Sanitation vehicle driving behavior recommendation method and system based on operation energy consumption
Through machine learning technology, the energy consumption data of sanitation vehicles is dynamically obtained and analyzed, and the driving behavior instructions with the lowest energy consumption are generated, which solves the problem of difficult energy consumption in traditional sanitation operations and realizes an efficient and energy-saving operation method.
Patent Information
- Application Number
- CN202510056862.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-14
- Publication Date
- 2025-05-13
AI Technical Summary
In traditional sanitation operations, it is difficult to achieve the lowest energy consumption while ensuring work quality, and there is a lack of effective means to constrain energy consumption by driver driving behavior.
Through machine learning, dynamically obtain vehicle energy consumption data, perform data cleaning and energy consumption level division, extract sample features and train AI models, generate driving behavior instructions with the lowest energy consumption and issue them to the vehicle end.
It has achieved the goal of reducing the energy consumption of sanitation vehicles, improving the cost-effectiveness of operations, and achieving the lowest energy consumption while ensuring work quality.
Smart Images
Figure CN119988726A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent sanitation technology, and in particular to a method and system for recommending driving behavior of a sanitation vehicle based on operating energy consumption. Background Art
[0002] With the development of the sanitation industry, more and more sanitation vehicles are connected to the Internet of Things. In this process, more and more data is generated. Relying on this part of data, the energy consumption of the vehicle can be calculated, and even the fuel consumption of each sanitation operation mode can be calculated at a more fine-grained level. Traditionally, in the process of sanitation operations, how to achieve the lowest energy consumption while ensuring the quality of work, and how to constrain the driver's driving behavior to achieve the lowest energy consumption are still blank in the market. Summary of the invention
[0003] In order to solve the above technical problems, the present invention constructs an energy-saving driving behavior recommendation for the sanitation field, dynamically obtains the vehicle energy consumption set with the help of machine learning, generates driving instructions based on the minimum energy consumption and sends them to the vehicle side, thereby influencing the operation side and constraining the driver's driving behavior, and finally realizing a cost-effective operation mode, achieving the lowest energy consumption and realizing the purpose of energy saving.
[0004] To achieve the above-mentioned purpose, the technical solution of the present invention provides a method for recommending driving behavior of sanitation vehicles based on operation energy consumption, which comprises: S1: obtaining a road section with high feasibility as an ideal road section through data cleaning; S2: performing energy consumption level classification; S3: performing sample feature extraction;
[0005] S4: Train the AI model based on samples and output the corresponding energy consumption; S5: Generate driving behavior instructions based on the minimum energy consumption and send them to the vehicle.
[0006] Furthermore, in step S1, the road section with high feasibility meets the following two conditions: (1) the network signal strength is the best to ensure that the gateway data is fully acquired; and (2) the operation mode is turned on for operation.
[0007] Furthermore, in step S1, the data cleaning process includes: the data platform obtains time series data from the gateway; multiple data sources are associated through time series timestamps and vehicle numbers to obtain a target time series wide table; the same type of vehicles are filtered, and the mean fuel consumption A1 and variance A2 of the section are obtained by aggregation calculation according to the road section dimension; the same type of vehicles are filtered, and the mean fuel consumption B1 and variance B2 of each vehicle in the section are obtained by aggregation calculation according to the road section and vehicle dimensions.
[0008] Furthermore, in step S2, vehicles whose mean B1 is smaller than mean A1 and whose variance B2 is smaller than variance A2 are marked as Class I energy consumption vehicles, otherwise they are marked as Class II energy consumption vehicles.
[0009] Furthermore, in step S3, the vehicle number set of Class I energy consumption vehicles is obtained, and the energy consumption distribution of these vehicles passing through the above-mentioned ideal road section every day is queried through SQL with the help of the data warehouse, where the sample extraction features include: average driving speed, washing and sweeping operation mode, operation standard, sweeping disk speed, whether dust reduction is used, and whether high-pressure water is used.
[0010] Furthermore, in step S3, the driving behavior instruction is used as an input with the help of the decision engine, and the corresponding vehicle weight set is obtained as another input; then the AI model is trained and deployed on the vehicle side / cloud platform to realize dynamic voice broadcasting of the driving behavior instructions with the lowest energy consumption when passing through the ideal road section.
