Vehicle fuel consumption analysis and early warning system and method based on vehicle-cloud data fusion

Abstract

This invention discloses a vehicle fuel consumption analysis and early warning system and method based on vehicle-cloud data fusion, relating to the field of intelligent connected vehicle operating status monitoring and fuel consumption early warning technology. The system includes a vehicle-side terminal and a cloud terminal connected by communication. The vehicle-side terminal collects on-board diagnostic data and inertial measurement data, extracts flat-road constant speed segment data after preprocessing, and forms data packets standardized according to 100 km / h, which are then uploaded to the cloud. The cloud terminal determines the validity of the data based on speed interval matching degree, generates segment confidence scores, updates the hierarchical benchmark, and calculates the linear regression slope, deviation value, and comprehensive anomaly score of fuel consumption change based on the fuel consumption values ​​per 100 km / h for each speed interval within a continuous statistical period. It then determines the anomaly type and whether it is a persistent or sudden anomaly, and further identifies candidate fault modules by combining auxiliary operating condition characteristics and maintenance manual rules. The system and method of this invention can improve the stability, comparability, and early warning reliability of fuel consumption analysis results.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle operation status monitoring and fuel consumption early warning technology, and in particular to a vehicle fuel consumption analysis and early warning system and method based on vehicle-cloud data fusion. Background Technology

[0002] With the development of vehicle networking, on-board diagnostics, and inertial measurement technologies, online analysis and health warnings of vehicle fuel consumption based on data collected from the vehicle have become an important direction for vehicle energy-saving management and condition monitoring. Existing technologies typically acquire operating data such as vehicle speed, fuel consumption, and mileage through on-board diagnostic interfaces, and combine this with historical statistical results to evaluate vehicle fuel consumption levels, thereby enabling fuel consumption anomaly identification, energy consumption management, and maintenance reminders.

[0003] However, current technologies for analyzing vehicle fuel consumption often rely on insufficiently filtered data from entire driving segments or simple comparisons based solely on average fuel consumption. This fails to effectively differentiate fuel consumption under various road conditions, speed ranges, and data quality levels. Particularly when vehicles experience complex conditions such as undulating roads, acceleration / deceleration, steering, and short-term fluctuations, directly including the corresponding data in the statistics can lead to inconsistent fuel consumption analysis standards, insufficient sample representativeness, and distorted benchmark comparisons. This results in low accuracy in anomaly identification, large fluctuations in warning results, and biased fault correlation.

[0004] Furthermore, existing solutions often lack an integrated processing mechanism for judging fuel consumption anomalies, encompassing effective data screening, benchmark updates, anomaly identification, and fault correlation. This results in insufficient validity of front-end data and a lack of reliable connection between back-end analysis results. Once invalid or low-reliability segments enter the analysis process, it not only contaminates the benchmark database but also leads to misjudgments of anomaly types and inaccurate maintenance recommendations, thereby affecting users' trust in the warning results and the effectiveness of subsequent maintenance decisions. Therefore, how to uniformly screen, evaluate the validity of, and update the hierarchical benchmark for flat road constant speed segment data under vehicle-cloud collaboration, and on this basis, achieve stable and accurate fuel consumption anomaly warnings, has become a core problem that existing technologies urgently need to solve. Summary of the Invention

[0005] To address the problems of inconsistent data standards, distorted fuel consumption analysis, insufficient accuracy of warnings, and unreliable fault correlation in existing fuel consumption warning systems, this invention proposes a vehicle fuel consumption analysis and warning system and method based on vehicle-cloud data fusion.

[0006] The present invention achieves the above objectives through the following technical solutions:

[0007] A vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion, including a vehicle terminal and a cloud terminal with communication connection;

[0008] The vehicle terminal is used to collect on-board diagnostic data and inertial measurement data. After preprocessing, it extracts flat road uniform speed segment data and forms data packets according to 100 km / h standardization and uploads them to the cloud.

[0009] The cloud includes:

[0010] The speed interval matching degree calculation module is used to calculate the interval assignment accuracy and the speed fluctuation coefficient within the interval based on the flat road uniform speed segment data, and then calculate the comprehensive matching degree, and determine the validity of the flat road uniform speed segment data based on the comprehensive matching degree.

[0011] The fragment confidence generation module is used to generate fragment confidence based on the comprehensive matching degree, the completeness of uploaded data, and the stability within a continuous statistical period.

[0012] The hierarchical benchmark library module is used to assign valid flat road uniform speed segment data to speed interval weights and update the hierarchical benchmark with segment confidence as the weight for entry into the library.

[0013] The anomaly warning module is used to calculate the linear regression slope of fuel consumption change, the deviation of the current fuel consumption value per 100 kilometers from the stratified benchmark, and the comprehensive anomaly score based on the fuel consumption value per 100 kilometers in each speed range within a continuous statistical period. When the comprehensive anomaly score is not lower than the preset anomaly threshold, the anomaly type is determined. Then, based on the number of consecutive verifications or the sudden changes in adjacent consecutive statistical periods, the continuous anomaly or sudden anomaly is determined.

[0014] The fault association module is used to determine candidate fault modules based on the type of abnormality, whether the abnormality is persistent or sudden, auxiliary operating condition characteristics, and the fault module association rules corresponding to the vehicle repair manual.

[0015] The results output module is used to output a comprehensive score, ranking among vehicles of the same model, defeat percentage, fuel consumption anomaly warning information, and vehicle health warning information based on the comprehensive anomaly score and candidate fault module.

[0016] As a preferred embodiment of the present invention, the vehicle end includes:

[0017] The acquisition module is used to acquire the on-board diagnostic data and inertial measurement data and attach timestamps;

[0018] The preprocessing module is used to perform time alignment and first-order low-pass filtering on the vehicle diagnostic data and inertial measurement data;

[0019] The attitude self-calibration module is used to estimate the direction of gravity based on preprocessed inertial measurement data and solve the rotation relationship between the sensor coordinate system and the vehicle coordinate system to obtain attitude self-calibration data.

[0020] The slope consistency verification module is used to estimate the inertial slope based on the preprocessed attitude self-calibration data, and compare the inertial slope with the external slope obtained from map elevation data and / or satellite positioning elevation data to obtain the slope consistency difference.

[0021] The segment extraction module is used to extract flat road uniform speed segment data that meets preset working conditions by using a sliding window based on preprocessed vehicle diagnostic data, inertial measurement data, and the slope consistency difference.

[0022] A caching module is used to cache the flat-road uniform speed segment data when communication is interrupted;

[0023] The upload module is used to upload the flat road uniform speed segment data to the cloud in a standardized data packet format of 100 kilometers.

[0024] As a preferred embodiment of the present invention, the preset operating conditions specifically satisfy the following conditions simultaneously:

[0025] The slope consistency difference is less than a preset slope threshold;

[0026] The absolute value of the horizontal acceleration within the sliding window is less than the preset horizontal acceleration threshold.

[0027] The absolute value of the difference between the vertical acceleration and the gravitational acceleration within the sliding window is less than the preset vertical acceleration threshold;

[0028] The total angular velocity within the sliding window is less than a preset angular velocity threshold.

[0029] The vehicle speed fluctuation within the sliding window is less than the preset vehicle speed fluctuation threshold;

[0030] Among them, flat road uniform speed segment data that do not meet the preset working conditions, have zero speed or abnormal fuel consumption are all removed.

[0031] The vehicle terminal will classify and cache the removed flat road constant speed segment data into speed ranges of 0-20km / h, 20-40km / h, 40-60km / h, 60-80km / h, 80-100km / h and above 100km / h, and output a standardized data packet for 100km after calculating the average fuel consumption of each speed range and the overall average fuel consumption when the cumulative mileage reaches 100km.

[0032] As a preferred embodiment of the present invention, the speed interval matching degree calculation module is used to determine the validity of the flat road uniform speed segment data according to the following rules:

[0033] A match is considered valid when the overall matching degree is not lower than the first matching threshold.

[0034] When the overall matching degree is lower than the first matching threshold but not lower than the second matching threshold, a supplementary judgment is made based on the driving time. If the driving time is not lower than the preset time threshold, it is determined to be valid.

[0035] When the overall matching degree is lower than the second matching threshold, it is determined to be invalid;

[0036] The first matching threshold is 85%, the second matching threshold is 70%, and the preset duration threshold is 30 seconds.

[0037] As a preferred embodiment of the present invention, the formula for calculating the interval assignment accuracy is as follows: ;

[0038] The formula for calculating the velocity fluctuation coefficient within the specified interval is as follows: ;

[0039] The formula for calculating the overall matching degree is: ;

[0040] in, Indicates the first The data for the flat road uniform speed segment corresponds to the first... The accuracy of assigning intervals to speed ranges; Indicates the first The data from the flat, uniform speed segment falls into the first... The number of sampling points in each speed range; Indicates the first The total number of sampling points for a flat road uniform speed segment data; Indicates the first The velocity fluctuation coefficient within the interval of a flat road uniform speed segment data; Indicates the first Standard deviation of vehicle speed for a flat road uniform speed segment; Indicates the first The average vehicle speed of a single flat road constant speed segment; Indicates the zero constant; Indicates the first The data for the flat road uniform speed segment corresponds to the first... Overall matching degree of each speed range; The weighting coefficients representing the accuracy of interval assignment. This represents the weighting coefficient of the velocity fluctuation coefficient within the interval.

