Vehicle fuel-efficient driving recommendation method, device, and computer-readable storage medium

Abstract

The present invention discloses a method, device, and computer-readable storage medium for recommending fuel-efficient driving for a vehicle, comprising: dividing historical message data into multiple start-stop cycles, constructing a feature vector corresponding to each start-stop cycle, the feature vector including quality information, road condition information, and fuel consumption influencing parameters; grouping the start-stop cycles according to the quality information and road condition information; obtaining a target feature vector for each group as reference data; processing the driver message data uploaded by the current vehicle to obtain a target feature vector for each group as driver driving data; obtaining the group in which the current vehicle is currently located, comparing the driver driving data corresponding to the group with the reference data, and providing a fuel-saving guide based on the comparison results. The present invention comprehensively considers quality and road condition information, and then provides a fuel-saving guide based on fuel consumption influencing parameters under specific quality and road conditions, so its fuel-saving recommendation effect is better.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle fuel consumption management, and in particular to a method and device for recommending fuel-efficient driving of a vehicle, and a computer-readable storage medium. Background Art

[0002] Economical driving is a new energy-saving technology applicable to the road traffic system. It is of great significance to improve vehicle fuel economy and reduce pollutant emissions. It is understood that the fuel consumption cost of medium and heavy trucks during the entire life cycle accounts for about 47% of the total cost. For a tractor with an annual operating mileage of 150,000 kilometers, if the oil consumption can be reduced by 1L per 100 kilometers, 9,000 yuan can be saved each year. According to statistics from the Ministry of Public Security, as of June 2021, the national motor vehicle ownership reached 384 million, of which the number of trucks reached 31.91 million. If each truck is calculated based on a saving of 9,000 yuan, at least 200 billion yuan can be saved each year, which is approximately the annual GDP of a small and medium-sized city in China. Therefore, achieving energy conservation and emission reduction for medium and heavy trucks has always been an important research direction in the field of vehicle science research.

[0003] Currently, fuel-saving driving behavior analysis is usually based on a single road condition or a fixed load, and then a fuel-saving plan is given to the driver based on the analysis results. Since a single road condition or a fixed load cannot accurately describe the fuel consumption during driving, the analysis results are less reliable and the fuel-saving effect is poor. Summary of the Invention

[0004] In view of this, it is necessary to provide a vehicle fuel-saving driving recommendation method, device and computer-readable storage medium, which can give fuel-saving suggestions by comprehensively analyzing information such as vehicle quality, road conditions, fuel consumption influencing parameters during driving, thereby improving the reliability of fuel-saving recommendations and improving fuel-saving effects.

[0005] To achieve the above objectives, an embodiment of the present invention provides a method for recommending fuel-efficient driving for a vehicle, comprising the following steps:

[0006] Acquire historical message data, and divide the historical message data into multiple start-stop cycles;

[0007] Constructing a feature vector corresponding to each start-stop cycle based on historical message data within the start-stop cycle, wherein the feature vector includes quality information, road condition information, and fuel consumption influencing parameters;

[0008] Grouping the start-stop cycles according to the quality information and the road condition information to obtain M*N groups, where M represents the number of groups of the quality information and N represents the number of groups of the road condition information;

[0009] For each group, obtaining the average instantaneous fuel consumption of each start-stop cycle in the group, and generating a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption, wherein the target feature vector includes a reference value of the fuel consumption influencing parameter;

[0010] Divide the driver message data uploaded by the current vehicle into multiple start-stop cycles, construct a feature vector for each start-stop cycle, group the start-stop cycles according to the quality information and road condition information, and generate a corresponding target feature vector for each group as driver driving data;

[0011] The group in which the current vehicle is currently located is obtained, the driving data of the driver corresponding to the group is compared with the reference data, and a fuel-saving guide is provided according to the comparison result.

[0012] Optionally, the step of generating a target feature vector corresponding to the group as reference data according to the average instantaneous fuel consumption includes:

[0013] For each group, a regression model of average instantaneous fuel consumption and fuel consumption influencing parameters is constructed based on the average instantaneous fuel consumption and characteristic vector of the start-stop cycle in the group;

[0014] Predicting the average instantaneous fuel consumption of each start-stop cycle in the group by using the regression model as the average instantaneous fuel consumption prediction value;

[0015] A target feature vector corresponding to the group is generated as reference data according to the average instantaneous fuel consumption prediction value.

[0016] Optionally, the reference data includes high fuel consumption reference data and low fuel consumption reference data, and the step of generating a target feature vector corresponding to the group as reference data according to the average instantaneous fuel consumption prediction value includes:

[0017] obtaining a first preset number of feature vectors having the highest average instantaneous fuel consumption prediction values ​​from the group and generating high fuel consumption target feature vectors based on the feature vectors, wherein the high fuel consumption target feature vectors serve as the high fuel consumption reference data;

[0018] obtaining a second preset number of feature vectors having the lowest average instantaneous fuel consumption prediction values ​​from the group and generating low fuel consumption target feature vectors based on the feature vectors, wherein the low fuel consumption target feature vectors serve as the low fuel consumption reference data;

[0019] Correspondingly, the step of comparing the driving data of the drivers corresponding to the group with the reference data and providing a fuel-saving guide based on the comparison result includes:

[0020] comparing the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data;

[0021] When the similarity is greater than a preset threshold, a fuel-saving guide is recommended to the current vehicle according to the low fuel consumption reference data.

[0022] Optionally, the step of obtaining a first preset number of feature vectors having the highest average instantaneous fuel consumption prediction values ​​from the group and generating a high fuel consumption target feature vector accordingly includes:

[0023] Obtaining a first preset number of feature vectors having the highest average instantaneous fuel consumption prediction values ​​from the group;

[0024] Calculating the mean of the fuel consumption influencing parameters of the first preset number of feature vectors, and using the feature vector formed by the mean as the high fuel consumption target feature vector;

[0025] The step of obtaining a second preset number of feature vectors with the lowest average instantaneous fuel consumption prediction values ​​from the group and generating a low fuel consumption target feature vector based on the feature vectors includes:

[0026] obtaining a second preset number of feature vectors having the lowest average instantaneous fuel consumption prediction values ​​from the group;

[0027] The mean value of the fuel consumption influencing parameters of the second preset number of feature vectors is calculated, and the feature vector formed by the mean value is used as the low fuel consumption target feature vector.

