Oil consumption analysis method based on T-Box data and elevation data

Abstract

The invention relates to the technical field of vehicle fuel consumption analysis, in particular to a fuel consumption analysis method based on T-Box data and elevation data, a fuel consumption prediction model is constructed based on an XGBoost model, fusion and feature extraction are carried out by collecting the T-Box data and the elevation data of a vehicle in real time, a data set suitable for machine learning model training is generated, and a fuel consumption prediction model is constructed. The method combines historical driving data and fuel consumption data to train and optimize the model, fully considers the influence of slope and topographic factors on vehicle fuel consumption, reveals the relationship among road slope, driving behavior and fuel consumption, analyzes the cause of high-fuel-consumption driving behavior through the model, provides a scientific basis for formulating a fuel-saving driving strategy, and optimizes the driving behavior.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle fuel consumption analysis, and in particular to a fuel consumption analysis method based on T-Box data and elevation data. Background Art

[0002] With the development of intelligent and automated automobile industry, fuel consumption analysis has become an important research direction to improve vehicle energy efficiency. Traditional fuel consumption analysis methods mainly rely on vehicle power system parameters (such as engine load, fuel flow, etc.) and driving status (such as vehicle speed, acceleration, deceleration, etc.). However, existing methods often ignore the impact of terrain on fuel consumption, especially factors such as slope and road type, which have a significant impact on fuel consumption during actual driving.

[0003] Existing technologies obtain terrain data through GPS and other positioning information, which can reflect the impact of geographical factors on fuel consumption to a certain extent, but lack real-time response to elevation changes and deep integration with vehicle data (such as engine load, acceleration, driving status, etc.). Therefore, how to accurately estimate fuel consumption in different terrains and different driving modes, especially in complex terrains (such as mountain roads, ramps, etc.), is a difficult problem that needs to be solved in current technology. Summary of the invention

[0004] The purpose of the present invention is to provide a fuel consumption analysis method based on T-Box data and elevation data, aiming to accurately predict fuel consumption and optimize driving behavior by deeply integrating vehicle real-time data and terrain information.

[0005] To achieve the above object, the present invention provides a fuel consumption analysis method based on T-Box data and elevation data, comprising the following steps:

[0006] Step 1: Install the T-Box device in the vehicle and collect T-Box data;

[0007] Step 2: Acquisition and processing of elevation data;

[0008] Step 3: Fuse T-Box data and elevation data and extract features;

[0009] Step 4: Build a fuel consumption prediction model and perform tuning and verification.

[0010] Optionally, the execution process of step 1 includes the following steps:

[0011] Step 1.1: Install the T-Box device in the vehicle, and the T-Box device collects the vehicle status data in real time;

[0012] Step 1.2: Connect the T-Box device to the vehicle's control system via the CAN bus to ensure continuous data collection;

[0013] Step 1.3: T-Box data is transmitted to the vehicle's control center or cloud platform through the on-board data transmission module for data storage and processing.

[0014] Optionally, the execution process of step 2 includes the following steps:

[0015] Step 2.1: Use the GPS module to obtain the latitude and longitude information of the vehicle;

[0016] Step 2.2: Based on the latitude and longitude information, use the national elevation data information file to query the road elevation data of the current location;

[0017] Step 2.3: Calculate the slope of the road by calculating the elevation difference before and after the current position and combining it with the horizontal distance;

[0018] Step 2.4: Determine the type of slope and update the road slope status in real time in combination with the vehicle's GPS location information.

[0019] Optionally, in step 2, determining the slope type, a method for calculating the slope and identifying the terrain is adopted, comprising the following steps:

[0020] Use interpolation methods to smooth elevation data;

[0021] Identify maxima and minima;

[0022] Merge the indexes of the identified maximum and minimum values, remove duplicates, and sort them;

[0023] Calculate the slope value by using the altitude difference and the mileage difference;

[0024] Slope terrain classification is performed based on slope values.

[0025] Optionally, in step 3, the T-Box data is fused with the elevation data to construct a multi-dimensional feature vector; the key features of the feature extraction are related to fuel consumption, namely, the impact of slope on vehicle fuel consumption, the interaction between vehicle speed and slope, the correlation between accelerator pedal position and slope, vehicle speed, and the impact of engine load and slope on fuel consumption.