[0041] In a preferred embodiment of the present invention, the hierarchical benchmark library module updates the hierarchical benchmark using a speed interval weighted allocation method, and the calculation formula is as follows:

[0042] ;

[0043] ;

[0044] In the formula, Indicates the first The data of the flat road uniform speed segment are compared with the first Weighted affiliation degree for each speed interval; Indicates the first The representative vehicle speed of a flat, uniform speed segment; Indicates the first The center value of each speed range; Represents any speed range The central value; This represents the speed range weighted assignment softening parameter; Indicates the total number of speed ranges; Indicates the first The hierarchical benchmark before the speed range update Indicates the first The hierarchical benchmark updated for each velocity range; Indicates the first Update gain for each speed range; This represents the weighted median operator; Indicates the first The fuel consumption value per 100 kilometers corresponding to the data of a flat road constant speed segment; Indicates the first Fragment confidence of a uniform speed segment on a flat road; Indicates the first The data of the flat road constant speed segment is written into the first... The weight of inbound data for each speed range.

[0045] As a preferred embodiment of the present invention, the formula for calculating the slope of the linear regression of fuel consumption change is as follows: ;

[0046] The formula for calculating the deviation between the current fuel consumption per 100 kilometers and the stratified benchmark is as follows: ;

[0047] The formula for calculating the comprehensive anomaly score is as follows: ;

[0048] In the formula, Indicates the first The slope of the linear regression of fuel consumption changes in each speed range; Indicates the total number of consecutive statistical periods; Indicates the first Within the 1st consecutive statistical period Fuel consumption per 100 kilometers for each speed range; Indicates the number of times within the current statistical period The deviation between the current fuel consumption per 100 kilometers in each speed range and the stratified benchmark; Indicates the number of times within the current statistical period Fuel consumption per 100 kilometers for each speed range; Indicates the first A tiered benchmark for each speed range; Indicates the first The overall anomaly score for each speed range; The weighting coefficients represent the slope of the linear regression of fuel consumption changes; The weighting coefficients represent the deviation values; This represents the normalization constant of the linear regression slope of fuel consumption changes; This represents the normalization constant for the deviation value;

[0049] When the comprehensive abnormal score When the value is not lower than the preset abnormal threshold, the first one is determined. The speed range has entered an abnormal judgment state;

[0050] Under the aforementioned abnormal condition, when the slope of the linear regression of fuel consumption changes... The deviation value is greater than the first slope threshold. When the deviation exceeds the first deviation threshold, it is determined to be the first abnormal type;

[0051] When the slope of the linear regression of fuel consumption changes The deviation value is greater than the second slope threshold and less than or equal to the first slope threshold. If the deviation exceeds the second deviation threshold, it is classified as the second abnormal type.

[0052] When deviation value If the value is less than the preset low threshold, it is determined to be the third type of abnormality;

[0053] When the same speed range meets the same anomaly type judgment condition within a consecutive statistical period of no less than a preset number of consecutive times, and the corresponding comprehensive anomaly score If all values ​​are not lower than the preset abnormal threshold, the condition is determined to be a persistent abnormality of the corresponding abnormal type.

[0054] When the absolute value of the difference between the fuel consumption values ​​per 100 kilometers in two consecutive statistical periods exceeds the preset mutation threshold, and the corresponding comprehensive matching degree is not lower than the preset matching threshold, and the corresponding comprehensive anomaly score is... If the value is not lower than the preset abnormal threshold, it is determined to be a sudden abnormality of the corresponding abnormality type.

[0055] As a preferred embodiment of the present invention, the auxiliary operating condition characteristics include at least one or more of the following: oxygen sensor signal characteristics, intake air volume characteristics, throttle opening characteristics, engine speed fluctuation characteristics, ignition advance angle deviation characteristics, fuel pressure characteristics, and fuel injection quantity characteristics.

[0056] The fault module association rules include:

[0057] When a sudden anomaly is identified as the first type of anomaly, and the oxygen sensor signal voltage fluctuation range is greater than 0.8V and the stabilization time is less than 0.5 seconds, the oxygen sensor fault is identified as a candidate fault module.

[0058] When a persistent abnormality is identified as the first type of abnormality, and the intake air volume is less than 2.5g / s, the throttle opening is greater than 20%, and the corresponding speed range is 60-80km / h, the air filter blockage is identified as a candidate fault module.

[0059] When a persistent abnormality is identified as the second type of abnormality, and the engine speed fluctuation is greater than 50 r / min and the ignition advance angle deviation is greater than 2 crankshaft rotation angles, spark plug aging is identified as a candidate fault module.

[0060] When a sudden anomaly is identified as the third type of anomaly, and the fuel pressure is greater than 3.5 bar and the fuel injection quantity is less than 10 mg per cycle, the fuel consumption sensor is identified as a candidate fault module.

[0061] When multiple candidate fault modules are triggered simultaneously, they are sorted according to priority: sensor fault, intake or fuel system fault, ignition or engine mechanical fault.

[0062] As a preferred embodiment of the present invention, the result output module is used to generate a comprehensive score based on the overall fuel consumption per 100 kilometers accounting for 60% and the comprehensive matching degree of each speed range accounting for 40%, and compare the comprehensive score with the comprehensive score of vehicles of the same model to obtain the ranking and defeat percentage of the same model.

[0063] The result output module is also used to output the overall fuel consumption per 100 kilometers, the fuel consumption per 100 kilometers for each speed range, the comprehensive matching degree for each speed range, the comprehensive anomaly score, the candidate fault module ranking results, and the corresponding maintenance suggestions.

[0064] Vehicle fuel consumption analysis and early warning methods based on vehicle-cloud data fusion include:

[0065] Collect on-board diagnostic data and inertial measurement data, and after preprocessing, extract flat road uniform speed segment data and standardize it according to 100 km / h to form a data package for uploading to the cloud;

[0066] A hierarchical benchmark is established and updated according to vehicle type and speed range. The interval assignment accuracy and speed fluctuation coefficient within the interval are calculated based on the flat road uniform speed segment data. Then, the comprehensive matching degree is calculated, and the validity of the flat road uniform speed segment data is determined based on the comprehensive matching degree.

[0067] Based on the comprehensive matching degree, the completeness of uploaded data, and the stability within a continuous statistical period, a segment confidence score is generated. Only valid flat road uniform speed segment data are assigned according to speed interval weights, and the segment confidence score is used as the weight for entering the database to update the stratification benchmark.

[0068] The linear regression slope of fuel consumption change, the deviation of the current fuel consumption value per 100 kilometers from the stratified benchmark, and the comprehensive anomaly score are calculated based on the fuel consumption value per 100 kilometers in each speed range within a continuous statistical period. When the comprehensive anomaly score is not lower than the preset anomaly threshold, the anomaly type is determined. Then, the continuous anomaly or sudden anomaly is determined based on the number of consecutive verifications or the sudden change in adjacent consecutive statistical periods.

[0069] Candidate fault modules are determined based on the type of anomaly, whether it is a persistent or sudden anomaly, auxiliary operating condition characteristics, and the fault module association rules corresponding to the vehicle repair manual.

[0070] Based on the comprehensive anomaly score and candidate fault module output, the comprehensive score, the ranking of the same model, the percentage of defeat, fuel consumption anomaly warning information, and vehicle health warning information are provided.

[0071] The beneficial effects of this invention are: it achieves unified control over the analysis criteria for vehicle fuel consumption, effectively avoiding interference from complex operating conditions, boundary fluctuations, and low-reliability data on the benchmark database and analysis results, thereby significantly improving the stability, comparability, and reliability of fuel consumption evaluation results. Based on valid data and hierarchical benchmarks, it calculates the linear regression slope, deviation value, and comprehensive anomaly score of fuel consumption changes based on fuel consumption values ​​per 100 kilometers in each speed range within a continuous statistical period. After determining the anomaly type, it combines persistent or sudden anomalies with auxiliary operating condition characteristics and the fault module association rules corresponding to the vehicle's maintenance manual to determine candidate fault modules. Finally, it outputs a comprehensive score, ranking among similar vehicle models, percentage of improvement, fuel consumption anomaly warning information, and vehicle health warning information. This not only improves the accuracy and consistency of fuel consumption anomaly identification but also enhances the correspondence between anomaly results and candidate fault modules, making the warning results more targeted and interpretable, which helps users quickly locate problems and take corresponding maintenance measures. Attached Figure Description

[0072] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the modular structure of the system of the present invention; Figure 2 This is a flowchart of the method in an embodiment of the present invention. Detailed Implementation

[0073] 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0074] like Figure 1 As shown, this is one embodiment of the present invention, which provides a vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion, including a vehicle terminal and a cloud terminal connected by communication. The vehicle terminal collects on-board diagnostic data and inertial measurement data, extracts flat road constant speed segment data after preprocessing, and forms data packets standardized according to 100 km / h, which are then uploaded to the cloud. The cloud terminal performs validity determination, hierarchical benchmark updates, anomaly warnings, fault correlation, and result output on the uploaded data packets.