[0028] Optionally, the step of comparing the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data includes:

[0029] Calculate the similarity between the driver's driving data and the high fuel consumption reference data according to the cosine similarity formula: cos (θ) = (A*B) / (|A|*|B|);

[0030] Among them, cos (θ) represents the similarity, A is the target feature vector in the driver driving data under the group, and B is the high fuel consumption target feature vector corresponding to the high fuel consumption reference data under the group. The larger the value of cos (θ), the higher the similarity between the driver driving data and the high fuel consumption reference data.

[0031] Optionally, the fuel consumption influencing parameters include at least two of engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and brake;

[0032] For each start-stop cycle, the road condition information in the feature vector is the mode of the road conditions contained in the start-stop cycle, the engine speed, engine torque, driving speed, accelerator pedal depth or gear position in the feature vector is the message proportion of the corresponding fuel consumption influencing parameters in different intervals contained in the start-stop cycle, and the braking in the feature vector is the message proportion of the braking operation in the start-stop cycle.

[0033] Optionally, before the step of constructing a feature vector corresponding to each start-stop cycle based on historical message data in each start-stop cycle, the step further includes:

[0034] Calculating vehicle quality based on historical message data within a start-stop cycle, and marking the start-stop cycle with quality based on the vehicle quality calculation result, wherein the marking result is quality information of the start-stop cycle;

[0035] The road condition is calculated based on the latitude and longitude information in the historical message data within the start-stop cycle, and the road condition of the start-stop cycle is marked based on the road condition calculation result, and the marking result is the road condition information of the start-stop cycle.

[0036] Optionally, the step of calculating the vehicle mass based on historical message data within the start-stop cycle includes:

[0037] Calculating a first mass of the vehicle according to a vehicle longitudinal dynamics mass estimation method based on historical message data of each start-stop cycle;

[0038] Calculating a second mass of the vehicle according to a mass estimation method based on the law of conservation of energy based on historical message data of each start-stop cycle;

[0039] A third mass obtained by taking the average of the first mass and the second mass is used as the mass of the start-stop cycle.

[0040] Another embodiment of the present invention provides a vehicle fuel saving recommendation device, comprising:

[0041] A data segmentation module is used to obtain historical message data and segment the historical message data into multiple start-stop cycles;

[0042] A vectorization module, configured to construct a feature vector corresponding to each start-stop cycle based on historical message data within the start-stop cycle, wherein the feature vector includes quality information, road condition information, and fuel consumption influencing parameters;

[0043] a grouping module, configured to group the start-stop cycles according to the quality information and the road condition information to obtain M*N groups, where M represents the number of groups of the quality information and N represents the number of groups of the road condition information;

[0044] a reference data generating module, configured to obtain, for each group, the average instantaneous fuel consumption of each start-stop cycle in the group, and generate a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption, wherein the target feature vector includes a reference value of the fuel consumption influencing parameter;

[0045] a driver data processing module, configured to divide the driver message data uploaded by the current vehicle into a plurality of start-stop cycles, construct a feature vector for each start-stop cycle, group the start-stop cycles according to the quality information and road condition information, and generate a corresponding target feature vector for each group as driver driving data;

[0046] The recommendation module is used to obtain the group in which the current vehicle is currently located, compare the driver driving data corresponding to the group with the reference data, and provide a fuel-saving guide based on the comparison result.

[0047] Yet another embodiment of the present invention provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, and the at least one instruction is loaded and executed by a processor to implement the steps of the above-mentioned vehicle fuel-efficient driving recommendation method.

[0048] Compared with the prior art, the vehicle fuel-saving driving recommendation method proposed in the embodiment of the present invention first divides the historical message data into multiple start-stop cycles, then constructs a feature vector for each start-stop cycle, and then uses quality information and road condition information to group the start-stop cycles, and obtains the target feature vector generated by the low-fuel-consumption start-stop cycle in each group as reference data. The reference data includes reference values ​​of fuel consumption influencing parameters, and then obtains the driver's driving data and the group in which the current vehicle is currently located. Finally, the driver's driving data in the current group is compared with the reference data, and a fuel-saving guide is given based on the comparison results. Since the reference data in the present invention is obtained by comprehensively considering the quality and road condition information, and then a fuel-saving guide is given for the fuel consumption influencing parameters under specific quality and road conditions, its fuel-saving recommendation effect is better and can better achieve the purpose of reducing fuel consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flow chart of an embodiment of a method for recommending fuel-efficient driving for a vehicle according to the present invention.

[0050] Figure 2 This is a structural block diagram of an embodiment of a vehicle fuel-saving driving recommendation device of the present invention.

[0051] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0052] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0053] The vehicle fuel-saving driving recommendation method of an embodiment of the present invention is used to recommend fuel-saving guidelines to drivers during vehicle operation. Since the fuel consumption cost accounts for a higher proportion of the total cost of commercial vehicles such as medium and heavy trucks during their entire life cycle, the vehicle fuel-saving driving recommendation method of the present invention, when applied to commercial vehicles, is of greater significance for improving fuel economy and reducing pollutant emissions.

[0054] The inventors have found through research that different driving behaviors can lead to huge differences in fuel consumption under the same road conditions and loads. Therefore, the significance of fuel-saving driving behavior research is particularly important. The core of fuel-saving driving behavior analysis technology is mainly based on a model based on both the driver's driving operation and the vehicle's fuel consumption. The driving operation mainly affects the changes in parameters such as vehicle speed, engine speed, torque, and accelerator pedal depth. Fuel consumption reflects the current kinetic energy of the vehicle, and the two affect each other. At the same time, during vehicle driving, the corresponding relationship between driving operation and fuel consumption under different road conditions and different loads is also very different. Therefore, fuel-saving driving behavior must be analyzed under multiple conditions to be more reliable. Based on this, the present invention provides a vehicle fuel-saving driving recommendation method, device, and computer-readable storage medium. Please refer to the description of the following embodiment for details.

[0055] Example 1

[0056] Please refer to Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for recommending fuel-efficient driving of a vehicle according to an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0057] Step S101: Acquire historical message data and divide the historical message data into multiple start-stop cycles.

[0058] In this embodiment, to recommend fuel-saving guidelines to drivers during vehicle operation, low-fuel-consumption reference data is first constructed based on historical message data. During operation, vehicles upload trip data in real time, for example, one trip data item per second. Each trip data item represents one message item. Reference data is constructed based on the large amount of collected historical message data. This historical message data typically includes multiple sets of historical message data uploaded by multiple vehicles of the same series. For example, if message data uploaded by 200 vehicles within a month is selected as historical message data, the historical message data will include 200 sets of historical message data corresponding to 200 vehicles.

