and method, which achieves high-precision fuel consumption measurement through multi-source data fusion and dynamic reliability assessment. The specific implementation process is as follows:

[0035] (I) System Architecture Deployment

[0036] The system consists of a heterogeneous sensor array, a dynamic reliability evaluation engine, a weight allocation unit, and a Kalman adaptive fusion module. The deployment specifications and technical parameters of each component are as follows:

[0037] Heterogeneous sensor group deployment

[0038] OBD Interface Module: Select the ELM327-V2.1 version, which supports mainstream protocols such as SAE J1939 and ISO 15765, ensuring compatibility with vehicles meeting the National VI and above emission standards. Connect through the vehicle's OBD-II interface to collect 12 key data items such as fuel injection volume, engine speed, and intake pressure. The data update frequency is 10Hz, and the transmission delay ≤ 50ms. When deploying, ensure good interface contact, use anti-electromagnetic interference cables, and the cable length ≤ 1.5m to avoid signal attenuation;

[0039] Ultrasonic Flowmeter: Adopt a clamping sensor with an accuracy of 0.2 level, a measurement range of 0 - 100L / h, a working temperature of -40°C to 85°C, and is suitable for fuel pipelines with a pipe diameter of 10 - 50mm; The installation position is selected on a straight pipe section ≥ 50cm away from the fuel pump and ≥ 30cm away from the fuel filter. The installation angle is parallel to the pipeline axis, with a deviation not exceeding ±5° to prevent turbulent interference. The seal uses a fluororubber gasket, and the tightening torque is 8 - 10N·m to avoid fuel leakage; After installation, an airtightness test needs to be carried out. The pressure is maintained at 0.3MPa, and no leakage within 30 minutes is considered qualified;

[0040] Triaxial Vibration Sensor: Adopt a low-power sensor using MEMS technology, with a measurement range of ±10g, a sensitivity of 1mg / LSB, a data update frequency of 50Hz, and a working voltage of 3.3V; Installed on the side of the engine block or the transmission housing, fixed by a titanium alloy bracket with a bracket thickness ≥ 3mm, and a shock pad is pasted at the bottom to reduce the impact of resonance. The X-axis of the sensor is along the vehicle's driving direction, the Y-axis is perpendicular to the ground, and the Z-axis is along the vehicle's width direction to ensure accurate collection of vibration data;

[0041] Core Processing Unit: Select the NVIDIA Jetson Nano embedded processor, integrated with an ARM Cortex-A57 quad-core CPU and a 128-core GPU, with an operation frequency of 1GHz, supporting parallel operation to meet the real-time data processing requirements. Equipped with 16GB eMMC flash memory and 32GB Micro SD card, supporting local caching of raw data and fusion results within 7 days, and at the same time realizing remote data upload through a 4G / 5G module, and the upload frequency can be configured (1 time / second to 1 time / minute);

[0042] Component Connection Relationship:

[0043] The output of the heterogeneous sensor group is connected to the input of the dynamic reliability evaluation engine via a CAN bus, with a transmission rate of 250kbps to ensure real-time data transmission. The output of the dynamic reliability evaluation engine is connected to the weight allocation unit via an SPI interface, with a data transmission delay of ≤2ms. The weight allocation unit and the Kalman adaptive fusion module are integrated through an internal bus to achieve seamless transmission of weight data. The output of the Kalman adaptive fusion module is divided into two paths: one path connects to the vehicle terminal via an RS485 interface for local viewing; the other path uploads data to the carbon metering MRV system via a 4G / 5G module to meet remote monitoring and verification requirements.

[0044] (II) Algorithm Execution Flow

[0045] Data Preprocessing: The raw data collected by the sensors first undergoes preprocessing to ensure data quality.

[0046] Outlier removal: The 3σ criterion is used to calculate the mean μ and standard deviation σ of each sensor data, and data points that exceed the range of [μ-3σ, μ+3σ] are removed to avoid interference from extreme values;

[0047] Data synchronization: Based on the timestamp of the vibration sensor (accuracy 1μs), OBD data and ultrasonic flow data are aligned using linear interpolation. The OBD data update frequency is 10Hz, and the ultrasonic data update frequency is 20Hz. After synchronization, they are output at a uniform frequency of 50Hz, with the synchronization error controlled within ±1ms.