[0075] In one embodiment, the vehicle-side includes a data acquisition module, a preprocessing module, an attitude self-calibration module, a slope consistency verification module, a segment extraction module, a caching module, and an upload module, with each module connected in the order of data processing.

[0076] The data acquisition module is used to collect on-board diagnostic data and inertial measurement data, along with timestamps. During acquisition, on-board diagnostic data and inertial measurement data can use different sampling frequencies, but both should include timestamps under a unified time base to ensure subsequent time alignment. On-board diagnostic data preferably includes one or more of the following: vehicle speed, cumulative mileage, fuel consumption, throttle opening, engine speed, intake air volume, fuel pressure, and fuel injection quantity; inertial measurement data preferably includes triaxial acceleration and triaxial angular velocity.

[0077] The preprocessing module performs time alignment and first-order low-pass filtering on the vehicle diagnostic data and inertial measurement data. Time alignment can be achieved using nearest-neighbor resampling, linear interpolation, or unified time grid reconstruction to map data from different sources and frequencies onto a unified time series. The first-order low-pass filtering suppresses high-frequency noise and transient jitter in the inertial measurement data, and its output serves as input for subsequent attitude self-calibration and operational condition identification.

[0078] The attitude self-calibration module estimates the direction of gravity based on preprocessed inertial measurement data and solves for the rotation relationship between the sensor coordinate system and the vehicle coordinate system to obtain attitude self-calibration data. Specifically, during periods when the vehicle is stationary or in low-dynamic operation, the direction of gravity is estimated based on the average output of the accelerometer over a period of time. This is then used to determine the attitude deviation between the sensor mounting coordinate system and the vehicle coordinate system, generating rotation matrices, Euler angle parameters, or quaternion parameters. The purpose of attitude self-calibration is to ensure that the subsequently calculated horizontal acceleration, vertical acceleration, and total angular velocity are all based on the vehicle coordinate system, improving the accuracy of operational condition identification.

[0079] The slope consistency verification module estimates the inertial slope based on preprocessed attitude self-calibration data and compares the inertial slope with the external slope obtained from map elevation data and / or satellite positioning elevation data to obtain the slope consistency difference. The inertial slope can be calculated from the relationship between the gravity component and the vehicle's longitudinal direction in the vehicle coordinate system after attitude self-calibration. The external slope can be inferred from the changes in map elevation or satellite positioning elevation corresponding to the vehicle's current position; the difference between the two is the slope consistency difference. This module is used to exclude data segments that, although the speed is relatively stable, are actually in an uphill or downhill state, preventing non-flat road conditions from being included in subsequent fuel consumption statistics.

[0080] The segment extraction module is used to extract flat road uniform speed segment data that meets preset operating conditions based on preprocessed vehicle diagnostic data, inertial measurement data, and gradient consistency difference, using a sliding window. Preferably, the sliding window can adopt a scanning method with a fixed window length and a fixed step size. For example, the window length can be 3 to 10 seconds, and the step size can be 0.5 to 2 seconds. For each window, the vehicle speed fluctuation, horizontal acceleration, vertical acceleration, total angular velocity, and gradient consistency difference are calculated, and it is determined whether the preset operating conditions are met. Window that meets the conditions is marked as flat road uniform speed segment data. If adjacent windows that meet the conditions are temporally continuous, they can be merged into a longer flat road uniform speed segment data to reduce the number of fragmented segments.

[0081] The caching module is used to cache flat-path, uniform-speed data segments when communication is interrupted. The cache can be implemented using local non-volatile storage and queued for storage according to the time order of the data segments to be uploaded. After communication is restored, the caching module, in conjunction with the upload module, re-uploads the data that was not successfully uploaded in time order, and can include a re-upload flag to avoid duplicate writing to the cloud.

[0082] The upload module is used to upload flat-road constant-speed segment data to the cloud in a standardized 100-kilometer data package. The 100-kilometer standardized data package refers to calculating the overall average fuel consumption and the average fuel consumption for each speed range using a unified method based on a preset cumulative mileage, and then encapsulating the calculation results along with the corresponding vehicle model identifier, time identifier, and statistical fields to form a standardized input that can be directly processed by the cloud. Preferably, the upload module first summarizes the valid segment data within the current cumulative mileage range on the vehicle side, forming a data package including the overall average fuel consumption, average fuel consumption for each speed range, cumulative mileage, segment start and end times, and vehicle model identifier, before uploading it to the cloud. This significantly reduces the amount of raw point data transmission while ensuring the complete availability of key statistics required for cloud analysis.

[0083] In this embodiment, the preset operating conditions require that the following conditions be met simultaneously:

[0084] The slope uniformity difference is less than the preset slope threshold;

[0085] The absolute value of the horizontal acceleration within the sliding window is less than the preset horizontal acceleration threshold.

[0086] The absolute value of the difference between the vertical acceleration and the gravitational acceleration within the sliding window is less than the preset vertical acceleration threshold;

[0087] The total angular velocity within the sliding window is less than a preset angular velocity threshold.

[0088] The vehicle speed fluctuation within the sliding window is less than the preset vehicle speed fluctuation threshold.

[0089] Only when all five conditions above are met simultaneously will the data corresponding to the current window be recognized as flat road uniform speed segment data. This parallel constraint method ensures that the selected segment meets the flat road uniform speed requirements in terms of road gradient, attitude change, vehicle speed stability, and dynamic disturbance.

[0090] Preferably, the sliding window length is 5 seconds, and the sliding step size is 1 second; when the inertial measurement data sampling frequency is 100Hz, each sliding window contains 500 sampling points. The horizontal acceleration threshold is preferably no greater than 0.1g, the vertical acceleration threshold is preferably no greater than 0.1g, the total angular velocity threshold is preferably no greater than 0.05rad / s, and the vehicle speed fluctuation threshold is preferably no greater than 5km / h. The slope consistency poor threshold is preferably 1° to 2°. These thresholds can be slightly adjusted according to the vehicle model, sensor accuracy, and installation error, but all are within the scope of conventional engineering adjustments disclosed in this embodiment.

[0091] During the segment selection process, flat road uniform speed segment data that does not meet the preset operating conditions, flat road uniform speed segment data with zero speed, and flat road uniform speed segment data with abnormal fuel consumption are all removed. Zero speed refers to an average vehicle speed of zero within the sliding window or a vehicle speed of zero at the vast majority of sampling points; abnormal fuel consumption refers to a fuel consumption signal that is continuously zero, remains constant for a long period, or significantly exceeds the vehicle's physically reasonable range. Identification of abnormal fuel consumption can be achieved by setting upper and lower bound thresholds and continuous constant value detection rules. For example, when the fuel consumption value remains completely unchanged for multiple consecutive sampling points and does not change with vehicle speed or mileage, it can be determined as abnormal sensor output. Only the data that has been removed enters the subsequent interval buffering and standardized statistics.

[0092] The vehicle-side data, after being filtered out, is categorized and cached into speed ranges: 0-20 km / h, 20-40 km / h, 40-60 km / h, 60-80 km / h, 80-100 km / h, and above 100 km / h. Preferably, a separate cache queue is set up for each speed range on the vehicle-side. For each segment of data representing a uniform speed on a flat road, its speed range can be determined based on the average vehicle speed of that segment, and the segment is then written into the corresponding cache queue. The purpose of categorizing and caching speed ranges on the vehicle-side is to facilitate subsequent separate statistical analysis of fuel consumption levels for each speed range and to maintain consistency with the cloud-based stratification benchmark.

[0093] When the cumulative mileage reaches 100km, the vehicle calculates the average fuel consumption for each speed range and the overall average fuel consumption, and outputs a standardized data packet for each 100km range. Specifically, for any speed range, the cumulative fuel consumption and cumulative mileage corresponding to all valid flat-road constant-speed segments within that speed range are statistically analyzed. The fuel consumption value for that speed range is obtained by dividing the cumulative fuel consumption by the cumulative mileage and then multiplying by 100. The same calculation is performed after summing the data for all speed ranges to obtain the overall average fuel consumption. The standardized data packet for each 100km range includes at least: the overall fuel consumption value for each 100km range, the fuel consumption value for each speed range, the number of valid samples for each speed range, the cumulative mileage, the start and end times, the vehicle model identifier, and the vehicle identifier. If the number of valid segments for a certain speed range is insufficient within the current 100km cycle, a missing or insufficient sample identifier can be retained in the data packet for the cloud to determine whether to participate in the hierarchical benchmark update.

[0094] Preferably, when the cumulative mileage has not yet reached 100km, the vehicle continues to cache valid flat road uniform speed segment data without outputting standardized data packets; when communication is interrupted, standardized data packets that have been formed but not successfully uploaded are temporarily stored by the caching module and re-uploaded by the uploading module after the network is restored.