[0059] After acquiring a large amount of historical message data, machine learning is required to identify the relationship between fuel consumption and various influencing factors. Prior to machine learning, the historical message data is first grouped and divided into multiple start-stop cycles. Machine learning is then performed on each start-stop cycle. Specifically, a start-stop cycle is the driving period corresponding to the time points when two adjacent vehicles reach zero speed and the engine speed also reaches zero. This driving period typically includes multiple pieces of historical message data, and therefore, each start-stop cycle typically also includes multiple pieces of historical message data.

[0060] Step S102 : constructing a feature vector corresponding to each start-stop cycle according to historical message data in each start-stop cycle, wherein the feature vector includes quality information, road condition information, and fuel consumption influencing parameters.

[0061] Fuel consumption is influenced by multiple factors during driving, including vehicle quality, road conditions, engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and braking. Fuel-saving recommendations require comprehensive consideration of these factors to achieve optimal fuel savings. Engine speed and torque are generally direct factors, while driving speed, accelerator pedal depth, gear position, and braking are indirect factors. Quality and road conditions are performance factors. Of these factors, quality and road conditions are primarily determined by the objective context and have little to do with driver behavior, making it difficult to achieve fuel savings by simply modifying them. Direct and indirect factors, however, are primarily driver-controllable and can be easily and flexibly adjusted. Therefore, in the fuel-saving recommendation method, quality and road conditions can be used to group start-stop cycles. Within each group, fuel savings can be achieved by adjusting factors such as engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and braking. Specifically, to group the start-stop cycles and identify the relationship between fuel consumption and different influencing factors, the start-stop cycles need to be vectorized, that is, a corresponding feature vector is generated for each start-stop cycle. As shown in Table 1, the feature vector corresponding to each start-stop cycle includes direct factors, indirect factors, and performance factors. The quality information in the feature vector is the vehicle mass calculated according to the vehicle mass calculation method of the present invention; the road condition information is the mode of the road condition information contained in each historical message data within the start-stop cycle; the engine speed, engine torque, driving speed, accelerator pedal depth, or gear position in the feature vector are the message proportions of the corresponding fuel consumption influencing parameters in different intervals within the start-stop cycle; and the braking in the feature vector is the message proportion of braking operations within the start-stop cycle.

[0062] Table 1 Parameters included in the feature vector

[0063]

[0064] Step S103 , grouping the start-stop cycles according to the quality information and the road condition information to obtain M*N groups, where M represents the number of groups of the quality information, and N represents the number of groups of the road condition information.

[0065] After constructing a feature vector for each start-stop cycle, the quality information and road condition information in the feature vector can be used to group all start-stop cycles. In this embodiment, the quality information includes four conditions: no load, half load, full load, and overload, and the road condition information includes four conditions: highway, national highway, provincial highway, and other. Therefore, M=N=4, and the start-stop cycles can be divided into 16 groups based on the quality information and road condition information. Of course, in other embodiments, the specific grouping of quality information and road condition information can also be other situations.

[0066] Step S104: For each group, obtain the average instantaneous fuel consumption of each start-stop cycle in the group, and generate a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption. The target feature vector includes a reference value of the fuel consumption influencing parameter.

[0067] In this embodiment, the fuel consumption-influencing parameters in each feature vector include at least two of the following: engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and brake position. When adjustment of a parameter is required, the feature vector includes the corresponding parameter. This embodiment uses the example of six fuel consumption-influencing parameters, namely, engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and brake position, as an example.

[0068] Instantaneous fuel consumption indicates the fuel consumption of a vehicle at a certain moment. When the vehicle is moving, the display unit is "liters per 100 kilometers (L / 100Km)". When the engine is running but the vehicle is stopped, the display unit is "liters per hour (L / H)". Since instantaneous fuel consumption is not very meaningful for daily driving guidance, it more reflects the vehicle's instantaneous working conditions. The working conditions of a car are ever-changing during driving, and it is impossible to pay attention to the fuel consumption at every moment. Therefore, what is more instructive for us is the average instantaneous fuel consumption. Average instantaneous fuel consumption refers to the amount of fuel consumed by a car when it travels a certain mileage. It is usually expressed in "liters per 100Km". The amount of fuel consumed from the time the car starts until it stops is divided by the mileage. The value obtained is the average instantaneous fuel consumption. In this embodiment, the analysis is performed in units of start-stop cycles, so it is necessary to calculate the average instantaneous fuel consumption of each start-stop cycle. The lower the average instantaneous fuel consumption rate, the better the fuel saving effect when the fuel consumption influencing parameter value of the start-stop cycle is adopted.

[0069] In one embodiment, the reference data includes low fuel consumption reference data. Using this low fuel consumption reference data, it is possible to determine the specific value or range of the fuel consumption influencing parameter for the corresponding group, which results in the lowest average instantaneous fuel consumption. In other embodiments, in addition to the low fuel consumption parameter data, the reference data may also include high fuel consumption reference data. Using this high fuel consumption reference data, it is possible to determine the specific value or range of the fuel consumption influencing parameter for the corresponding group, which results in the highest average instantaneous fuel consumption.

[0070] In this embodiment, the fuel consumption influencing parameters in each start-stop cycle are recorded in the form of feature vectors, so the reference data is also recorded in the form of feature vectors. Specifically, the step of generating a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption includes:

[0071] (1) For each group, a regression model of average instantaneous fuel consumption and fuel consumption influencing parameters is constructed based on the average instantaneous fuel consumption and characteristic vector of the start-stop cycle in the group;

[0072] Specifically, the regression model between fuel consumption influencing parameters X and average instantaneous fuel consumption Y is expressed as Y=F(X). This represents the impact of each fuel consumption influencing parameter X on average instantaneous fuel consumption Y. In this embodiment, the regression model's fitting accuracy reaches 91.58%. In specific implementations, machine learning methods (such as Catboost) are used to find the corresponding relationship F between X and Y through model learning. This method of obtaining a regression model through machine learning is conventional and will not be described in detail here.

[0073] (2) Predicting the average instantaneous fuel consumption of each start-stop cycle under the group by using a regression model as the average instantaneous fuel consumption prediction value, and generating a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption prediction value.