[0026] Optionally, during the execution of step 3, the fused data needs to be preprocessed and cleaned, and finally the processed feature vectors are sorted in time series to generate a data set suitable for machine learning model training;

[0027] The preprocessing and cleaning process includes removing invalid or abnormal data, data interpolation, de-noising, standardization or normalization;

[0028] Each data point in the dataset includes the following information: timestamp, T-Box data, elevation data, and target variable.

[0029] Optionally, the execution process of step 4 includes the following steps:

[0030] Step 4.1: Fuel consumption prediction based on the machine learning model XGBoost model;

[0031] Step 4.2: Train the XGBoost model using historical driving data and fuel consumption data;

[0032] Step 4.3: Tune the training model and use the 5-fold cross-validation method to optimize the model's hyperparameters;

[0033] Step 4.4: After training is completed, the model is integrated into the vehicle for real-time data acquisition and fuel consumption prediction;

[0034] Step 4.5: During the driving process of the vehicle, the T-Box data and elevation data are obtained in real time, and the data are input into the trained fuel consumption prediction model to predict the fuel consumption under the current driving state;

[0035] Step 4.6: The model prediction results are fed back to the driver, providing real-time fuel consumption information to help the driver optimize driving behavior.

[0036] Optionally, during the process of tuning the training model, a k-fold cross-validation method is used to divide the training data into k subsets, use k-1 subsets for training, and use the remaining subset for validation, and repeat k times; then, the hyperparameters of the model are tuned through grid search or Bayesian optimization.

[0037] The present invention provides a fuel consumption analysis method based on T-Box data and elevation data, constructs a fuel consumption prediction model based on an XGBoost model, generates a data set suitable for machine learning model training by real-time collection of vehicle T-Box data and elevation data for fusion and feature extraction, trains and optimizes the model in combination with historical driving data and fuel consumption data, fully considers the influence of slope and terrain factors on vehicle fuel consumption, reveals the relationship between road slope, driving behavior and fuel consumption, analyzes the causes of high fuel consumption driving behavior through the model, provides a scientific basis for formulating fuel-saving driving strategies, and optimizes driving behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is a schematic diagram of the steps of a fuel consumption analysis method based on T-Box data and elevation data of the present invention.

[0040] Figure 2 This is a method step for calculating slope and identifying terrain in a fuel consumption analysis method based on T-Box data and elevation data of the present invention. DETAILED DESCRIPTION

[0041] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0042] See also Figure 1 The present invention provides a fuel consumption analysis method based on T-Box data and elevation data, comprising the following steps:

[0043] S1: Install T-Box device in the vehicle and collect T-Box data;

[0044] S2: Acquisition and processing of elevation data;

[0045] S3: Fusion and feature extraction of T-Box data and elevation data;

[0046] S4: Build a fuel consumption prediction model and perform tuning and verification.

[0047] The following is a further explanation based on the specific implementation steps:

[0048] Step S1 is the process of T-Box data acquisition, which includes the following steps:

[0049] S1.1 Install a T-Box device in the vehicle, which can collect vehicle status data in real time. The T-Box device is connected to the vehicle electronic system through the vehicle communication system to obtain the following data:

[0050] Vehicle speed, acceleration, accelerator pedal opening, engine net output torque, engine speed, friction torque, information collection time, atmospheric pressure, engine fuel flow, engine coolant temperature, fuel tank level, positioning status, longitude and latitude, cumulative mileage, engine torque mode, cumulative fuel consumption, etc., including but not limited to the above data information.

[0051] The T-box device is a vehicle-mounted terminal device that is connected to the vehicle via the CAN bus to obtain various data information of the vehicle, and the data collected is the T-Box data.

[0052] S1.2 is connected to the vehicle's control system in real time via the CAN bus, ensuring that the T-Box can continuously and accurately collect and transmit vehicle data.

[0053] S1.3 T-Box data is transmitted to the vehicle's control center or cloud platform through the vehicle's onboard data transmission module (such as Wi-Fi or LTE module) for data storage and processing.

[0054] Step S2: Acquisition and processing of elevation data;

[0055] S2.1 Use the GPS module to obtain the latitude and longitude information of the vehicle. When the vehicle is driving, the GPS module continuously updates the current location information.

[0056] S2.2 Based on the latitude and longitude information, use the national elevation data information file to query the road elevation data at the current location. The elevation data reflects the change in ground altitude.