[0095] In this embodiment, the various modules on the vehicle side can be implemented without relying on a training model. If a data-driven algorithm is used for attitude estimation or filtering enhancement in a specific implementation, it should be ensured that the input is inertial measurement data and time information, and the output is attitude self-calibration parameters or filtered inertial data, without changing the basic connection relationships of acquisition, preprocessing, attitude self-calibration, slope consistency verification, segment extraction, caching, and uploading.

[0096] In another preferred embodiment, the cloud includes at least a speed interval matching degree calculation module, a segment confidence degree generation module, a hierarchical benchmark library module, an anomaly warning module, a fault association module, and a result output module.

[0097] The speed interval matching degree calculation module is used to determine the validity of flat road uniform speed segment data. Specifically, it calculates the interval assignment accuracy based on the proportion of sampling points in the flat road uniform speed segment data falling into the target speed interval, calculates the speed fluctuation coefficient within the interval based on the standard deviation and average value of traffic congestion within the segment, and calculates the comprehensive matching degree accordingly. The comprehensive matching degree is used to characterize the representativeness and stability of the flat road uniform speed segment data within the target speed interval. Only when the comprehensive matching degree meets the preset requirements is the flat road uniform speed segment data recognized as valid data.

[0098] The formula for calculating the interval assignment accuracy is: ;

[0099] The formula for calculating the velocity fluctuation coefficient within the interval is: ;

[0100] The formula for calculating the overall matching degree is: ;

[0101] In the formula, Indicates the first The data for the flat road uniform speed segment corresponds to the first... The accuracy of assigning intervals to speed ranges; Indicates the first The data from the flat, uniform speed segment falls into the first... The number of sampling points in each speed range; Indicates the first The total number of sampling points for a flat, uniform speed segment of data; due to and Both represent the number of sampling points; both are counts and are dimensionless. It is a dimensionless quantity, with a value range from 0 to 1.

[0102] Indicates the first The velocity fluctuation coefficient within the interval of a flat road uniform speed segment data; Indicates the first Standard deviation of vehicle speed for a flat road uniform speed segment; Indicates the first The average vehicle speed of a single flat road constant speed segment; Indicates a zero constant; preferably, , If all units are km / h, then the numerator and denominator have the same dimensions, and the velocity fluctuation coefficient... It is a dimensionless quantity. (Prevents zero constant) To avoid division by zero at extremely low average vehicle speeds, the unit is consistent with the vehicle speed unit, preferably 0.1 km / h to 1 km / h. In this embodiment, since segments with zero speed have already been removed from the vehicle-side data of the flat road uniform speed segments, therefore... It is mainly used to improve the stability of numerical calculations.

[0103] Indicates the first The data for the flat road uniform speed segment corresponds to the first... Overall matching degree of each speed range; The weighting coefficients representing the accuracy of interval assignment. This represents the weighting coefficient for the velocity fluctuation coefficient within the interval. Because... and All are dimensionless quantities. , It is also a dimensionless weighting coefficient, therefore It is a dimensionless quantity. Preferably, For example, it is advisable , This ensures that the overall matching degree takes into account both interval belonging and speed stability; of course, the weights of this group can also be adjusted by engineering according to the vehicle model or application scenario.

[0104] Among them, the standard deviation of vehicle speed The average vehicle speed can be calculated using conventional statistical formulas. The average speed of all sampling points within a sliding window can be used to obtain the data. If the uniform speed segment data on a flat road is formed by merging multiple consecutive windows, all sampling points after merging can be used as the first... The calculation is performed using a sample set of data from a flat, uniform speed segment. If the same segment needs to be evaluated across multiple speed ranges, the corresponding calculation is performed separately for each speed range. and .

[0105] The speed interval matching degree calculation module determines the validity of flat road uniform speed segment data according to the following three rules:

[0106] When the overall matching degree is not lower than the first matching threshold, the data of the flat road uniform speed segment is directly determined to be valid;

[0107] When the overall matching degree is lower than the first matching threshold but not lower than the second matching threshold, the flat road uniform speed segment data is considered to be in a critically reliable state and needs to be supplemented by the driving time. If the continuous driving time of the flat road uniform speed segment data is not lower than the preset time threshold, it is determined to be valid; otherwise, it is determined to be invalid.

[0108] When the overall matching degree is lower than the second matching threshold, it is directly determined to be invalid.

[0109] In this embodiment, the first matching threshold is preferably set to 85%, the second matching threshold is preferably set to 70%, and the preset duration threshold is preferably set to 30 seconds. The reason for setting these three levels of judgment logic is as follows: when the overall matching degree is high, it means that the segment is highly representative of the target speed range and can be directly used for subsequent hierarchical benchmark updates; when the overall matching degree is in the middle range, although the segment has certain range fluctuations, if the duration is long enough, it can still be regarded as having statistical significance for the speed range, so it is allowed to be supplemented by "driving time" for confirmation; when the overall matching degree is too low, it means that the segment is neither stable nor representative and should not be included in the subsequent analysis chain.

[0110] In practical implementation, the travel time can be directly calculated from the start and end times of the flat road constant speed segment data. If a segment is formed by merging multiple continuous windows, the length of the merged continuous valid time is taken as the travel time. If there are invalid windows in the middle of a segment that exceed a preset interval time, the segment can be split into multiple sub-segments and judged separately. Those skilled in the art can use timestamp difference or window accumulation methods to calculate the travel time.

[0111] To facilitate subsequent fragment confidence generation and active re-sampling control, the validity determination result in this embodiment can be further encoded into a three-level status identifier:

[0112] The first-level state corresponds to direct validity;

[0113] The second-level state becomes effective after supplementary judgment;

[0114] Level 3 states are invalid.

[0115] In this way, subsequent modules can use either the binary determination result of valid / invalid data or the more granular status identifier to distinguish valid data from different sources, thereby improving the subsequent data quality control capabilities.

[0116] The segment confidence generation module generates segment confidence based on comprehensive matching degree, uploaded data completeness, and stability within a continuous statistical period. Uploaded data completeness reflects at least whether the overall average fuel consumption, average fuel consumption for each speed range, cumulative mileage, and necessary identifier fields are complete within the uploaded data packet. Stability within a continuous statistical period reflects whether the fuel consumption values ​​for the same vehicle in corresponding speed ranges show abnormal dispersion within adjacent statistical periods. A higher segment confidence indicates that the flat-road uniform speed segment data is more suitable for subsequent benchmark updates and anomaly analysis.

[0117] In one specific embodiment, the fragment confidence score is obtained by combining three scores, namely the overall matching score. Completeness score and stability score All three scores are restricted to a closed interval between 0 and 1, thus ensuring that the output of subsequent combined calculations remains within a comparable and uniform range.

[0118] Among them, the overall matching score The overall matching degree calculated by the speed range matching degree calculation module is normalized. If the overall matching degree is output as a decimal between 0 and 1, then the overall matching degree score is directly set. This equals the overall matching score; if the overall matching score is output as a percentage, it is normalized to the range of 0 to 1 by dividing the overall matching score by 100. Since the overall matching score itself is dimensionless, the normalized overall matching score... It is still a dimensionless quantity;

[0119] Completeness score This is used to characterize the completeness of preset essential fields in the uploaded data packet. Preferably, the preset essential fields include at least the overall fuel consumption per 100 kilometers, fuel consumption per 100 kilometers for each speed range, cumulative mileage, vehicle model identifier, time identifier, and the number of samples for each speed range. The completeness score can be calculated using the following formula: ; Indicates the first Each flat-road constant-speed segment of data corresponds to the actual number of pre-defined required fields in the uploaded data packet. This indicates the total number of pre-defined required fields. Because... and All represent the number of fields, and are dimensionless quantities. It is a dimensionless quantity, with a value range from 0 to 1. When all required fields are present... When a field is missing, Decrease proportionally;

[0120] Stability score This is used to characterize the fluctuation of fuel consumption per 100 kilometers for the same vehicle within the same speed range over several recent consecutive statistical periods. Preferably, the historical fuel consumption variation coefficient is calculated first. This is then mapped to a stability score. Specifically, the following formula can be used: ; ; Indicates the first The velocity interval number to which each uniform speed segment on a flat road belongs; Indicates the same vehicle in the most recent Within the 1st consecutive statistical period Historical standard deviation of fuel consumption per 100 kilometers for each speed range, superscript Indicates historical statistics; Indicates the same vehicle in the most recent Within the 1st consecutive statistical period Historical average fuel consumption per 100 kilometers for each speed range; Indicates the zero constant; This represents the stability normalization constant. Preferably, and If L / 100km is used as the unit, then It is a dimensionless quantity. It is also a dimensionless quantity, therefore As a dimensionless quantity, it satisfies the input requirements of an exponential function. The stability score is calculated accordingly. It is a dimensionless quantity between 0 and 1, and The smaller the value, the more stable the historical fuel consumption. The closer it is to 1.