[0074] Specifically, for each start-stop cycle, the relevant information of the fuel consumption influencing parameters in the feature vector of the start-stop cycle is input into the regression model, and the predicted value predicted by the regression model can be obtained. This predicted value is recorded as the average instantaneous fuel consumption predicted value. By replacing the original average instantaneous fuel consumption of the start-stop cycle with the average instantaneous fuel consumption predicted value, abnormal fuel consumption data caused by some special cases in the original data can be eliminated. Afterwards, the required parameter data can be generated based on the average instantaneous fuel consumption predicted value and the feature vector of each start-stop cycle. Specifically, when the reference data includes high fuel consumption reference data and low fuel consumption reference data, the step of generating a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption predicted value includes:

[0075] (a) A first predetermined number of feature vectors with the highest average instantaneous fuel consumption predicted values ​​are obtained from the group and used to generate a high fuel consumption target feature vector. The high fuel consumption target feature vector serves as the high fuel consumption reference data. Specifically, the average instantaneous fuel consumption predicted values ​​are sorted in a certain order, such as from largest to smallest or from smallest to largest, and then a first predetermined number of feature vectors with the highest average instantaneous fuel consumption predicted values ​​are obtained. The mean of the fuel consumption influencing parameters of the first predetermined number of feature vectors is then calculated, and a feature vector formed by this mean value serves as the high fuel consumption target feature vector. The first predetermined number is a value pre-set based on the desired fuel saving effect. For example, in one embodiment, the first predetermined number is expressed as a percentage, such as 30%. For example, if the eigenvector C corresponding to the start-stop cycle is expressed as C = (mass, road condition, engine speed, engine torque, driving speed, accelerator pedal depth, gear position, brake), then the high fuel consumption target eigenvector B1 = (engine speed, engine torque, driving speed, accelerator pedal depth, gear position, brake). The value of each fuel consumption influencing parameter in B1 is the average value of the corresponding fuel consumption influencing parameters of the 30% eigenvectors C with the highest average instantaneous fuel consumption prediction values.

[0076] (b) Obtaining a second preset number of feature vectors with the lowest average instantaneous fuel consumption prediction values ​​from the group and generating a low fuel consumption target feature vector based thereon, with the low fuel consumption target feature vector serving as the low fuel consumption reference data. In a specific implementation, the average instantaneous fuel consumption prediction values ​​are sorted in a certain order, such as from large to small or from small to large, and then a second preset number of feature vectors with the lowest average instantaneous fuel consumption prediction values ​​are obtained. The mean of the fuel consumption influencing parameters of the second preset number of feature vectors is then calculated, and the feature vector formed by this mean value serves as the low fuel consumption target feature vector. The second preset number is a value pre-set based on the desired fuel saving effect. For example, in one embodiment, the second preset number is expressed as a percentage, which is 30%. It should be noted that the second preset number may be the same as or different from the first preset number. Similarly, taking the characteristic vector C corresponding to the start-stop cycle expressed as C=(mass, road condition, engine speed, engine torque, driving speed, accelerator pedal depth, gear position, brake) as an example, in the low fuel consumption target characteristic vector B2=(engine speed, engine torque, driving speed, accelerator pedal depth, gear position, brake), the value of each fuel consumption influencing parameter is the average value of the corresponding fuel consumption influencing parameters of the 30% characteristic vectors C with the lowest average instantaneous fuel consumption prediction values.

[0077] Step S105, dividing the driver message data uploaded by the current vehicle into multiple start-stop cycles, constructing a feature vector for each start-stop cycle, grouping the start-stop cycles according to the quality information and road condition information, and generating a corresponding target feature vector for each group as driver driving data.

[0078] In this step, when a fuel-saving guide is required for a vehicle, the driving habits of the vehicle under different road conditions and vehicle weights must first be obtained. Then, based on the vehicle's current road weights and vehicle weights, the comparison is performed with the low fuel consumption reference data of the corresponding group to provide a fuel-saving guide. The method for obtaining the driving habits of a vehicle includes:

[0079] (1) Divide the driver message data uploaded by the current vehicle into multiple start-stop cycles;

[0080] (2) Constructing a feature vector for each start-stop cycle and grouping the start-stop cycles according to the quality information and road condition information; the feature vector in this step has the same structure as the feature vector in step S102, both including quality information, road condition information, and fuel consumption influencing parameters.

[0081] (3) Generate a corresponding target feature vector for each group as the driver's driving data, which is used to characterize the driving habits of the vehicle.

[0082] Compared to the eigenvector, the target eigenvector no longer includes quality information and road condition information, but only includes fuel consumption influencing parameters that can be adjusted by the driver, such as engine speed, engine torque, driving speed, accelerator pedal depth, gear position, brake, and other information. Specifically, the step of generating a corresponding target eigenvector for each group is achieved by calculating the mean of each fuel consumption influencing parameter of all eigenvectors in the group, and generating the target eigenvector based on the mean. For example, the target eigenvector A of a certain group is expressed as A = (engine speed, engine torque, driving speed, accelerator pedal depth, gear position, brake), then each fuel consumption influencing parameter, such as engine speed, represents the mean of the engine speed of the eigenvectors of all start-stop cycles in the group.

[0083] Step S106 , obtaining the group that the current vehicle is currently in, comparing the driver's driving data corresponding to the group with the reference data, and providing a fuel-saving guide based on the comparison result.

[0084] Specifically, when making fuel-saving recommendations for a vehicle, the vehicle's current quality and road conditions are used to determine its group. The driver's driving data for that group is then compared with reference data to provide fuel-saving guidance. The current vehicle quality and road conditions are obtained based on the data from the most recent start-stop cycle of at least 300 seconds. The current vehicle quality and road conditions are calculated using the message data from that start-stop cycle.

[0085] Among them, when the reference data only includes low-fuel consumption reference data, the corresponding driver driving data under the group is compared with the reference data, and the implementation principle of the fuel-saving guide is given according to the comparison results: the driver driving data is compared with the low-fuel consumption reference data to judge the difference. When the difference between the two is greater, it indicates that the current vehicle driving behavior is more different from the low-fuel consumption driving behavior. At this time, a fuel-saving guide needs to be recommended. Therefore, it can be decided whether to make a fuel-saving recommendation based on the degree of difference in the comparison results. The specific fuel-saving guide can be to recommend the reference values ​​of each fuel consumption influencing parameter in the low-fuel consumption parameter data to the driver.