[0011] Furthermore, for each driving behavior instruction, the weight at the starting point of the corresponding ideal road section is associated, and the weight is represented by the remaining oil and water volume conversion, that is, the remaining oil and water volume are also added as features in the feature extraction.
[0012] Furthermore, in step S4, the training process includes:
[0013] Each feature in the sample is labeled with an independent variable, and the correlation coefficient is added:
[0014] y=w1x1+w2x2+w3x3+w4x4+....+w n x n
[0015] m samples form m n-variable linear equations;
[0016] Then use sklearn to solve the normal equation, call the API, and pass in the X data and y target value;
[0017] Set the intercept parameter to true / false according to the scenario. In this scenario, there is no intercept and it is set to false. Finally, the solution of the normal equation is obtained by the following method: numpy.linalg.inv(XTdot(X)).dot(X,T).dot(y);
[0018] Thus we can obtain the above w1,w2,w3,w4,...,w n The value of the coefficient.
[0019] Further, in step S5, the remaining fuel amount is obtained when the starting point of the operation section is reached, and according to the remaining fuel amount, the operation parameters with the lowest energy consumption are output in the linear regression model, the corresponding energy consumption is output, and then the input features corresponding to the lowest energy consumption are obtained.
[0020] The technical solution of the present invention also provides a sanitation vehicle driving behavior recommendation system based on operational energy consumption, which includes: an ideal section acquisition module, which is used to obtain a highly feasible section as an ideal section through data cleaning; an energy consumption level classification module, which is used to perform energy consumption level classification; a sample feature extraction module, which is used to perform sample feature extraction; a model training module, which is used to train an AI model based on samples and output corresponding energy consumption; a driving behavior instruction dynamic generation module, which is used to generate driving behavior instructions based on the minimum energy consumption and send them to the vehicle end. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 is a flow chart for recommending optimal energy consumption driving parameters of the present invention;
[0022] Figure 2 This is a schematic diagram of capturing an ideal road section according to the present invention. DETAILED DESCRIPTION
[0023] The technical solution of the present invention is further described below in conjunction with specific embodiments, but the present invention is not limited to these embodiments.
[0024] The present invention aims to construct an energy-saving driving behavior recommendation for the sanitation field. With the help of machine learning, the vehicle energy consumption set is dynamically acquired, and driving instructions are generated based on the minimum energy consumption and sent to the vehicle side, so as to influence the operation side and constrain the driver's driving behavior, and finally realize a cost-effective operation mode, achieve the lowest energy consumption, and realize the purpose of energy saving.
[0025] The key technical points of the present invention are: ideal road section feature extraction, linear regression training AI model output energy consumption set, and recommendation of driving parameters with the lowest energy consumption based on operating parameters.
[0026] In the technical solution of the present invention, a method for recommending driving behavior of sanitation vehicles based on operational energy consumption includes: S1: obtaining a road section with high feasibility as an ideal road section through data cleaning; S2: classifying energy consumption levels; S3: extracting sample features; S4: training an AI model based on the samples and outputting the corresponding energy consumption; S5: generating driving behavior instructions based on the minimum energy consumption and sending them to the vehicle end.
[0027] Furthermore, in step S1, the road section with high feasibility meets the following two conditions: (1) the network signal strength is the best to ensure that the gateway data is fully acquired; and (2) the operation mode is turned on for operation.
[0028] Furthermore, in step S1, the data cleaning process includes: the data platform obtains time series data from the gateway; multiple data sources are associated through time series timestamps and vehicle numbers to obtain a target time series wide table; the same type of vehicles are filtered, and the mean fuel consumption A1 and variance A2 of the section are obtained by aggregation calculation according to the road section dimension; the same type of vehicles are filtered, and the mean fuel consumption B1 and variance B2 of each vehicle in the section are obtained by aggregation calculation according to the road section and vehicle dimensions.
[0029] Furthermore, in step S2, vehicles whose mean B1 is smaller than mean A1 and whose variance B2 is smaller than variance A2 are marked as Class I energy consumption vehicles, otherwise they are marked as Class II energy consumption vehicles.
[0030] Furthermore, in step S3, the vehicle number set of Class I energy consumption vehicles is obtained, and the energy consumption distribution of these vehicles passing through the above-mentioned ideal road section every day is queried through SQL with the help of the data warehouse, where the sample extraction features include: average driving speed, washing and sweeping operation mode, operation standard, sweeping disk speed, whether dust reduction is used, and whether high-pressure water is used.