[0057] S2.3 Calculate the slope of the road by calculating the elevation difference before and after the current position and combining it with the horizontal distance. The slope calculation formula is:

[0058]

[0059] S2.4 determines the type of slope and updates the road slope status in real time in combination with the vehicle's GPS location information.

[0060] Furthermore, the process of obtaining elevation data in S2.2 includes the following steps:

[0061] S2.2.1 Prepare data: Get the real-time longitude and latitude data from the vehicle system and provide it as input data to the get_elevation function. For example, this data may come from GPS or T-Box module.

[0062] S2.2.2 Query elevation data: For each longitude and latitude point, obtain the corresponding elevation data by calling the get_elevation function. This function will first check whether the longitude and latitude are in the cached data. If not, it will query multiple GeoTIFF files to obtain the elevation.

[0063] S2.2.3 Processing elevation data:

[0064] The GeoTIFF file is opened using the rasterio library in the get_elevation function.

[0065] Use Transformer to convert the longitude and latitude of the WGS 84 coordinate system to the coordinates of the UTM coordinate system.

[0066] Convert the UTM coordinates to the pixel coordinates of the raster data, and obtain the corresponding pixel row and column positions through the pixel coordinates.

[0067] Read the elevation value of the corresponding position from the raster data. If the reading is successful, return the elevation value.

[0068] Also, see Figure 2 , which is a method step for calculating slope and identifying terrain in a fuel consumption analysis method based on T-Box data and elevation data of the present invention. In the process of step S2.3, the following steps are specifically included:

[0069] S2.3.1 Smoothing elevation data:

[0070] Input: a smoothed elevation column (elevation_data) and a cumulative mileage column (mileage_data), where the elevation column contains the elevation data of each location and the cumulative mileage column represents the cumulative distance traveled by the vehicle.

[0071] For data points with an elevation value of 0 (which may be caused by measurement errors or missing data), use an interpolation method (such as linear interpolation or polynomial interpolation) to fill in these missing values ​​and ensure data continuity.

[0072] S2.3.2 Identify maxima and minima:

[0073] Use the find_peaks function from the scipy library to identify all local maxima (ascending phases) and minima (descending phases) in the elevation data. For the ascending phase in the elevation data, use the find_peaks function directly to identify the maxima.

[0074] For the descending phase, the altitude data is first negative, and then the find_peaks function is used to identify the minimum value (corresponding to the actual altitude decrease phase).

[0075] Since the find_peaks function does not process the first and last points by default, you need to manually determine whether the first and last points are extreme values. If so, manually add them to the extreme value index list.

[0076] S2.3.3 Merge and sort extreme value indices:

[0077] The indexes of all the maximum and minimum values ​​identified in the previous step are merged, and then sorted after removing duplicates. The sorted result ensures that the index order of the extreme values ​​is consistent with the order of actual terrain changes.

[0078] Further calculation of the slope, for each extreme interval, calculate the slope, defined as the ratio of the elevation difference to the mileage difference between two adjacent points:

[0079]

[0080] in, is the altitude difference between two adjacent points (unit: meter); is the mileage difference between two adjacent points (unit: meter).

[0081] S2.3.4 Terrain classification:

[0082] Based on the absolute value of the calculated slope, the terrain is divided into three categories:

[0083] Plain: slope ≤ 5°;

[0084] Micro-hill: 5° < slope ≤ 15°;

[0085] Heavy hills: slope > 15°.

[0086] Traverse each slope interval, update the cumulative distance of the corresponding terrain type according to the category of each slope value, and calculate its proportion in the total distance.

[0087] Step S2.4 Output terrain classification results and proportions:

[0088] S2.4.1 Input data:

[0089] Smoothed elevation data: elevation_data, containing the elevation information of each location.

[0090] Cumulative mileage data: mileage_data, records the cumulative distance traveled by the vehicle.

[0091] Interpolated elevation column: fills in the missing elevation data to ensure data continuity.

[0092] S2.4.2 Calculate the terrain type for each slope segment:

[0093] Traverse the sorted extreme value indexes, calculate the slope value based on the adjacent extreme value points, and classify the slope value into plain, micro-hill or heavy hill. After each slope is calculated, update the cumulative distance of the corresponding terrain type according to its slope value.

[0094] S2.4.3 Calculate the cumulative distance and proportion:

[0095] For each terrain type (plain, light hill, heavy hill), calculate its cumulative distance.