[0121] In this embodiment, fragment confidence A weighted summation method can be used to calculate it: ;

[0122] in, Indicates the first Fragment confidence of a uniform speed segment on a flat road; , and These represent the weighting coefficients of the overall matching score, completeness score, and stability score, respectively. Preferably, , , Since all three scores are between 0 and 1, and the sum of the three weight coefficients is 1, the fragment confidence obtained by weighted summation is... It must also be between 0 and 1.

[0123] In another implementation, the fragment confidence level can also be calculated using a multiplicative combination method: The multiplicative combination method, compared to the weighted summation method, places greater emphasis on the "weakest link effect," meaning that when any one of the three scores is significantly low, the confidence level of the segment will decrease more noticeably. Regardless of whether the weighted summation method or the multiplicative combination method is used, this embodiment requires that only one method be consistently used within the same system to ensure consistency in scoring criteria.

[0124] In this embodiment, the segment confidence output range is explicitly limited to 0 to 1. Preferably, a high confidence threshold is set. Low confidence threshold .when When the corresponding flat-road uniform speed segment data is identified as high-confidence data, it is allowed to participate in both hierarchical benchmark updates and anomaly analysis; when At that time, it was identified as data with reduced weight, and only... As an input weight, it participates in the tiered benchmark update, and preferably is not used as the sole trigger for a new round of anomaly warnings; when In such cases, the data is classified as low-confidence data and excluded from stratified benchmark updates and anomaly analysis. This approach aims to fully utilize high-quality data, make limited use of medium-quality data, and directly isolate low-quality data, thereby reducing the interference of invalid or unstable data with the benchmark database and early warning results.

[0125] In the implementation process, the subsequent uses of fragment confidence include at least two aspects:

[0126] Firstly, the weighting factor used for updating the stratified baseline, together with the weighted affiliation degree of the speed interval, determines the contribution of a certain flat road uniform speed segment of data to the stratified baseline.

[0127] Secondly, as an admission control condition for subsequent anomaly analysis, it is used to shield low-confidence segments from interfering with the overall anomaly score and fault association rule matching.

[0128] With the above settings, fragment confidence is no longer just an auxiliary score, but directly affects the two main chains of hierarchical baseline update and anomaly analysis, thereby improving the overall robustness of the system.

[0129] Preferably, recently In a series of consecutive statistical periods The value can be 3 to 10, and is more preferably 5; the stability normalization constant The constant can be taken as 0.10 to 0.30, and is more preferably 0.15; zero constant is prevented. The values ​​can range from 0.01L / 100km to 0.10L / 100km. These parameters are standard engineering settings used by those skilled in the art based on vehicle data stability and do not affect the implementation of this invention.

[0130] The hierarchical benchmark library module maintains hierarchical benchmarks for different vehicle models and speed ranges. The hierarchical benchmarks are preferably stored using the median or weighted median. This module only updates valid flat-road constant-speed segments. During the update, the flat-road constant-speed segment data is written into the corresponding speed range according to its weighted affiliation, and then the segment confidence level is used as the weight for updating the hierarchical benchmark. This approach reduces the abrupt impact of speed range boundary segments on single-range benchmarks and minimizes the contamination of the benchmark library by low-confidence data.

[0131] In this embodiment, the hierarchical benchmark library module first constructs a hierarchical benchmark library. The hierarchical benchmark library is established according to vehicle type and speed range, with the speed range preferably set as six intervals: 0-20km / h, 20-40km / h, 40-60km / h, 60-80km / h, 80-100km / h, and above 100km / h. For any vehicle type and any speed range, the hierarchical benchmark library stores the statistical value of fuel consumption per 100 kilometers corresponding to historical valid flat road constant speed segments of that vehicle type within the corresponding speed range.

[0132] During the initial database construction phase, when a vehicle model accumulates at least a preset number of valid flat-road constant speed segment data within a certain speed range, the corresponding fuel consumption values ​​per 100 kilometers for these flat-road constant speed segment data are extracted, and the median is used as the initial stratification benchmark for that vehicle model in that speed range. The preset number of samples can be set according to the sample size of the vehicle model, for example, taking 20, 30, or 50 valid flat-road constant speed segment data. When the accumulated sample number does not reach the preset sample number, the initial stratification benchmark for that speed range is not generated temporarily, and subsequent valid flat-road constant speed segment data continues to be accumulated.

[0133] During the hierarchical benchmark update phase, only flat road constant speed segment data meeting the following conditions are allowed to enter the hierarchical benchmark update process: 1) it has been determined to be valid through speed interval matching; 2) the corresponding segment confidence score has been generated; and 3) a fuel consumption value per 100 kilometers has been generated. For flat road constant speed segment data that meets the conditions, the hierarchical benchmark library module calculates its weighted affiliation to each speed interval based on its representative vehicle speed, and generates the entry weight for each speed interval by combining the segment confidence score, and then updates the hierarchical benchmark for each speed interval, as shown in the following formula:

[0134] ;

[0135] ;

[0136] In the formula, Indicates the first The data of the flat road uniform speed segment are compared with the first Weighted affiliation degree for each speed interval; Indicates the first The representative vehicle speed of a flat, uniform speed segment; Indicates the first The center value of each speed range; Represents any speed range The central value; This represents the speed range weighted assignment softening parameter; Indicates the total number of speed ranges; preferably, , , , If all units are km / h, then the exponential function... It is a dimensionless quantity; the entire fraction is output. It is a dimensionless quantity and satisfies and for the same have ;

[0137] In this embodiment, Preferred selection of the first The average vehicle speed at all sampling points within a single flat road constant speed segment; cic_{i}ci can be taken as 10km / h, 30km / h, 50km / h, 70km / h, 90km / h, and 110km / h, respectively, corresponding to the center values ​​of the six speed ranges above 100km / h (0-20km / h, 20-40km / h, 40-60km / h, 60-80km / h, 80-100km / h, and 100km / h); Softening parameter The preferred speed is 5 km / h to 15 km / h. If Smaller values ​​indicate a closer approximation to a hard partition; if... A larger value indicates a higher degree of sharing between adjacent speed intervals; the choice can be made based on the data distribution.

[0138] Indicates the first The hierarchical benchmark before the speed range update Indicates the first The hierarchical benchmark updated for each velocity range; Indicates the first Update gain for each speed range; This represents the weighted median operator; Indicates the first The fuel consumption value per 100 kilometers corresponding to the data of a flat road constant speed segment; Indicates the first Fragment confidence of a uniform speed segment on a flat road; Indicates the first The data of the flat road constant speed segment is written into the first... Inbound weights for each speed range;

[0139] In terms of dimensions, , and All figures represent fuel consumption per 100 kilometers, with L / 100km being the preferred unit. It is a dimensionless quantity, and its value range is preferably from 0 to 1; and All are dimensionless quantities, therefore the weight for inclusion in the inventory is... It is also a dimensionless quantity; the weighted median operator Output dimensions and The same, also L / 100km.

[0140] In practice, the following steps can be followed:

[0141] Step 1: Read the current number to be updated. The representative vehicle speed of the flat road constant speed segment data fuel consumption per 100 kilometers and fragment confidence ;

[0142] Step 2: Calculate the weighted affiliation degree of the segment to each velocity range. ;

[0143] Step 3: Calculate the corresponding inbound weight for each speed range. ;

[0144] Step 4: Input the fuel consumption per 100 kilometers of the current set of segments to be updated and the corresponding entry weight into the weighted median operator to obtain the weighted median result for each speed range;

[0145] Step 5: Update the gain based on the previous stratification baseline. Calculate the updated stratified baseline.

[0146] The weighted median can be achieved using conventional methods in this field, namely, sorting the samples by fuel consumption per 100 kilometers and determining the median at the position where the cumulative weight reaches half of the total weight.

[0147] Preferably, update gain It can be set to a fixed value, such as 0.05 to 0.3; or it can be dynamically adjusted according to the number of new valid samples added in the current period within that speed range. If the number of new samples added in a certain speed range is small in the current statistical period, it can be reduced. To enhance the stability of the historical stratification benchmark; if there are many new samples, the value can be appropriately increased. This improves the sensitivity to baseline updates. Regardless of whether a fixed or dynamic approach is used, it essentially falls within the scope of conventional engineering implementations disclosed in this embodiment.

[0148] The hierarchical benchmark library module can be updated in real time when new 100km standardized data packets are received, or it can be updated in batches according to a preset statistical period. Regardless of whether real-time or batch updates are used, the corresponding hierarchical benchmark update operation is performed for each speed range.

[0149] The anomaly warning module calculates the linear regression slope of fuel consumption change, the deviation of the current fuel consumption per 100 kilometers from the stratified benchmark, and a comprehensive anomaly score based on the fuel consumption values ​​per 100 kilometers for each speed range within a continuous statistical period. The comprehensive anomaly score characterizes the combined degree of trend anomalies and deviation anomalies. When the comprehensive anomaly score is not lower than a preset anomaly threshold, the module enters anomaly judgment state. In the anomaly judgment state, the anomaly type is determined based on the combination of the slope and the deviation value. Then, combined with the number of consecutive verifications or the abrupt changes between adjacent consecutive statistical periods, it is determined whether the anomaly is persistent or sudden. Thus, the anomaly warning result includes information at both the anomaly type and anomaly pattern levels.