[0031] Furthermore, in step S3, the driving behavior instruction is used as an input with the help of the decision engine, and the corresponding vehicle weight set is obtained as another input; then the AI model is trained and deployed on the vehicle side / cloud platform to realize dynamic voice broadcasting of the driving behavior instructions with the lowest energy consumption when passing through the ideal road section.
[0032] Furthermore, for each driving behavior instruction, the weight at the starting point of the corresponding ideal road section is associated, and the weight is represented by the remaining oil and water volume conversion, that is, the remaining oil and water volume are also added as features in the feature extraction.
[0033] Furthermore, in step S4, the training process includes:
[0034] Each feature in the sample is labeled with an independent variable, and the correlation coefficient is added:
[0035] y=w1x1+w2x2+w3x3+w4x4+....+w n x n
[0036] m samples form m n-variable linear equations;
[0037] Then use sklearn to solve the normal equation, call the API, and pass in the X data and y target value;
[0038] Set the intercept parameter to true / false according to the scenario. In this scenario, there is no intercept and it is set to false. Finally, the solution of the normal equation is obtained by the following method: numpy.linalg.inv(XTdot(X)).dot(X,T).dot(y);
[0039] Thus we can obtain the above w1,w2,w3,w4,...,w n The value of the coefficient.
[0040] Further, in step S5, the remaining fuel amount is obtained when the starting point of the operation section is reached, and according to the remaining fuel amount, the operation parameters with the lowest energy consumption are output in the linear regression model, the corresponding energy consumption is output, and then the input features corresponding to the lowest energy consumption are obtained.
[0041] The detailed steps of the technical solution of the present invention are:
[0042] 1. Obtaining the ideal road section
[0043] An ideal road segment needs to meet two conditions:
[0044] (1) Select a road section with good signal to ensure complete gateway data acquisition
[0045] like Figure 2 As shown in the figure, CSQ can reflect the network signal strength (signal quality, example: 31, 99; the value before the comma represents the CSQ value, between 0 and 31, the larger the value, the better the signal quality, the value after the comma represents the channel, the value is between 0 and 99, otherwise you should check whether the antenna or SIM card is installed correctly), the CSQ value is between 0 and 31, the larger the value, the better the signal quality, if 99 appears, it means the channel is invalid, the figure shows the distribution of signal strength on the road section, the color depth represents the signal strength, the lighter the color, the stronger the signal, the heavier the color, the weaker the signal. This sample selects the road section with the best signal, such as Figure 2 The selected road segment is shown.
[0046] (2) Select the road section where the operation mode is turned on.
[0047] On this road section, a single working mode operation is required and the operating mileage meets the preset mileage. For example, this road section is a road section where a sweeper performs full sweeping and washing operations.
[0048] Data cleaning process: The data platform obtains time series data from the gateway; multiple data sources are associated through time series timestamps and vehicle numbers to obtain the target time series wide table, the content of which can be timestamp, vehicle number, longitude coordinates, latitude coordinates, remaining fuel consumption, instantaneous speed, speed, working status, etc.; filter the same type of vehicles, such as 18-ton sanitation sweepers, and obtain the mean fuel consumption A1 and variance A2 of the section according to the above-mentioned road section dimension aggregation calculation; filter the above-mentioned vehicles of the same type, and obtain the mean fuel consumption B1 and variance B2 of each vehicle in the section according to the road section and vehicle dimensions.
[0049] In particular, the energy consumption here is measured by fuel consumption, because fuel consumption can be obtained from the vehicle-side fuel consumption sensor, or it can be other measurements, such as the degree of wear of the sweeping disc after the operation is completed.
[0050] 2. Energy consumption level classification
[0051] The vehicles with the above mean B1 less than the mean A1 and the variance B2 less than the variance A2 are marked as Class I energy consumption vehicles, otherwise they are marked as Class II energy consumption vehicles. Here, Class I energy consumption is assumed to be the lowest, and so on and so forth (only Class I energy consumption is extracted later, so only two levels are needed here).