[0048] Data smoothing: A moving average filter with a window size of 5 data points is used to smooth the synchronized OBD data and ultrasonic flow data to reduce the impact of random noise. The filter formula is: y_smooth(i)=(y(i-2)+y(i-1)+y(i)+y(i+1)+y(i+2)) / 5, where i is the data index;

[0049] The credibility factor calculation dynamic credibility evaluation engine calculates the real-time credibility factor α based on preprocessed data through the following steps. i :

[0050] Initialization parameters: Set the initial historical error standard deviation according to the sensor type. For the OBD interface module, σ1 = 5%, and for the ultrasonic flow meter, σ2 = 0.3%. The initial value of the attenuation coefficient β is set to 0.01, and will be dynamically adjusted according to the vibration intensity.

[0051] Real-time parameter acquisition: Data is acquired every 10ms with a time delay Δt. The time difference between the sensor data timestamp and the system time difference of the core processing unit is used for calculation. Simultaneously, the triaxial acceleration values ​​of the vibration sensor are acquired, and the total vibration intensity a = √(a_x) is obtained through vector synthesis. 2 +a_y2 +a_z 2 )。

[0052] Dynamically adjust the attenuation coefficient β: Use the PID algorithm to achieve linear positive correlation adjustment between β and the vibration intensity. The adjustment formula is:

[0053] β = β0 + Kp×(a - a0), where β0 = 0.01, a0 = 0.5g, and Kp = 0.008. When the vibration intensity a ≤ 0.5g, β remains 0.01; when 0.5g < a < 3.0g, β increases linearly with a; when a ≥ 3.0g, the upper limit of β is set to 0.05 to avoid excessive attenuation.

[0054] Calculate the credibility factor:

[0055] Substitute into the model α i = 1 / σ o 2 ×e^(-β·|Δt|), and calculate the credibility factor α_obd of the OBD data and the credibility factor α_ultra of the ultrasonic flow data respectively;

[0056] For example, when the OBD data delay Δt = 30ms and the vibration intensity a = 0.8g, β = 0.01 + 0.008×(0.8 - 0.5) = 0.0124,

[0057] α_obd = 1 / (5% 2 )×e^(-0.0124×30) ≈ 400×0.69 ≈ 276.

[0058] Dynamic weight allocation The weight allocation unit adjusts and outputs the fusion weight according to the following steps:

[0059] Vibration intensity judgment: Obtain the total vibration intensity a of the vibration sensor in real time. If a > 3.0g, trigger the weight downscaling mechanism; if a ≤ 3.0g, maintain the current credibility factor unchanged.

[0060] Weight downscaling execution: After triggering the downscaling, use the exponential decay method to gradually adjust the weight to avoid sudden changes. The OBD weight is reduced from the current value to 6% (downscaled by 40%), and the ultrasonic flowmeter weight is reduced to 80% (downscaled by 20%). The downscaling process lasts for 200ms, and the decay formula is:

[0061] α(t) = α0 × e^(-t / τ), where τ = 50ms and t is the downscaling time (0 ≤ t ≤ 200ms).

[0062] Time delay compensation: Calculate the time delay compensation coefficient k_obd=e^(-0.01·Δt) for OBD data and the time delay compensation coefficient k_ultra=e^(-0.005·Δt) for ultrasonic flow data. Multiply these coefficients by the corresponding reliability factors to obtain the compensated reliability factors α_obd'=α_obd×k_obd and α_ultra'=α_ultra×k_ultra.

[0063] Normalization: The compensated credibility factor is normalized, and the fusion weights w_obd=α_obd' / (α_obd'+α_ultra') and w_ultra=α_ultra' / (α_obd'+α_ultra') are output, ensuring that w_obd+w_ultra=1. For example, if after compensation α_obd'=250 and α_ultra'=600, then w_obd=250 / (250+600)≈0.29 and w_ultra≈0.71.