[0086] When the reference data includes both high fuel consumption reference data and low fuel consumption reference data, the steps of comparing the driving data of the corresponding driver in the group with the reference data and providing a fuel saving guide based on the comparison results include:

[0087] (1) Comparing the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data; the specific implementation principle includes: calculating the similarity between the driver driving data and the high fuel consumption reference data according to the cosine similarity formula: cos(θ)=(A*B) / (|A|*|B|), wherein cos(θ) represents the similarity, A is the target feature vector in the driver driving data in the group, and B is the high fuel consumption target feature vector corresponding to the high fuel consumption reference data in the corresponding group (that is, B1 mentioned above). The larger the value of cos(θ), the higher the similarity between the driver driving data and the high fuel consumption reference data. The higher the similarity, the more necessary it is to make a fuel-saving recommendation.

[0088] (2) When the similarity is greater than a preset threshold, a fuel-saving guide is recommended for the current vehicle based on the low fuel consumption reference data. That is, when a fuel-saving recommendation is needed, a fuel-saving guide is generated using the low fuel consumption reference data and recommended to the user. The fuel-saving guide is used to guide the driver to adjust the current fuel consumption influencing parameters so that they are infinitely close to the values ​​or ranges of the fuel consumption influencing parameters in the low fuel consumption parameter data, thereby adjusting the fuel consumption of the current vehicle.

[0089] Compared with the prior art, the vehicle fuel-saving driving recommendation method of the present invention first divides the historical message data into multiple start-stop cycles, then constructs a feature vector for each start-stop cycle, and then uses quality information and road condition information to group the start-stop cycles, and obtains reference data under each group, which includes reference values ​​of fuel consumption influencing parameters. Then, the driver's driving data and the group in which the current vehicle is located at the current moment are obtained, and finally the driver's driving data under the current group is compared with the reference data, and a fuel-saving guide is given according to the comparison results. Since the division of the reference data in the present invention comprehensively considers the quality and road condition information, and then gives a fuel-saving guide for the fuel consumption influencing parameters under specific quality and road conditions, its fuel-saving recommendation effect is better and can better achieve the purpose of reducing fuel consumption.

[0090] It should be noted that the message data uploaded by the vehicle typically includes travel information such as vehicle speed (driving speed), engine speed, torque, instantaneous fuel consumption, brake pedal depth, geographic coordinates, and time, but does not directly include quality information or road condition information. Therefore, before constructing a feature vector corresponding to each start-stop cycle based on the historical message data within each start-stop cycle, the following steps are also included:

[0091] (1) Calculating the vehicle mass based on historical message data within a start-stop cycle, and marking the start-stop cycle with a mass according to the vehicle mass calculation result, wherein the marking result is the mass information of the start-stop cycle; specifically, the vehicle mass calculation process includes the following steps: calculating the first mass of the vehicle based on the historical message data of each start-stop cycle according to a mass estimation method of vehicle longitudinal dynamics; calculating the second mass of the vehicle based on the historical message data of each start-stop cycle according to a mass estimation method of the law of conservation of energy; and averaging the first mass and the second mass to obtain a third mass as the mass of the start-stop cycle.

[0092] (2) Calculate the road condition based on the longitude and latitude information in the historical message data within the start-stop cycle, and mark the road condition of the start-stop cycle based on the road condition calculation result. The marking result is the road condition information of the start-stop cycle. In this embodiment, the geographical coordinates of the historical message data include longitude and latitude information. The road condition calculation is performed based on the longitude and latitude information. The road condition calculation results include four situations: expressway, national highway, provincial highway, and other. The road condition information is then converted into numerical data. For example, the corresponding numerical values ​​of expressway, national highway, provincial highway, and other are 1, 2, 3, and 4 respectively. Then, a road condition mark is added to the historical message data based on the conversion result. The road condition mark of each historical message data is 1, 2, 3, or 4. In order to realize the grouping of the start-stop cycle according to the road condition information, the mode of the road condition marks in the historical message data contained in the start-stop cycle is used as the road condition information of the start-stop cycle.

[0093] In the embodiment of the present invention, before calculating the vehicle mass, the data needs to be pre-processed. Specifically, the data includes two parts:

[0094] 1) Utilize T-BOX to collect message data transmitted by the vehicle terminal, including vehicle speed, engine speed, torque, instantaneous fuel consumption, brake pedal depth, geographic coordinates, time and other information.

[0095] 2) Vehicle static data, including basic information such as engine gear ratio, transmission efficiency, tire radius, moment of inertia, maximum torque, and maximum torque speed.

[0096] After obtaining the above data, the message data needs to be preprocessed because it is easily interfered with during the collection process and may contain missing, duplicate or abnormal data. The preprocessing process mainly includes three steps: 1) dimensional transformation, converting units such as driving, mileage, and time; 2) outlier filtering, which mainly filters two types of data: data that clearly exceeds physical limits (for example, vehicle speed reaches 600km / h) and transmission abnormal error data; 3) elimination of missing and duplicate values.

[0097] The following describes the causes and types of missing data, duplicate data, and abnormal data.

[0098] 1) Missing data: GPS signals are lost due to high-rise buildings or tunnels, so such data are eliminated.

[0099] 2) Duplicate Data: Data duplication significantly impacts the calculation of acceleration, distance, and fuel consumption. Data overlap can occur in the raw data, manifesting itself in the time domain, where two or more sets of identical data appear at the same sampling time. Duplicate values ​​are removed using time as the index.

[0100] 3) Abnormal data: Abnormal values ​​may be caused by strong noise generated by the system itself or by external interference during the data collection process. They do not provide meaningful information and only reduce the quality of the data and subsequent analysis. The following abnormal data are eliminated. Abnormal data include:

[0101] a) Invalid data, error data generated during data collection, inf or null value;

[0102] b) Abnormal time: due to signal loss, the time to provide data is less than 1s;

[0103] c) Acceleration abnormality: if the acceleration is less than -8 m / s^2 or greater than 3.968 m / s^2, it is considered abnormal;

[0104] d) Abnormal parking: during engine shutdown, the engine speed is 0, but the vehicle speed is not 0;

[0105] e) Abnormal vehicle speed, exceeding the maximum speed the vehicle can travel;

[0106] f) Abnormal torque, the vehicle torque output percentage is negative or greater than 1;

[0107] After preprocessing the data, the preprocessed data is divided into driving conditions. The division of driving conditions is, to a certain extent, a simulation of driver task identification. In this embodiment, the driving conditions are divided into start-stop cycles and motion conditions. The definitions of the start-stop cycle and motion conditions are as follows:

[0108] 1) Start-Stop Cycle: During vehicle operation, there is a lifecycle from when the vehicle starts to when it stops. Therefore, the driving process from one stop to the next is defined as a start-stop cycle. Simply put, a start-stop cycle is the driving period corresponding to two consecutive points in time when the vehicle's speed is zero.