[0096] According to the ratio of cumulative distance to total mileage, the proportion of each terrain type is calculated as follows:

[0097]

[0098] S2.4.4 Output results:

[0099] The accumulated distance of each terrain type and its proportion in the total distance are output as input features for subsequent fuel consumption prediction.

[0100] Step S3 is the process of data fusion and feature extraction, which includes the following steps:

[0101] S3.1 Fusion of T-Box data and elevation data In this section, dynamic vehicle data from the onboard T-Box is combined with elevation data to generate a multi-dimensional feature vector. The purpose of data fusion is to fully utilize the advantages of each data type and improve the prediction ability of the fuel consumption prediction model. The fused data includes but is not limited to the following aspects:

[0102] Vehicle speed: reflects the current speed of the vehicle, in km / h. Vehicle speed is an important factor affecting fuel consumption, and higher vehicle speed usually leads to higher fuel consumption.

[0103] Acceleration: The acceleration or deceleration of the vehicle, measured in m / s². Acceleration determines the acceleration process of the vehicle and directly affects the engine load and fuel consumption.

[0104] Accelerator pedal position: The throttle opening reflects the driver's acceleration needs. The position of the accelerator pedal is directly related to the vehicle's power output.

[0105] Engine load: The vehicle's engine load reflects the intensity of the engine's work. A higher engine load usually means higher fuel consumption.

[0106] Road slope: Slope information calculated through elevation data, terrain proportion obtained through slope. Different terrains have different effects on fuel consumption.

[0107] Data fusion method:

[0108] Using weighted averaging, features from different sources are combined into a multi-dimensional feature vector.

[0109] When fusion is performed, it is necessary to ensure that the different data sources are time-aligned, that is, to match vehicle data and geographic data at the same time point.

[0110] S3.2 Extraction of key features related to fuel consumption In the feature extraction stage, it is necessary to extract features closely related to fuel consumption prediction from the fused data. In particular, the following key features:

[0111] Impact of terrain ratio on vehicle fuel consumption: Different terrains such as plains, slight hills, and heavy hills will significantly affect fuel consumption. Further explanation: When going uphill under different terrains, the greater the slope, the more power output the vehicle needs to overcome gravity. Slopes are classified (such as gentle, steep, etc.) and their relationship with fuel consumption is analyzed.

[0112] The interaction between speed and slope: The interaction between speed and slope is a complex factor affecting fuel consumption. When going uphill, high speed may lead to higher fuel consumption, while when going downhill, the car is at a constant speed or idling, and fuel consumption does not change significantly. Especially in complex terrain, the interaction between speed and slope needs special consideration.

[0113] The relationship between the accelerator pedal position, slope and vehicle speed: The accelerator pedal opening directly affects the engine load and fuel consumption, while the combination of the accelerator pedal position, slope and vehicle speed can reflect the driving needs of the vehicle. When the slope is large, the driver may need a larger accelerator opening to overcome gravity, resulting in higher fuel consumption. Use the combined features of the accelerator pedal position, vehicle speed and slope to model.

[0114] Effect of engine load and grade on fuel consumption: Engine load reflects the demand on the vehicle's power system. Higher loads generally result in higher fuel consumption, especially on steep slopes or under acceleration. Calculate the relationship between engine load and grade, especially in complex terrain.

[0115] S3.3 Data preprocessing and cleaning After data collection, the data needs to be preprocessed to ensure the accuracy and completeness of the input features. The steps include:

[0116] Removing invalid or abnormal data: Detecting and removing abnormal values ​​caused by sensor failure, data loss, etc. For example, points where the vehicle speed is 0 or the acceleration is abnormal need to be corrected or deleted.

[0117] Data interpolation: For missing points in elevation data, interpolation is used to fill them in. The interpolation method can be linear interpolation, spline interpolation, etc., to ensure the smoothness and coherence of the data.

[0118] Denoising: Remove high-frequency noise, especially small fluctuations in GPS data or elevation data, through smoothing techniques.

[0119] Standardization or normalization: Standardize different features to ensure numerical stability during model training, especially features with large values ​​such as vehicle speed and acceleration.