[0064] Kalman Adaptive Fusion: The Kalman adaptive fusion module performs data fusion according to the following equation: in:

[0065] y_obd is the preprocessed OBD fuel data, and y_ultra is the preprocessed ultrasonic flow data;

[0066] Kk is the gain matrix, which is dynamically adjusted according to the confidence level. The adjustment formula is: Where Pk is the state covariance matrix, Rk is the observation noise matrix, and H is the observation matrix (H = [1]);

[0067] zk is the observed value of the working condition disturbance provided by the vibration sensor, which is converted to zk = 0.01 × a through vibration intensity a;

[0068] Fusion results This is the final fuel consumption data, output at a frequency of 50Hz. It also stores the original data and the fusion result for subsequent verification.

[0069] (III) Testing, Verification and Application Deployment

[0070] Test environment and equipment

[0071] Test vehicles: Three mainstream models were selected: Dongfeng Tianlong KL heavy truck (China VI diesel engine, rated power 350kW, fuel tank capacity 600L), Foton Aumark light truck (China VI gasoline engine, rated power 120kW, fuel tank capacity 120L), and Wuling Hongguang minivan (China VI gasoline engine, rated power 73kW, fuel tank capacity 45L), covering mobile sources with different loads and power types.

[0072] Test conditions include: smooth cruising (vehicle speed 80km / h, engine speed 1500rpm, vibration acceleration ≤0.5g), rapid acceleration and deceleration in mountainous areas (vehicle speed 30-90km / h alternating, engine speed 1200-2500rpm, peak vibration acceleration 4.2g), low temperature conditions (ambient temperature -25℃, started after 12 hours of standing, and ran continuously for 4 hours), high load conditions (fully loaded climbing, slope 15°, continuous driving for 30 minutes), and long-term stability conditions (continuous operation for 30 days, 8 hours a day, covering urban delivery, trunk transportation and other scenarios).

[0073] Test equipment: high-precision fuel consumption meter (accuracy ±0.1%, used as a standard value for comparison), NI cDAQ-9178 data acquisition instrument (sampling rate 100Hz, recording raw data), high-precision weather station (monitoring ambient temperature, humidity, and air pressure), vibration table (simulating vibrations of different intensities to verify the weighting mechanism).

[0074] Test Results and Analysis

[0075] Basic operating condition test: Under stable cruise conditions, the average error of the system of this invention is 0.41%, the error of a single OBD is 5.7%, and the error of a single ultrasonic flow meter is 0.35%. In terms of peak error during rapid acceleration, the system of this invention is only 0.63%, which is much lower than the 22.3% of a single OBD and the 1.8% of a single ultrasonic flow meter, meeting the requirement of MRV error ≤1%.

[0076] Extreme operating condition test: Under low temperature conditions (-25℃), the average error of the system of the present invention is 0.52%, the error of a single OBD is 8.3%, and the error of a single ultrasonic flow meter is 2.1%; under high load conditions, the average error of the system of the present invention is 0.47%, the peak error of rapid acceleration is 0.71%, and the error of a single ultrasonic flow meter is 2.3%, demonstrating excellent adaptability to operating conditions.

[0077] Long-term stability test: During 30 days of continuous operation, the error drift of the system of this invention was ≤0.05%, the data transmission success rate was 99.9%, and there were no sensor fault alarms, meeting the requirements for long-term industrial operation. The test report (No. CTI-2024-0217) issued by the National Institute of Metrology of China verifies that the maximum measurement error of the system is 0.48%, which fully complies with the relevant standards of carbon trading MRV.

[0078] In practical applications, this invention has been deployed on logistics vehicles of numerous logistics companies nationwide, covering multiple provinces and cities across the country, and involving various scenarios such as trunk transportation, urban delivery, and cold chain transportation. Adaptation solutions for different vehicle models are as follows:

[0079] Heavy-duty truck models: The ultrasonic flow meter is installed on the main fuel line at the frame beam, and the vibration sensor is installed in the engine block;

[0080] Light truck models: The ultrasonic flow meter is installed at the fuel filter outlet, and the vibration sensor is installed in the gearbox housing;

[0081] Minivan: The ultrasonic flow meter adopts a miniaturized design and is installed near the oil outlet of the fuel tank, while the vibration sensor is installed on the inner wall of the engine compartment.

[0082] The deployment has yielded significant results: for over 3,000 long-haul logistics vehicles, the fuel consumption accounting dispute rate dropped from 12% before deployment to 0.8% after deployment. The system enables real-time integration of fuel consumption data with refueling data from transportation companies, improving accounting efficiency by 40% and reducing verification time from 3 days to 1 day. To date, the system has accumulated over 12 million kilometers of operation, reducing fuel consumption accounting disputes by over 20 million tons annually, providing stable and reliable fuel consumption measurement support for vehicle fuel consumption accounting and the mobile source carbon trading market.