[0150] The formula for calculating the slope of the linear regression of fuel consumption changes is: ;

[0151] The formula for calculating the deviation between the current fuel consumption per 100 kilometers and the stratified benchmark is as follows: ;

[0152] The formula for calculating the comprehensive anomaly score is as follows: ;

[0153] In the formula, Indicates the first The slope of the linear regression of fuel consumption changes in each speed range; Indicates the total number of consecutive statistical periods; Indicates the first Within the 1st consecutive statistical period Fuel consumption per 100 kilometers in each speed range; preferably, Using L / 100km as the unit, because In this embodiment, the statistical period number is represented and is considered a dimensionless ordinal number. The dimension of this can be expressed as "L / 100km per statistical period"; if the statistical period adopts a fixed duration, such as updating once every 100km or every fixed number of days, then this slope represents the trend of fuel consumption in the corresponding speed range with the statistical period.

[0154] Indicates the number of times within the current statistical period The deviation between the current fuel consumption per 100 kilometers in each speed range and the stratified benchmark; Indicates the number of times within the current statistical period Fuel consumption per 100 kilometers for each speed range; Indicates the first A tiered benchmark for each speed range; Indicates the first The overall anomaly score for each speed range; The weighting coefficients represent the slope of the linear regression of fuel consumption changes; The weighting coefficients represent the deviation values; This represents the normalization constant of the linear regression slope of fuel consumption changes; This represents the normalization constant for the deviation value; preferably, For example, it is advisable , Different weights can be assigned to the trend and deviation terms depending on the scenario. Normalization constant. The unit should be with The same, meaning "L / 100km per statistical period"; normalization constant. The unit should be with The same, that is, L / 100km.

[0155] When the comprehensive abnormal score When the value is not lower than the preset abnormal threshold, the first one is determined. A speed range enters an anomaly detection state; preferably, the preset anomaly threshold can be 1.0, 1.2, or other dimensionless thresholds obtained by calibration using historical data. The purpose of the comprehensive anomaly score is to uniformly quantify trend anomalies and deviation anomalies. Only when the comprehensive degree of the two meets the preset requirements will the next step of anomaly type determination be carried out, thereby avoiding false alarms caused by single instantaneous fluctuations.

[0156] After entering the anomaly assessment state, the anomaly type is determined according to the following rules:

[0157] When the slope of the linear regression of fuel consumption changes The deviation value is greater than the first slope threshold. When the deviation exceeds the first deviation threshold, it is determined to be the first abnormal type;

[0158] When the slope of the linear regression of fuel consumption changes The deviation value is greater than the second slope threshold and less than or equal to the first slope threshold. If the deviation exceeds the second deviation threshold, it is classified as the second abnormal type.

[0159] When deviation value If the value is less than the preset low threshold, it is determined to be the third type of abnormality;

[0160] In this embodiment, the first anomaly type corresponds to a high-value anomaly with a significant upward trend and large deviation; the second anomaly type corresponds to a high-value anomaly with a slower upward trend but a deviation that has reached a preset level; and the third anomaly type corresponds to a low-value anomaly where the current fuel consumption value is significantly lower than the stratification benchmark. Preferably, each threshold can be obtained through statistical analysis of the historical distribution of the same vehicle model; for example, the first slope threshold can be higher than the second slope threshold, the first deviation threshold can be higher than the second deviation threshold, and the preset low-value threshold is a negative number.

[0161] After determining the exception type, further distinguish the exception mode:

[0162] When the same speed range meets the same anomaly type judgment condition within a consecutive statistical period of no less than a preset number of consecutive times, and the corresponding comprehensive anomaly score If all values ​​are not lower than the preset abnormal threshold, the condition is determined to be a persistent abnormality of the corresponding abnormality type.

[0163] When the absolute value of the difference between the fuel consumption values ​​per 100 kilometers in two consecutive statistical periods exceeds the preset mutation threshold, and the corresponding comprehensive matching degree is not lower than the preset matching threshold, and the corresponding comprehensive anomaly score is... If the value is not lower than the preset abnormal threshold, it is determined to be a sudden abnormality of the corresponding abnormality type.

[0164] Therefore, anomaly determination forms a two-layer structure: the first layer identifies the anomaly type, and the second layer identifies the anomaly pattern. This allows the anomaly results to reflect both the direction and degree of deviation, as well as the manner and temporal characteristics of occurrence. It is important to note that the comprehensive anomaly score is not merely used for ranking, but rather serves as an entry threshold for entering the anomaly determination state. It is also a common constraint for the establishment of persistent and sudden anomalies. Persistent and sudden anomalies are not independent classifications parallel to anomaly types, but rather a temporal pattern division based on the already determined anomaly types.

[0165] The fault association module is used to determine candidate fault modules based on anomaly type, persistent or sudden anomalies, auxiliary operating condition characteristics, and fault module association rules corresponding to the vehicle's repair manual. Auxiliary operating condition characteristics may include one or more of the following: oxygen sensor signal characteristics, intake air volume characteristics, throttle opening characteristics, engine speed fluctuation characteristics, ignition advance angle deviation characteristics, fuel pressure characteristics, and fuel injection quantity characteristics. The fault association module does not employ a black-box learning model; instead, it determines candidate fault modules based on confirmed diagnostic experience and rule relationships found in the vehicle's repair manual. Therefore, it can be directly implemented by those skilled in the art simply by reading the corresponding rules from the vehicle's repair manual.

[0166] The rules for associating faulty modules include:

[0167] When a sudden anomaly is identified as the first type of anomaly, and the oxygen sensor signal voltage fluctuation range is greater than 0.8V and the stabilization time is less than 0.5 seconds, the oxygen sensor fault is identified as a candidate fault module.

[0168] When a persistent abnormality is identified as the first type of abnormality, and the intake air volume is less than 2.5g / s, the throttle opening is greater than 20%, and the corresponding speed range is 60-80km / h, the air filter blockage is identified as a candidate fault module.

[0169] When a persistent abnormality is identified as the second type of abnormality, and the engine speed fluctuation is greater than 50 r / min and the ignition advance angle deviation is greater than 2 crankshaft rotation angles, spark plug aging is identified as a candidate fault module.

[0170] When a sudden anomaly is identified as the third type of anomaly, and the fuel pressure is greater than 3.5 bar and the fuel injection quantity is less than 10 mg / cycle, the fuel consumption sensor is identified as a candidate fault module.

[0171] Wherein, 0.8V represents the oxygen sensor signal swing threshold, 0.5 seconds represents the time threshold required for the oxygen sensor to reach a stable state, 2.5g / s represents the intake air mass flow rate threshold, 20% represents the throttle opening threshold, 50r / min represents the engine speed fluctuation threshold, 2 crankshaft rotation angles represents the ignition advance angle deviation threshold, 3.5bar represents the fuel pressure threshold, and 10mg / cycle represents the fuel injection quantity threshold. Each threshold can be adjusted according to the specific vehicle model's repair manual, but as long as the associated structure of abnormal warning result + auxiliary operating condition characteristics + repair rules remains unchanged, it falls within the scope of this embodiment.

[0172] When multiple candidate fault modules are triggered simultaneously, the fault association modules are prioritized according to the following order: sensor fault, intake or fuel system fault, and ignition or engine mechanical fault. This prioritization is based on the following reasons: sensor faults may directly cause abnormal fuel consumption readings or control feedback, and should be investigated first; intake or fuel system faults usually directly affect fuel consumption; and ignition or engine mechanical faults often manifest as continuous performance degradation. By pre-setting priorities, the system can provide a ranking result that better aligns with maintenance practices when multiple fault candidates exist simultaneously.

[0173] In this embodiment, to ensure the clarity and completeness of the implementation process, the following execution steps can be adopted:

[0174] Step 1: Read the exception type and exception mode output by the exception warning module;

[0175] Step 2: Read the auxiliary operating condition characteristics of the corresponding speed range within the current statistical period;

[0176] Step 3: Select the corresponding fault module association rule in the repair manual according to the vehicle model identification;

[0177] Step 4: Match each rule one by one and record the candidate fault modules that meet the conditions;

[0178] Step 5: Sort the candidate fault modules according to the preset priority and output the sorting results.

[0179] The results output module is used to output a comprehensive score, ranking among vehicles of the same model, defeat percentage, fuel consumption anomaly warning information, and vehicle health warning information based on the comprehensive anomaly score and candidate fault module.

[0180] In this embodiment, the result output module generates a comprehensive score based on 60% of the overall fuel consumption per 100 kilometers and 40% of the overall matching degree for each speed range. To ensure comparability, the overall fuel consumption per 100 kilometers and the overall matching degree for each speed range are preferably normalized before being weighted. For example, the overall fuel consumption per 100 kilometers can be mapped to a score range of 0 to 100 based on the best and worst values ​​from historical samples of the same vehicle model; the overall matching degree for each speed range can be directly expressed as a percentage or converted from 0 to 100 points; and then the comprehensive score is calculated as follows. : ;

[0181] in, This represents the fuel consumption score obtained by normalizing the overall fuel consumption per 100 kilometers. This represents the matching score obtained by summing up the overall matching degree of each speed range.