[0109] 2) Motion Conditions: To simulate the frequent starting, acceleration, deceleration, and cruising conditions of a car during actual driving, and to facilitate vehicle status analysis, a start-stop cycle is further divided into acceleration, deceleration, cruising, and other conditions. The raw data is preprocessed to improve its quality, but due to the low sampling frequency and quality of the data itself, errors and noise still exist in the processed speed data. Therefore, two definitions are introduced: speed span V = Vm - Vn and average acceleration a = V / (tm - tn). The specific division details are as follows:

[0110] Acceleration condition: Satisfied ;

[0111] Deceleration condition: Satisfied ;

[0112] Cruise condition: Mean(v)>=16.66m / s, where Mean represents the average value.

[0113] The rest of the cases are classified as other working conditions;

[0114] Among them, min(V) represents the minimum value of the speed span, and Quantile a represents the acceleration quantile. That is, the accelerations are sorted from small to large, and then the corresponding cumulative percentile is calculated. The quantile of the acceleration in the acceleration or deceleration condition can be modified accordingly based on the research content; similarly, Quantile V represents the speed span quantile.

[0115] Based on the above definitions of the start-stop cycle and the driving condition, the driving condition division in the embodiment of the present invention includes the following steps when implemented:

[0116] (1) Dividing the driving condition into multiple start-stop cycles based on the speed information of the historical message data, wherein the start-stop cycle is the driving time period corresponding to two adjacent time points when the vehicle speed is zero and the engine speed is also zero;

[0117] (2) The start-stop cycle is divided into acceleration condition, deceleration condition, cruising condition and other conditions according to the speed information of the historical message data in the start-stop cycle.

[0118] After the driving conditions are divided, vehicle mass calculation is performed using a start-up cycle as the minimum unit, and the message data used is the message data for the acceleration condition within the start-up cycle. Specifically, the steps for calculating vehicle mass based on historical message data within the start-stop cycle are as follows: The vehicle mass is calculated based on the historical message data corresponding to the acceleration condition within the start-stop cycle. Because the road slope under acceleration conditions is generally zero or very small and can be ignored, and the difference in initial and final speeds under acceleration conditions is large, energy conversion is better reflected and the impact of other driving resistance on mass identification is reduced. Therefore, in the embodiments of the present invention, message data for acceleration conditions is preferentially used for vehicle mass calculation. Furthermore, only acceleration condition message data is selected as input for vehicle mass calculation, allowing the mass estimation algorithm to estimate vehicle load using a small amount of message data.

[0119] Specifically, when calculating vehicle mass, the embodiments of the present invention combine the mass estimation method of vehicle longitudinal dynamics and the mass estimation method based on the law of conservation of energy. By optimizing the two mass estimation methods, the accuracy of vehicle mass estimation is further improved, making the algorithm not limited to the guidance vehicle model. Model parameters can be modified according to the specific vehicle model, transfer learning, and rapid application. Specifically, the vehicle mass is calculated based on historical message data within the start-stop cycle, and the start-stop cycle is marked as quality based on the vehicle mass calculation result. The steps of the marking result as the quality information of the start-stop cycle include:

[0120] (1) Based on historical message data corresponding to acceleration conditions during a start-stop cycle, a first mass of the vehicle is calculated according to a mass estimation method based on vehicle longitudinal dynamics. The first mass is the vehicle mass calculated according to the mass estimation method based on vehicle longitudinal dynamics.

[0121] The longitudinal nonlinear dynamic equation of the vehicle can be described by the following equation:

[0122] Formula (1)

[0123] Transform the formula and express it as:

[0124] Formula (2)

[0125] Where M is the vehicle mass, g is the acceleration due to gravity, and f is the drag coefficient. is the air resistance coefficient, A is the frontal area of ​​the car, ρ is the air density, v is the longitudinal speed of the vehicle, represents acceleration, is the slope angle, is the engine torque, is the final reduction gear ratio, is the transmission ratio, is the efficiency of the power system, is the tire radius, and n represents the engine speed. The left side of Equation (2) can be considered the system output Y, and the right side of Equation (2) can be considered the observation vector X. Finally, the least squares method is used to estimate the mass M. The various parameters are substituted into the calculation based on the specific vehicle information, and the resulting M is recorded as the first mass.

[0126] (2) Based on historical message data corresponding to the acceleration condition during the start-stop cycle, a second mass of the vehicle is calculated according to a mass estimation method based on the law of conservation of energy. The second mass is the vehicle mass calculated according to the mass estimation method based on the law of conservation of energy.

[0127] Based on the mass estimation of the law of conservation of energy, the output power of the engine can be obtained by multiplying the speed and torque. Converted into the corresponding formula:

[0128] Formula (3)

[0129] Where M is the vehicle mass, g is the acceleration due to gravity, and f is the drag coefficient. Indicates torque, represents the speed, s is the distance traveled, is the air resistance coefficient, A is the frontal area of ​​the car, ρ is the air density, is the slope angle, represents the terminal speed, Indicates the starting speed, Indicates the start time, Indicates the end time.

[0130] During vehicle movement, air resistance consumes less energy than road resistance and vehicle kinetic energy. Therefore, the formula can be converted to:

[0131] Formula (4)

[0132] The left side of formula (4) can be regarded as the output Y, and the right side of the formula except the vehicle mass can be regarded as X. The mass can be regarded as the parameter to be estimated, and M is solved by the least squares method. This M is recorded as the second mass.

[0133] (3) The third mass obtained by taking the average of the first mass and the second mass is used as the mass of the start-stop cycle.

[0134] It should be noted that in some embodiments, after calculating the first and second masses, outlier processing is required to address clearly inaccurate or erroneous calculation results. For example, if the fully loaded vehicle mass is X and the calculated mass is 120%X, this indicates a clear error in the calculation result, and therefore the mass needs to be processed. Specifically, outlier processing can be performed by assigning masses with calculated results greater than the 90th percentile mass value. This means that all calculated masses are sorted from small to large, and then the 10% with the largest values ​​in the mass data are assigned the value of the 90th percentile mass.

[0135] (4) Marking the start-stop cycle based on the calculated mass. That is, after calculating the vehicle mass of a start-stop cycle, the start-stop cycle is marked based on the vehicle mass. The quality marking results are empty, half-loaded, fully loaded, and overloaded, so that the mass information of the later start-stop period can be directly used for grouping. It should be noted that the vehicle calculation method in the embodiment of the present invention integrates the vehicle longitudinal dynamics and energy conservation models, taking advantage of their strengths and compensating for their weaknesses, thereby increasing the reliability and stability of mass estimation, and the model recognition accuracy reaches over 80%.