[0052] 3. Feature extraction
[0053] Obtain the vehicle number set of the above-mentioned Class I energy consumption vehicles, and use SQL to query the energy consumption distribution of these vehicles when they pass through the above-mentioned ideal road section every day with the help of the data warehouse. The sample extraction features can be: average driving speed, washing and sweeping operation mode, operation standard, sweeping disk speed, whether dust reduction, whether high-pressure water, etc.; the sample can be sample 1: ideal road section, time period 2022-09-01 11:00:00 to 2022-09-01 11:09:00, driver Wang, vehicle number 80009, section length 1km, average driving speed 1.75m / s, full washing and sweeping operation, standard sweeping, sweeping disk speed F5, dust reduction, high-pressure water, etc.; sample 2: ideal road section, time period 2022-11-30 10:30:00 to 2022-11-30 10:40:40, driver Li, vehicle number 80010, section length 1km, average driving speed 1.63m / s, full washing and sweeping operation, strong sweeping, sweeping disk speed F7, dust reduction, high-pressure water, etc. The current value label is the minimum energy consumption corresponding to the vehicle number of the above sample.
[0054] The above instructions can be triggered by a strong vehicle end or a strong platform. For example, there are many ideal road sections in reality, which will increase the workload of the team leader and the driver. It is unrealistic to rely solely on the strong vehicle end. For example, the weight of the vehicle body is different when passing through different ideal road sections (the fuel tank capacity and the water tank water level will affect the weight of the vehicle body), and the above driving behavior instructions will lose their meaning.
[0055] Therefore, the above driving behavior instructions are used as an input with the help of the decision engine, and then the corresponding vehicle weight (here, the corresponding remaining fuel consumption in the fuel tank and the corresponding characterization weight of the water level in the water tank) set is obtained as another input. Then the model is trained and deployed on the vehicle side / cloud platform (if deployed on the cloud platform, instructions need to be issued), so as to realize the dynamic voice broadcast of the driving behavior instructions with the lowest energy consumption when passing through the ideal road section. The specific process can be: for each of the above driving behavior instructions, the weight at the starting moment of the corresponding ideal road section is associated, and the weight is characterized by the conversion of the remaining oil and water, that is, the remaining oil and remaining water are also added as features in the above feature extraction. The less the remaining weight, the lighter the car, and vice versa. The decision engine constructed in this way, such as responding to the vehicle weight of 17 tons, generates driving instructions such as full washing and sweeping operations, strong sweeping, sweeping disk speed F2, and turning on dust reduction and high-pressure water.
[0056] 4. Driving behavior recommendations
[0057] The AI model is trained based on the above samples and the corresponding energy consumption is output.
[0058] The training process can be:
[0059] Each feature in the sample is labeled with an independent variable, and the correlation coefficient is added:
[0060] y=w1x1+w2x2+w3x3+w4x4+....+w n x n
[0061] m samples form m n-variable linear equations. The above variables need to be assigned uniformly. For example, if the average driving speed is 1.63m / s, then each sample is substituted into the variable after removing the unit, which is 1.63 here. Another example is the sweeping speed F7, 7 is uniformly substituted into the variable. This is considered a regulation but needs to be unified for each sample.
[0062] Then use sklearn to solve the normal equation, call its API, pass in the X data and y target value. Set the intercept parameter to true / false according to the scenario. In this scenario, there is no intercept and it is set to false. Finally, the solution of the normal equation is obtained by the following method: numpy.linalg.inv(XTdot(X)).dot(X,T).dot(y);
[0063] Thus we can obtain the above w1,w2,w3,w4,...,w n The value of the coefficient.
[0064] 5. Dynamic generation of driving instructions
[0065] After the above linear regression training, the remaining fuel volume is obtained when the starting point of the working section is reached. According to the remaining fuel volume, the linear regression model is used to find the operating parameters that output the lowest energy consumption. The corresponding energy consumption is output, and then the input features corresponding to the lowest energy consumption are obtained. The driving behavior instructions extracted based on the above features are such as an average driving speed of 1.63m / s, full sweeping operation, strong sweeping, sweeping disk speed F7, dust reduction, and high-pressure water. The captain of the operation side vehicle broadcasts this driving behavior instruction through the intercom when passing the ideal section. This strong vehicle-side method is used to control the energy consumption of the continuous working section to be the lowest.