[0182] In a preferred implementation The result can be obtained by weighting the overall matching degree of each speed range according to the range weights; if it is not necessary to distinguish the range weights, a simple average can also be used. The range weights can be set according to the sample size of each speed range or the distribution of common working conditions of the same vehicle model, but regardless of the method used, the overall matching degree of each speed range should be normalized first, and then the weighting should be performed.

[0183] The results output module further compares the overall score with the overall scores of vehicles of the same model to obtain the ranking and percentage of vehicles outperformed within the same model. The ranking refers to the current vehicle's position within the sample set of vehicles of the same model after sorting by overall score; the percentage of vehicles outperformed indicates the proportion of vehicles of the same model whose overall score is higher than the current vehicle's. Preferably, the ranking and percentage of vehicles outperformed are calculated based on valid data of the same model within the most recent preset time range to maintain the timeliness and comparability of the comparison samples.

[0184] In addition, the results output module also outputs the overall fuel consumption per 100 kilometers, fuel consumption per 100 kilometers for each speed range, overall matching degree for each speed range, overall anomaly score, candidate fault module ranking results, and corresponding maintenance suggestions. Specifically, the overall fuel consumption per 100 kilometers and the fuel consumption per 100 kilometers for each speed range display the vehicle's current fuel consumption level; the overall matching degree for each speed range displays data quality and range representativeness; the overall anomaly score displays the current degree of anomaly; the candidate fault module ranking results indicate the priority direction for troubleshooting; and the corresponding maintenance suggestions are automatically generated based on preset maintenance instructions matched with the candidate fault modules. In this way, the results output module not only provides numerical evaluations but also explanations of causes and maintenance suggestions, facilitating understanding and handling by users and maintenance personnel.

[0185] In this embodiment, the output of the result output module can be displayed through an in-vehicle terminal, a mobile terminal, or a service backend interface, or it can be provided to an external system through message push, report generation, or interface return.

[0186] It should be noted that the comprehensive score, ranking, and defeat percentage are all calculated based solely on the standardized data packets formed from the flat road uniform speed segment data after validity assessment. When the number of valid samples in a certain speed range is insufficient, the comprehensive matching degree of that range can be displayed, but it can be weighted down or excluded from the calculation according to preset rules during scoring. The ranking results of candidate fault modules and maintenance recommendations are achieved through a preset mapping relationship of one-to-one or one-to-many correspondence.

[0187] like Figure 2 As shown, another embodiment of the present invention provides a vehicle fuel consumption analysis and early warning method based on vehicle-cloud data fusion, including the following steps:

[0188] S1: Collect on-board diagnostic data and inertial measurement data, extract flat road uniform speed segment data after preprocessing, and form data packets according to 100 km / h standardization and upload them to the cloud;

[0189] S2: Establish and update the hierarchical benchmark according to vehicle type and speed range, calculate the interval assignment accuracy and speed fluctuation coefficient within the interval based on the flat road uniform speed segment data, and then calculate the comprehensive matching degree, and determine the validity of the flat road uniform speed segment data based on the comprehensive matching degree.

[0190] S3: Based on the comprehensive matching degree, the completeness of uploaded data and the stability within the continuous statistical period, generate segment confidence scores, assign valid flat road uniform speed segment data to speed intervals, and update the stratification benchmark with segment confidence scores as the weight for entry into the database.

[0191] S4: Calculate the linear regression slope of fuel consumption change, the deviation of the current fuel consumption value per 100 kilometers from the stratified benchmark, and the comprehensive anomaly score based on the fuel consumption value per 100 kilometers in each speed range within a continuous statistical period. When the comprehensive anomaly score is not lower than the preset anomaly threshold, the anomaly type is determined. Then, based on the number of consecutive verifications or the sudden changes in adjacent consecutive statistical periods, the continuous anomaly or sudden anomaly is determined.

[0192] S5: Determine candidate fault modules based on anomaly type, persistent or sudden anomaly, auxiliary operating condition characteristics, and fault module association rules corresponding to the vehicle model repair manual.

[0193] S6: Outputs a comprehensive score, ranking among vehicles of the same model, percentage of vehicles defeated, fuel consumption anomaly warning information, and vehicle health warning information based on the comprehensive anomaly score and candidate fault modules.

[0194] The collection, transmission, storage, analysis, and use of vehicle operation data, driving-related data, and their derived statistical results involved in this application embodiment have all been agreed to by the user and comply with legal and regulatory requirements. For facial information, biometric information, user behavior data, and other personal information that may be involved in specific implementations, processing should be carried out according to the principle of minimum necessity after obtaining the corresponding authorization and consent, and data security should be ensured through measures such as de-identification, encrypted transmission, access control, permission isolation, and security auditing. The analysis, early warning, scoring, ranking, fault correlation, and suggestion generation rules in each embodiment are all based on vehicle operating status, working condition characteristics, statistical results, and maintenance rules. The algorithm design must not contain discriminatory rules, and must not make differentiated processing based on personal attributes unrelated to technical issues.

[0195] In summary, this invention improves the consistency of vehicle fuel consumption analysis, the accuracy of anomaly identification, and the reliability of early warning results by uniformly screening, evaluating the effectiveness of, updating the hierarchical benchmark, providing anomaly warnings, and associating faults with data from flat road uniform speed segments. This enables the invention to more accurately reflect the true fuel consumption status of vehicles and assist in maintenance decisions.

[0196] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all 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 vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion, characterized in that, This includes the vehicle-side and cloud-side communication connections; The vehicle terminal is used to collect on-board diagnostic data and inertial measurement data. After preprocessing, it extracts flat road uniform speed segment data and forms data packets according to 100 km / h standardization and uploads them to the cloud. The cloud includes: The speed interval matching degree calculation module is used to calculate the interval assignment accuracy and the speed fluctuation coefficient within the interval based on the flat road uniform speed segment data, and then calculate the comprehensive matching degree, and determine the validity of the flat road uniform speed segment data based on the comprehensive matching degree. The fragment confidence generation module is used to generate fragment confidence based on the comprehensive matching degree, the completeness of uploaded data, and the stability within a continuous statistical period. The hierarchical benchmark library module is used to assign valid flat road uniform speed segment data to speed interval weights and update the hierarchical benchmark with segment confidence as the weight for entry into the library. The anomaly warning module is used to calculate the linear regression slope of fuel consumption change, the deviation of the current fuel consumption value per 100 kilometers from the stratified benchmark, and the comprehensive anomaly score based on the fuel consumption value per 100 kilometers in each speed range within a continuous statistical period. When the comprehensive anomaly score is not lower than the preset anomaly threshold, the anomaly type is determined. Then, based on the number of consecutive verifications or the sudden changes in adjacent consecutive statistical periods, the continuous anomaly or sudden anomaly is determined. The fault association module is used to determine candidate fault modules based on the type of abnormality, whether the abnormality is persistent or sudden, auxiliary operating condition characteristics, and the fault module association rules corresponding to the vehicle repair manual. The results output module is used to output a comprehensive score, ranking among vehicles of the same model, defeat percentage, fuel consumption anomaly warning information, and vehicle health warning information based on the comprehensive anomaly score and candidate fault module.

2. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 1, characterized in that, The vehicle end includes: The acquisition module is used to acquire the on-board diagnostic data and inertial measurement data and attach timestamps; The preprocessing module is used to perform time alignment and first-order low-pass filtering on the vehicle diagnostic data and inertial measurement data; The attitude self-calibration module is used to estimate the direction of gravity based on preprocessed inertial measurement data and solve the rotation relationship between the sensor coordinate system and the vehicle coordinate system to obtain attitude self-calibration data. The slope consistency verification module is used to estimate the inertial slope based on the preprocessed attitude self-calibration data, and compare the inertial slope with the external slope obtained from map elevation data and / or satellite positioning elevation data to obtain the slope consistency difference. The segment extraction module is used to extract flat road uniform speed segment data that meets preset working conditions by using a sliding window based on preprocessed vehicle diagnostic data, inertial measurement data, and the slope consistency difference. A caching module is used to cache the flat-road uniform speed segment data when communication is interrupted; The upload module is used to upload the flat road uniform speed segment data to the cloud in a standardized data packet format of 100 kilometers.

3. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 2, characterized in that, The preset operating conditions specifically require that the following conditions be met simultaneously: The slope consistency difference is less than a preset slope threshold; The absolute value of the horizontal acceleration within the sliding window is less than the preset horizontal acceleration threshold. The absolute value of the difference between the vertical acceleration and the gravitational acceleration within the sliding window is less than the preset vertical acceleration threshold; The total angular velocity within the sliding window is less than a preset angular velocity threshold. The vehicle speed fluctuation within the sliding window is less than the preset vehicle speed fluctuation threshold; Among them, flat road uniform speed segment data that do not meet the preset working conditions, have zero speed or abnormal fuel consumption are all removed. The vehicle terminal will classify and cache the removed flat road constant speed segment data into speed ranges of 0-20km / h, 20-40km / h, 40-60km / h, 60-80km / h, 80-100km / h and above 100km / h, and output a standardized data packet for 100km after calculating the average fuel consumption of each speed range and the overall average fuel consumption when the cumulative mileage reaches 100km.

4. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 3, characterized in that, The speed interval matching degree calculation module is used to determine the validity of the flat road uniform speed segment data according to the following rules: A match is considered valid when the overall matching degree is not lower than the first matching threshold. When the overall matching degree is lower than the first matching threshold but not lower than the second matching threshold, a supplementary judgment is made based on the driving time. If the driving time is not lower than the preset time threshold, it is determined to be valid. When the overall matching degree is lower than the second matching threshold, it is determined to be invalid; The first matching threshold is 85%, the second matching threshold is 70%, and the preset duration threshold is 30 seconds.

5. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 4, characterized in that, The formula for calculating the interval assignment accuracy is as follows: ; The formula for calculating the velocity fluctuation coefficient within the specified interval is as follows: ; The formula for calculating the overall matching degree is: ; in, Indicates the first The data for the flat road uniform speed segment corresponds to the first... The accuracy of assigning intervals to speed ranges; Indicates the first The data from the flat, uniform speed segment falls into the first... The number of sampling points in each speed range; Indicates the first The total number of sampling points for a flat road uniform speed segment data; Indicates the first The velocity fluctuation coefficient within the interval of a flat road uniform speed segment data; Indicates the first Standard deviation of vehicle speed for a flat road uniform speed segment; Indicates the first The average vehicle speed of a single flat road constant speed segment; Indicates the zero constant; Indicates the first The data for the flat road uniform speed segment corresponds to the first... Overall matching degree of each speed range; The weighting coefficients representing the accuracy of interval assignment. This represents the weighting coefficient of the velocity fluctuation coefficient within the interval.

6. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 1, characterized in that, The hierarchical benchmark library module updates the hierarchical benchmark using a speed interval weighted attribution method, and the calculation formula is as follows: ； ； In the formula, Indicates the first The data of the flat road uniform speed segment are compared with the first Weighted affiliation degree for each speed interval; Indicates the first The representative vehicle speed of a flat, uniform speed segment; Indicates the first The center value of each speed range; Represents any speed range The central value; This represents the speed range weighted assignment softening parameter; Indicates the total number of speed ranges; Indicates the first The hierarchical benchmark before the speed range update Indicates the first The hierarchical benchmark updated for each velocity range; Indicates the first Update gain for each speed range; This represents the weighted median operator; Indicates the first The fuel consumption value per 100 kilometers corresponding to the data of a flat road constant speed segment; Indicates the first Fragment confidence of a uniform speed segment on a flat road; Indicates the first The data of the flat road constant speed segment is written into the first... The weight of inbound data for each speed range.

7. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 1, characterized in that, The formula for calculating the slope of the linear regression of fuel consumption change is: ; The formula for calculating the deviation between the current fuel consumption per 100 kilometers and the stratified benchmark is as follows: ; The formula for calculating the comprehensive anomaly score is as follows: ; In the formula, Indicates the first The slope of the linear regression of fuel consumption changes in each speed range; Indicates the total number of consecutive statistical periods; Indicates the first Within the 1st consecutive statistical period Fuel consumption per 100 kilometers for each speed range; Indicates the number of times within the current statistical period The deviation between the current fuel consumption per 100 kilometers in each speed range and the stratified benchmark; Indicates the number of times within the current statistical period Fuel consumption per 100 kilometers for each speed range; Indicates the first A tiered benchmark for each speed range; Indicates the first The overall anomaly score for each speed range; The weighting coefficients represent the slope of the linear regression of fuel consumption changes; The weighting coefficients represent the deviation values; This represents the normalization constant of the linear regression slope of fuel consumption changes; This represents the normalization constant for the deviation value; When the comprehensive abnormal score When the value is not lower than the preset abnormal threshold, the first one is determined. The speed range has entered an abnormal judgment state; Under the aforementioned abnormal condition, when the slope of the linear regression of fuel consumption changes... The deviation value is greater than the first slope threshold. When the deviation exceeds the first deviation threshold, it is determined to be the first abnormal type; When the slope of the linear regression of fuel consumption changes The deviation value is greater than the second slope threshold and less than or equal to the first slope threshold. If the deviation exceeds the second deviation threshold, it is classified as the second abnormal type. When deviation value If the value is less than the preset low threshold, it is determined to be the third type of abnormality; When the same speed range meets the same anomaly type judgment condition within a consecutive statistical period of no less than a preset number of consecutive times, and the corresponding comprehensive anomaly score If all values ​​are not lower than the preset abnormal threshold, the condition is determined to be a persistent abnormality of the corresponding abnormal type. When the absolute value of the difference between the fuel consumption values ​​per 100 kilometers in two consecutive statistical periods exceeds the preset mutation threshold, and the corresponding comprehensive matching degree is not lower than the preset matching threshold, and the corresponding comprehensive anomaly score is... If the value is not lower than the preset abnormal threshold, it is determined to be a sudden abnormality of the corresponding abnormality type.

8. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 7, characterized in that, The auxiliary operating condition characteristics include at least one or more of the following: oxygen sensor signal characteristics, intake air volume characteristics, throttle opening characteristics, engine speed fluctuation characteristics, ignition advance angle deviation characteristics, fuel pressure characteristics, and fuel injection quantity characteristics. The fault module association rules include: When a sudden anomaly is identified as the first type of anomaly, and the oxygen sensor signal voltage fluctuation range is greater than 0.8V and the stabilization time is less than 0.5 seconds, the oxygen sensor fault is identified as a candidate fault module. When a persistent abnormality is identified as the first type of abnormality, and the intake air volume is less than 2.5g / s, the throttle opening is greater than 20%, and the corresponding speed range is 60-80km / h, the air filter blockage is identified as a candidate fault module. When a persistent abnormality is identified as the second type of abnormality, and the engine speed fluctuation is greater than 50 r / min and the ignition advance angle deviation is greater than 2 crankshaft rotation angles, spark plug aging is identified as a candidate fault module. When a sudden anomaly is identified as the third type of anomaly, and the fuel pressure is greater than 3.5 bar and the fuel injection quantity is less than 10 mg per cycle, the fuel consumption sensor is identified as a candidate fault module. When multiple candidate fault modules are triggered simultaneously, they are sorted according to priority: sensor fault, intake or fuel system fault, ignition or engine mechanical fault.

9. The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion according to claim 5, characterized in that, The result output module is used to generate a comprehensive score based on the overall fuel consumption per 100 kilometers accounting for 60% and the comprehensive matching degree of each speed range accounting for 40%. The comprehensive score is then compared with the comprehensive score of vehicles of the same model to obtain the ranking and percentage of vehicles outperforming each other. The result output module is also used to output the overall fuel consumption per 100 kilometers, the fuel consumption per 100 kilometers for each speed range, the comprehensive matching degree for each speed range, the comprehensive anomaly score, the candidate fault module ranking results, and the corresponding maintenance suggestions.

10. A vehicle fuel consumption analysis and early warning method based on vehicle-cloud data fusion, characterized in that, The vehicle fuel consumption analysis and early warning system based on vehicle-cloud data fusion as described in any one of claims 1-9 includes: Collect on-board diagnostic data and inertial measurement data, and after preprocessing, extract flat road uniform speed segment data and standardize it according to 100 km / h to form a data package for uploading to the cloud; A hierarchical benchmark is established and updated according to vehicle type and speed range. The interval assignment accuracy and speed fluctuation coefficient within the interval are calculated based on the flat road uniform speed segment data. Then, the comprehensive matching degree is calculated, and the validity of the flat road uniform speed segment data is determined based on the comprehensive matching degree. Based on the comprehensive matching degree, the completeness of uploaded data, and the stability within a continuous statistical period, a segment confidence score is generated. Only valid flat road uniform speed segment data are assigned according to speed interval weights, and the segment confidence score is used as the weight for entering the database to update the stratification benchmark. The linear regression slope of fuel consumption change, the deviation of the current fuel consumption value per 100 kilometers from the stratified benchmark, and the comprehensive anomaly score are calculated based on the fuel consumption value per 100 kilometers in each speed range within a continuous statistical period. When the comprehensive anomaly score is not lower than the preset anomaly threshold, the anomaly type is determined. Then, the continuous anomaly or sudden anomaly is determined based on the number of consecutive verifications or the sudden change in adjacent consecutive statistical periods. Candidate fault modules are determined based on the type of anomaly, whether it is a persistent or sudden anomaly, auxiliary operating condition characteristics, and the fault module association rules corresponding to the vehicle repair manual. Based on the comprehensive anomaly score and candidate fault module output, the comprehensive score, the ranking of the same model, the percentage of defeat, fuel consumption anomaly warning information, and vehicle health warning information are provided.

Images