[0136] Example 2

[0137] This embodiment provides a vehicle fuel-efficient driving recommendation device 100, which includes:

[0138] The data segmentation module 10 is used to obtain historical message data and segment the historical message data into multiple start-stop cycles;

[0139] A vectorization module 11 is used to construct a feature vector corresponding to each start-stop cycle based on historical message data in the start-stop cycle, wherein the feature vector includes quality information, road condition information, and fuel consumption influencing parameters;

[0140] a grouping module 12, configured to group the start-stop cycles according to the quality information and the road condition information to obtain M*N groups, where M represents the number of groups of the quality information and N represents the number of groups of the road condition information;

[0141] a reference data generating module 13 configured to obtain, for each group, the average instantaneous fuel consumption of each start-stop cycle in the group, and generate a target feature vector corresponding to the group as reference data based on the average instantaneous fuel consumption, wherein the target feature vector includes a reference value of the fuel consumption influencing parameter;

[0142] a driver data processing module 14 for dividing the driver message data uploaded by the current vehicle into a plurality of start-stop cycles, constructing a feature vector for each start-stop cycle, grouping the start-stop cycles according to the quality information and road condition information, and generating a corresponding target feature vector for each group as driver driving data; and

[0143] The recommendation module 15 is configured to obtain the group in which the current vehicle is currently located, compare the driver's driving data corresponding to the group with the reference data, and provide a fuel-saving guide based on the comparison result.

[0144] Specifically, the reference data generation module 13 is specifically used to: for each group, construct a regression model of average instantaneous fuel consumption and fuel consumption influencing parameters based on the average instantaneous fuel consumption and characteristic vector of the start-stop cycle under the group; predict the average instantaneous fuel consumption of each start-stop cycle under the group through the regression model as the average instantaneous fuel consumption prediction value; generate a target characteristic vector corresponding to the group as reference data based on the average instantaneous fuel consumption prediction value.

[0145] In this embodiment, the reference data includes high fuel consumption reference data and low fuel consumption reference data. At this time, the reference data generation module 13 is specifically used to: obtain a first preset number of feature vectors with the highest average instantaneous fuel consumption prediction value from the group and generate a high fuel consumption target feature vector based on it, and the high fuel consumption target feature vector serves as the high fuel consumption reference data; obtain a second preset number of feature vectors with the lowest average instantaneous fuel consumption prediction value from the group and generate a low fuel consumption target feature vector based on it, and the low fuel consumption target feature vector serves as the low fuel consumption reference data.

[0146] Correspondingly, the recommendation module 15 is specifically configured to: compare the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data; when the similarity is greater than a preset threshold, recommend a fuel-saving guide for the current vehicle based on the low fuel consumption reference data. The principle of comparing the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data is as follows: Calculate the similarity between the driver driving data and the high fuel consumption reference data according to the cosine similarity formula: cos(θ)=(A*B) / (|A|*|B|); wherein cos(θ) represents the similarity, A is the target feature vector in the driver driving data in the group, and B is the high fuel consumption target feature vector corresponding to the high fuel consumption reference data in the group. A larger value of cos(θ) indicates a higher similarity between the driver driving data and the high fuel consumption reference data.

[0147] In this embodiment, the fuel consumption influencing parameters include at least two of the engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and brake. For each start-stop cycle, the road condition information in the feature vector is the mode of the road conditions contained in the start-stop cycle. The engine speed, engine torque, driving speed, accelerator pedal depth or gear position in the feature vector is the message proportion of the corresponding fuel consumption influencing parameters in different intervals contained in the start-stop cycle. The brake in the feature vector is the message proportion of the braking operation in the start-stop cycle.

[0148] In addition, the fuel-saving driving recommendation device according to the embodiment of the present invention further includes:

[0149] A quality calculation module, configured to calculate vehicle quality based on historical message data within a start-stop cycle, and to mark the quality of the start-stop cycle based on the vehicle quality calculation result, wherein the marking result is the quality information of the start-stop cycle;

[0150] The road condition calculation module is used to calculate the road condition based on the latitude and longitude information in the historical message data within the start-stop cycle, and mark the road condition of the start-stop cycle according to the road condition calculation result, and the marking result is the road condition information of the start-stop cycle.

[0151] Specifically, the mass calculation module is specifically used to: calculate the first mass of the vehicle based on the historical message data of each start-stop cycle and the mass estimation method of the vehicle longitudinal dynamics; calculate the second mass of the vehicle based on the historical message data of each start-stop cycle and the mass estimation method of the law of conservation of energy; and take the third mass obtained by taking the average of the first mass and the second mass as the mass of the start-stop cycle.

[0152] It should be noted that the specific implementation process of the data screening module, reference data establishment module, driving data establishment module, recommendation module, quality calculation module, and working condition division module in the vehicle fuel-saving driving recommendation device of the embodiment of the present invention can be referred to Figure 1 The relevant descriptions in the illustrated embodiments will not be described in detail here.

[0153] Example 3

[0154] This embodiment provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the steps of the method for recommending fuel-efficient driving for a vehicle as described in Example 1. Because the specific implementation of the computer-readable storage medium in this embodiment is substantially the same as that of the aforementioned method for recommending fuel-efficient driving for a vehicle, detailed description thereof is omitted here.

[0155] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0156] The above are only preferred embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for recommending fuel-efficient driving of a vehicle, characterized in that: The following steps are involved: Acquire historical message data, and divide the historical message data into multiple start-stop cycles; Constructing a feature vector corresponding to each start-stop cycle based on historical message data within the start-stop cycle, wherein the feature vector includes quality information, road condition information, and fuel consumption influencing parameters; Grouping the start-stop cycles according to the quality information and the road condition information to obtain M*N groups, where M represents the number of groups of the quality information and N represents the number of groups of the road condition information; For each group, obtaining the average instantaneous fuel consumption of each start-stop cycle in the group, and constructing a regression model of the average instantaneous fuel consumption and fuel consumption influencing parameters based on the average instantaneous fuel consumption and the characteristic vector of the start-stop cycle in the group; Predicting the average instantaneous fuel consumption of each start-stop cycle in the group by using the regression model as the average instantaneous fuel consumption prediction value; obtaining a first preset number of feature vectors with the highest average instantaneous fuel consumption prediction values ​​from the group and generating high fuel consumption target feature vectors based on the feature vectors, wherein the high fuel consumption target feature vectors serve as high fuel consumption reference data; obtaining a second preset number of feature vectors having the lowest average instantaneous fuel consumption prediction values ​​from the group and generating a low fuel consumption target feature vector based on the feature vectors, the low fuel consumption target feature vector serving as low fuel consumption reference data; the target feature vector including the reference value of the fuel consumption influencing parameter; Divide the driver message data uploaded by the current vehicle into multiple start-stop cycles, construct a feature vector for each start-stop cycle, group the start-stop cycles according to the quality information and road condition information, and generate a corresponding target feature vector for each group as driver driving data; Obtaining the group in which the current vehicle is currently located, comparing the driver driving data corresponding to the group with the reference data, and comparing the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data; When the similarity is greater than a preset threshold, a fuel-saving guide is recommended to the current vehicle according to the low fuel consumption reference data.