[0066] Beneficial technical effects of the present invention
[0067] The present invention implements a driving behavior recommendation technology based on the energy consumption of sanitation operations. With the help of data warehouse data cleaning, the road sections with high feasibility are obtained as ideal sections, and then the energy consumption level classification is obtained, that is, a high-quality sample set is obtained. Based on this sample set, an AI model is obtained with the help of Sklearn training, and finally the vehicle energy consumption set is dynamically obtained, and driving instructions are generated based on the minimum energy consumption and sent to the vehicle side. In summary, the present invention uses the AI model to dynamically recommend driving instructions to the vehicle side, realizes refined energy consumption control, and is of great significance to energy saving and cost reduction.
[0068] The above is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, which all belong to the protection scope of the present invention.
Claims
1. A method for recommending driving behavior of sanitation vehicles based on operation energy consumption, characterized in that: include: S1: The road sections with high feasibility are obtained through data cleaning as ideal road sections; S2: Classify energy consumption levels; S3: extract sample features; S4: Train the AI model based on the samples and output the corresponding energy consumption; S5: Generate driving behavior instructions based on minimum energy consumption and send them to the vehicle.
2. The method according to claim 1, wherein: In step S1, the road sections with high feasibility meet the following two conditions: (1) The network signal strength is the best to ensure the completeness of gateway data acquisition; (2) Operation is performed with the operation mode turned on.
3. The method according to claim 2, wherein: In step S1, the data cleaning process includes: The data platform obtains time series data from the gateway; Multiple data sources are associated through time series timestamps and vehicle numbers to obtain a target time series wide table; Filter the same type of vehicles and aggregate and calculate the mean A1 and variance A2 of the fuel consumption of the road section according to the road section dimension; The same type of vehicles mentioned above are filtered out, and the mean fuel consumption B1 and variance B2 of each vehicle on the road section are obtained by aggregation calculation according to the road section and vehicle dimensions.
4. The method according to claim 3, wherein: In step S2, vehicles whose mean B1 is less than mean A1 and whose variance B2 is less than variance A2 are marked as Class I energy consumption vehicles, otherwise they are marked as Class II energy consumption vehicles.
5. The method according to claim 4, wherein: In step S3, the vehicle number set of Class I energy-consuming vehicles is obtained, and the energy consumption distribution of these vehicles passing through the above-mentioned ideal road section every day is queried through SQL with the help of the data warehouse, where the sample extraction features include: average driving speed, washing and sweeping operation mode, operation standard, sweeping disk speed, whether dust reduction is used, and whether high-pressure water is used.
6. The method according to claim 5, wherein: In step S3, the decision engine takes the driving behavior instruction as an input and obtains the corresponding vehicle weight set as another input; then the AI model is trained and deployed on the vehicle side / cloud platform to realize dynamic voice broadcasting of the driving behavior instructions with the lowest energy consumption when passing through the ideal road section.
7. The method according to claim 6, wherein: For each driving behavior instruction, the weight at the starting point of the corresponding ideal road section is associated, and the weight is represented by the remaining oil and water volume conversion, that is, the remaining oil and water volume are also added as features in the feature extraction.
8. The method according to claim 7, wherein: In step S4, the training process includes: Each feature in the sample is labeled with an independent variable, and the correlation coefficient is added: y=w1x1+w2x2+w3x3+w4x4+....+w n x n m samples form m n-variable linear equations; Then use sklearn to solve the normal equation, call the API, and pass in the X data and y target value; Set the intercept parameter to true / false according to the scenario. In this scenario, there is no intercept and it is set to false. Finally, the solution of the normal equation is obtained by the following method: numpy.linalg.inv(XTdot(X)).dot(X,T).dot(y); Thus we can obtain the above w1,w2,w3,w4,...,w n The value of the coefficient.
9. The method according to claim 8, wherein: In step S5, the remaining fuel volume is obtained when the starting point of the operation section is reached. According to the remaining fuel volume, the operation parameters with the lowest energy consumption are output in the linear regression model, the corresponding energy consumption is output, and then the input features corresponding to the lowest energy consumption are obtained.
10. A sanitation vehicle driving behavior recommendation system based on operation energy consumption, characterized in that: include: An ideal road section acquisition module is used to obtain a road section with high feasibility as an ideal road section through data cleaning; Energy consumption level classification module, used for energy consumption level classification; A sample feature extraction module is used to extract sample features; Model training module, used to train AI models based on samples and output corresponding energy consumption; The driving behavior instruction dynamic generation module is used to generate driving behavior instructions based on the minimum energy consumption and send them to the vehicle side.