[0120] S3.4 Dataset Construction and Time Series Sorting For machine learning models, the time order of data is very important. Therefore, at this stage, all processed data needs to be sorted in time order. Each data point contains the following information:

[0121] Timestamp, T-Box data (vehicle speed, acceleration, accelerator pedal position, engine load, etc.), elevation data (slope, altitude), target variable (fuel consumption)

[0122] Finally, the present invention packages these data into a training set for training the machine learning model. The data set should contain different driving scenarios (such as cities, mountains, highways, etc.) to ensure that the model can handle diverse driving situations.

[0123] The execution process of step S4 includes the following steps:

[0124] S4.1 Selecting a machine learning model for fuel consumption prediction The XGBoost model can handle regression problems and predict the fuel consumption of a vehicle.

[0125] S4.2 Use historical driving data and fuel consumption data to train the model. The training data should include various terrain proportions, driving conditions, and corresponding fuel consumption, including but not limited to the above data.

[0126] S4.3 Tune the training model and use the 5-fold cross-validation method to optimize the model's hyperparameters to improve the model's accuracy and generalization ability.

[0127] S4.4 After training is completed, the model is integrated into the vehicle's onboard computing unit or cloud platform to ensure that it can obtain data in real time and make fuel consumption predictions.

[0128] S4.5 During vehicle driving, T-Box data and elevation data are acquired in real time, and the data are input into the trained fuel consumption prediction model to predict the fuel consumption under the current driving state.

[0129] The S4.6 model prediction results are fed back to the driver, providing real-time fuel consumption information and helping the driver optimize driving behavior (such as deceleration, adjusting the accelerator pedal position, etc.), thereby improving the vehicle's energy efficiency.

[0130] Furthermore, in the process of step S4.1, XGBoost selects the XGBoost algorithm that has better performance in regression problems in fuel consumption prediction, which can effectively model complex nonlinear relationships and provide efficient and accurate prediction results, especially when dealing with large-scale data and feature engineering.

[0131] The XGBoost algorithm is based on Gradient Boosting Decision Trees and improves prediction accuracy by combining multiple weak classifiers (i.e., decision trees). XGBoost can automatically handle missing values, supports a variety of regularization techniques, prevents overfitting, and improves the generalization ability of the model.

[0132] The historical driving data and fuel consumption data used in step S4.2 are used for training data: including the historical driving data and fuel consumption data of the vehicle, including data on different terrain proportions and driving conditions, but not limited to the above data. These data can be obtained from the vehicle T-Box or historical driving records.

[0133] Target variable: Fuel consumption data (fuel consumption) is used as the predicted target variable.

[0134] Training process:

[0135] All preprocessed feature vectors are used for model training to ensure that the model can make predictions under a variety of different driving conditions.

[0136] The data is divided into training set, validation set and test set to ensure that the model performs stably on different data sets.

[0137] The steps for tuning step S4.3 are as follows:

[0138] Cross-validation: Use the k-fold cross-validation method to divide the training data into k subsets, use k-1 subsets for training, and use the remaining subset for validation, repeating k times.

[0139] Hyperparameter optimization: Tuning model hyperparameters (such as learning rate, maximum tree depth, regularization coefficient, etc.) through grid search or Bayesian optimization

[0140] Model deployment and real-time prediction in step S4.4 include two aspects:

[0141] Integration into the vehicle computing unit: Embed the trained XGBoost model into the vehicle’s computing unit to obtain data in real time and make fuel consumption predictions.

[0142] Cloud platform deployment: Unified management and prediction of multiple vehicles are carried out through the cloud platform, and cloud storage and analysis of data are also supported.

[0143] In summary, the present invention has the following advantages by fusing T-Box data with elevation data:

[0144] 1. Improve fuel efficiency: In-depth analysis of fuel consumption characteristics under different road slope conditions and comprehensive consideration of terrain factors can identify high-fuel-consumption driving behaviors, making fuel consumption predictions more accurate, thereby proposing fuel-saving driving strategies, significantly reducing fuel consumption and improving fuel efficiency.

[0145] 2. Real-time optimization of driving behavior: Through real-time feedback of fuel consumption information, it helps the driver understand how to adjust driving behavior under different slope conditions to reduce fuel consumption, so as to reduce unnecessary energy loss, achieve more economical driving and optimize energy use.

[0146] 3. Improve transportation efficiency: By reducing fuel consumption, transportation costs can be reduced and transportation efficiency can be improved, which is particularly important for the logistics and transportation industries.