2. The vehicle fuel-efficient driving recommendation method according to claim 1, characterized in that: The step of obtaining a first preset number of feature vectors with the highest average instantaneous fuel consumption prediction values ​​from the group and generating a high fuel consumption target feature vector based on the feature vectors includes: Obtaining a first preset number of feature vectors having the highest average instantaneous fuel consumption prediction values ​​from the group; Calculating the mean of the fuel consumption influencing parameters of the first preset number of feature vectors, and using the feature vector formed by the mean as the high fuel consumption target feature vector; The step of obtaining a second preset number of feature vectors with the lowest average instantaneous fuel consumption prediction values ​​from the group and generating a low fuel consumption target feature vector based on the feature vectors includes: obtaining a second preset number of feature vectors having the lowest average instantaneous fuel consumption prediction values ​​from the group; The mean value of the fuel consumption influencing parameters of the second preset number of feature vectors is calculated, and the feature vector formed by the mean value is used as the low fuel consumption target feature vector.

3. The vehicle fuel-efficient driving recommendation method according to claim 1, characterized in that: The step of comparing the driver driving data in the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data includes: Calculate the similarity between the driver's driving data and the high fuel consumption reference data according to the cosine similarity formula: cos(θ)=(A*B) / (|A|*|B|); Among them, cos (θ) represents the similarity, A is the target feature vector in the driver driving data under the group, and B is the high fuel consumption target feature vector corresponding to the high fuel consumption reference data under the group. The larger the value of cos (θ), the higher the similarity between the driver driving data and the high fuel consumption reference data.

4. The vehicle fuel-efficient driving recommendation method according to claim 1, characterized in that: The fuel consumption influencing parameters include at least two of the following: engine speed, engine torque, driving speed, accelerator pedal depth, gear position, and brake; For each start-stop cycle, the road condition information in the feature vector is the mode of the road conditions contained in the start-stop cycle, the engine speed, engine torque, driving speed, accelerator pedal depth or gear position in the feature vector is the message proportion of the corresponding fuel consumption influencing parameters in different intervals contained in the start-stop cycle, and the braking in the feature vector is the message proportion of the braking operation in the start-stop cycle.

5. The vehicle fuel-efficient driving recommendation method according to claim 1, characterized in that: Before the step of constructing a feature vector corresponding to each start-stop cycle based on historical message data in each start-stop cycle, the step further includes: Calculating vehicle quality based on historical message data within a start-stop cycle, and marking the start-stop cycle with a quality result based on the vehicle quality calculation result, wherein the quality marking result is quality information of the start-stop cycle; The road condition is calculated based on the latitude and longitude information in the historical message data within the start-stop cycle, and the road condition is marked for the start-stop cycle based on the road condition calculation result, and the road condition marking result is the road condition information of the start-stop cycle.

6. The vehicle fuel-efficient driving recommendation method according to claim 5, characterized in that: The step of calculating the vehicle mass based on the historical message data within the start-stop cycle includes: Calculating a first mass of the vehicle according to a vehicle longitudinal dynamics mass estimation method based on historical message data of each start-stop cycle; Calculating a second mass of the vehicle according to a mass estimation method based on the law of conservation of energy based on historical message data of each start-stop cycle; A third mass obtained by taking the average of the first mass and the second mass is used as the mass of the start-stop cycle.

7. A vehicle fuel saving recommendation device, characterized in that: include: A data segmentation module is used to obtain historical message data and segment the historical message data into multiple start-stop cycles; A vectorization module, configured to construct a feature vector corresponding to each start-stop cycle based on historical message data within the start-stop cycle, wherein the feature vector includes quality information, road condition information, and fuel consumption influencing parameters; a grouping module, configured to group the start-stop cycles according to the quality information and the road condition information to obtain M*N groups, where M represents the number of groups of the quality information and N represents the number of groups of the road condition information; a reference data generation module, configured to obtain, for each group, the average instantaneous fuel consumption of each start-stop cycle in the group, and construct a regression model of the average instantaneous fuel consumption and fuel consumption influencing parameters based on the average instantaneous fuel consumption of the start-stop cycle in the group and the characteristic vector; predict the average instantaneous fuel consumption of each start-stop cycle in the group using the regression model as the average instantaneous fuel consumption prediction value; obtain a first preset number of characteristic vectors with the highest average instantaneous fuel consumption prediction values ​​from the group and generate a high fuel consumption target characteristic vector based on the obtained value, the high fuel consumption target characteristic vector serving as high fuel consumption reference data; obtain a second preset number of characteristic vectors with the lowest average instantaneous fuel consumption prediction values ​​from the group and generate a low fuel consumption target characteristic vector based on the obtained value, the low fuel consumption target characteristic vector serving as low fuel consumption reference data; the target characteristic vector includes the reference value of the fuel consumption influencing parameter; a driver data processing module, configured to divide the driver message data uploaded by the current vehicle into a plurality of start-stop cycles, construct a feature vector for each start-stop cycle, group the start-stop cycles according to the quality information and road condition information, and generate a corresponding target feature vector for each group as driver driving data; The recommendation module is used to obtain the group in which the current vehicle is currently located, compare the driver driving data corresponding to the group with the reference data, and compare the driver driving data under the group with the high fuel consumption reference data to determine the similarity between the driver driving data and the high fuel consumption reference data; when the similarity is greater than a preset threshold, recommend a fuel-saving guide to the current vehicle based on the low fuel consumption reference data.

8. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the steps of the vehicle fuel-efficient driving recommendation method according to any one of claims 1 to 6.

Images