[0083] Maintenance and Calibration Program

[0084] Regular calibration: The sensor is calibrated on-site every 6 months, using a standard fuel flow generator (accuracy ±0.05%) as a reference, and the historical error standard deviation σi of the OBD and ultrasonic flow meter is adjusted; the gain matrix parameters of the Kalman fusion algorithm are optimized remotely every quarter to adapt to changes in different operating conditions.

[0085] Routine maintenance: Check the sensor installation status monthly to ensure there is no looseness or leakage; clean the probe surface of the ultrasonic flow meter every 3 months to remove oil and impurities; the local data of the embedded processor is automatically backed up to the cloud every 7 days to prevent data loss; the system supports remote diagnostics, can monitor the working status of each component in real time, and automatically alarm and push maintenance prompts when a fault occurs.

[0086] Test Results

[0087] The test results are shown in the table below. The system of this invention exhibits excellent measurement accuracy under both operating conditions and meets MRV compliance requirements:

[0088] Monitoring methods average error Rapid acceleration peak error MRV compliance Single OBD 5.7％ 22.3％ × Single ultrasound 0.35％ 1.8％ △ This invention system 0.41％ 0.63％ √

[0089] Actual measurements have verified that the maximum measurement error of the system of this invention is 0.48%, which complies with the relevant standards of carbon trading MRV.

[0090] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A multi-source heterogeneous sensor mobile source fuel consumption dynamic credibility evaluation system, characterized in that, include: The heterogeneous sensor group includes an OBD interface module, an ultrasonic flow meter with an accuracy of 0.2, and a triaxial vibration sensor, which are used to collect OBD fuel data, ultrasonic flow data, and vibration acceleration data. A dynamic credibility evaluation engine, an input end of which is connected with an output end of the heterogeneous sensor group, is used for calculating a real-time credibility factor α based on a historical error standard deviation σ i of the sensor, a data time delay Δt and a vibration intensity i . The weight allocation unit, whose input is connected to the output of the dynamic credibility evaluation engine, is used to dynamically adjust and output the normalized fusion weights [α_obd, α_ultra] based on the vibration intensity and data delay. The Kalman adaptive fusion module, with its input connected to the output of the weight allocation unit, is used to combine the observed values ​​of operating condition disturbances to output fuel consumption data that meets the MRV requirements.

2. The dynamic reliability evaluation system for fuel consumption of a multi-source heterogeneous sensor mobile source according to claim 1, characterized in that: The credibility factor a i The calculation model for: α i = 1 / σ i 2 × e^(-β·|Δt|); where σ i is the standard deviation of the sensor historical error, σ1=5% for the OBD interface module and σ2=0.3% for the ultrasonic flow meter; Δt is the data time delay (ms); and β is the attenuation coefficient, which is positively correlated with the vibration intensity collected by the vibration sensor.

3. A dynamic weight allocation method based on the system of claim 1, characterized in that, Includes the following steps: S1: Obtain the acceleration value collected by the triaxial vibration sensor; S2: If the acceleration value is >3.0g, trigger the OBD weight reduction by 40% and the ultrasonic flow meter weight reduction by 20%; S3: Calculate the OBD data delay compensation coefficient e^(-0.01·Δt) and the ultrasonic flow meter data delay compensation coefficient e^(-0.005·Δt), and apply them to the corresponding reliability factors respectively; S4: Normalize the adjusted credibility factor and output the fusion weights [α_obd, α_ultra] to the Kalman adaptive fusion module.

4. A Kalman adaptive fusion method based on the system of claim 1, characterized in that: The fusion equation is: where y_obd is the OBD fuel data, y_ultra is the ultrasonic flow data, K k is a gain matrix dynamically adjusted with the credibility, z k is the working condition disturbance observation value provided by the vibration sensor, and H is an observation matrix.

5. The application of the system described in any one of claims 1-2, the dynamic weight allocation method described in claim 3, or the Kalman adaptive fusion method described in claim 4 in carbon trading MRV.

Images