[0147] 4. Improved economic benefits: Reducing fuel consumption is directly related to improved economic benefits. For commercial vehicle operators, this can reduce operating costs and increase profits.

[0148] 5. Reduced environmental impact: Reducing fuel consumption means reducing greenhouse gas emissions, which has a positive effect on environmental protection and reducing air pollution.

[0149] 6. Data-driven decision support: By integrating T-Box data and elevation data and applying machine learning algorithms, data-driven decision support tools are provided to help automotive equipment users and drivers make more reasonable operational decisions based on actual data.

[0150] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A fuel consumption analysis method based on T-Box data and elevation data, characterized in that: The following steps are involved: Step 1: Install the T-Box device in the vehicle and collect T-Box data; Step 2: Acquisition and processing of elevation data; Step 3: Fuse T-Box data and elevation data and extract features; Step 4: Build a fuel consumption prediction model and perform tuning and verification.

2. The fuel consumption analysis method based on T-Box data and elevation data according to claim 1, characterized in that: The execution process of step 1 includes the following steps: Step 1.1: Install the T-Box device in the vehicle, and the T-Box device collects the vehicle status data in real time; Step 1.2: Connect the T-Box device to the vehicle's control system via the CAN bus to ensure continuous data collection; Step 1.3: T-Box data is transmitted to the vehicle's control center or cloud platform through the on-board data transmission module for data storage and processing.

3. The fuel consumption analysis method based on T-Box data and elevation data according to claim 2, characterized in that: The execution process of step 2 includes the following steps: Step 2.1: Use the GPS module to obtain the latitude and longitude information of the vehicle; Step 2.2: Based on the latitude and longitude information, use the national elevation data information file to query the road elevation data of the current location; Step 2.3: Calculate the slope of the road by calculating the elevation difference before and after the current position and combining it with the horizontal distance; Step 2.4: Determine the type of slope and update the road slope status in real time in combination with the vehicle's GPS location information.

4. The fuel consumption analysis method based on T-Box data and elevation data according to claim 3, characterized in that: Step 2 determines the slope type by using a method for calculating the slope and identifying the terrain, including the following steps: Use interpolation methods to smooth elevation data; Identify maxima and minima; Merge the indexes of the identified maximum and minimum values, remove duplicates, and sort them; Calculate the slope value by using the altitude difference and the mileage difference; Slope terrain classification is performed based on slope values.

5. The fuel consumption analysis method based on T-Box data and elevation data according to claim 4, characterized in that: In step 3, the T-Box data is fused with the elevation data to construct a multi-dimensional feature vector; the key features of feature extraction are related to fuel consumption, namely the impact of slope on vehicle fuel consumption, the interaction between vehicle speed and slope, the correlation between accelerator pedal position and slope and vehicle speed, and the impact of engine load and slope on fuel consumption.

6. The fuel consumption analysis method based on T-Box data and elevation data according to claim 5, characterized in that: During the execution of step 3, the fused data needs to be preprocessed and cleaned, and finally the processed feature vectors are sorted in time series to generate a data set suitable for machine learning model training; The preprocessing and cleaning process includes removing invalid or abnormal data, data interpolation, de-noising, standardization or normalization; Each data point in the dataset includes the following information: timestamp, T-Box data, elevation data, and target variable.

7. The fuel consumption analysis method based on T-Box data and elevation data according to claim 6, characterized in that: The execution process of step 4 includes the following steps: Step 4.1: Fuel consumption prediction based on the machine learning model XGBoost model; Step 4.2: Train the XGBoost model using historical driving data and fuel consumption data; Step 4.3: Tune the training model and use the 5-fold cross-validation method to optimize the model's hyperparameters; Step 4.4: After training is completed, the model is integrated into the vehicle for real-time data acquisition and fuel consumption prediction; Step 4.5: During the driving process of the vehicle, the T-Box data and elevation data are obtained in real time, and the data are input into the trained fuel consumption prediction model to predict the fuel consumption under the current driving state; Step 4.6: The model prediction results are fed back to the driver, providing real-time fuel consumption information to help the driver optimize driving behavior.

8. The fuel consumption analysis method based on T-Box data and elevation data according to claim 7, characterized in that: In the process of tuning the training model, the k-fold cross-validation method is used to divide the training data into k subsets, use k-1 subsets for training, and use the remaining subset for validation, and repeat k times; then use grid search or Bayesian optimization to tune the model's hyperparameters.

Images