A Big Data-Based Evaluation System and Method for Ecological Driving of Bus Drivers

Abstract

This invention discloses a big data-based ecological driving evaluation system for bus drivers, comprising: a data preprocessing module, which classifies long-distance and short-distance trips based on matching results; a quality estimation module, which receives information from the data preprocessing module and estimates the vehicle quality for short-distance trips and classifies the quality level for long-distance trips; a working condition identification and classification module, which receives information from the data preprocessing module and determines the traffic category for short-distance trips and the traffic condition classification result for long-distance trips; and a scoring module, which receives and inputs the quantitative results from the quality estimation module and the working condition identification and classification module, establishes a correspondence between energy consumption and driver skill level, and evaluates the driver's ecological driving level. This system uses a scoring method to evaluate the ecological driving level of bus drivers. This invention also provides a big data-based method for evaluating the ecological driving of bus drivers.

Description

Technical Field

[0001] This invention relates to a big data-based ecological driving evaluation system and method for bus drivers, belonging to the field of driving behavior evaluation technology. Background Technology

[0002] The transportation industry is a major energy consumer, and this enormous energy consumption leads to significant emissions of greenhouse gases such as carbon dioxide. To achieve the dual carbon goals of peak carbon emissions and carbon neutrality as soon as possible, it is essential to promote the green transformation of the transportation sector. However, limited by traditional fuel-saving technologies that have almost reached their fuel-saving potential limits, the difficulty in improving urban congestion in the short term, and the slow market penetration of electric vehicles, eco-driving has become a good way to reduce vehicle energy consumption in the short term.

[0003] Eco-driving refers to changing poor driving behaviors and cultivating energy-saving driving habits, starting with the driver themselves. Related research shows that energy consumption differences caused by different drivers' driving habits can reach up to 32%. Eco-driving can reduce vehicle energy consumption and carbon dioxide emissions, but its degradation characteristics necessitate continuous feedback. Research indicates that unless people can immediately see the beneficial consequences of their actions, it is difficult to maintain such behavior over time. Therefore, providing drivers with fair and reasonable feedback on the ecological impact of their driving behavior helps maintain eco-driving behavior in the long term. Meanwhile, for public transportation companies, economic incentive systems are typically established to encourage drivers to drive energy-efficiently and reduce their energy costs. However, the biggest challenge facing such systems is how to fairly and reasonably evaluate drivers' eco-driving levels. Therefore, establishing an eco-driving assessment method for public transportation company drivers is of great significance.

[0004] Current research primarily evaluates drivers' ecological driving based on driving events or characteristic indicators, neglecting in-depth analysis of the driving process itself. External factors unrelated to driving behavior, such as traffic and weather, significantly influence these events or indicators. Using traditional ecological assessment methods, a driver with high energy efficiency might receive a lower ecological score in congested traffic due to frequent acceleration or deceleration. Furthermore, traditional methods rely heavily on expert experience, increasing the subjectivity of the evaluation results. Summary of the Invention

[0005] This invention designs and develops a big data-based ecological driving evaluation system for bus drivers. The system comprehensively considers the impact of vehicle weight changes, road traffic conditions, and weather on bus energy consumption, and uses a scoring method to evaluate the ecological driving level of bus drivers.

[0006] This invention also designed and developed a big data-based method for evaluating the ecological driving of bus drivers. By utilizing the operational characteristics of buses, a correlation was established between energy consumption and the driver's ecological driving level, and the ecological evaluation of driving behavior was summarized and fed back to the driver.

[0007] The technical solution provided by this invention is as follows:

[0008] A big data-based evaluation system for the ecological driving of bus drivers includes:

[0009] The data preprocessing module is used to clean and complete the big data of buses, and to perform map matching on the bus operation routes, and then divide the long trips and micro trips based on the matching results.

[0010] The quality estimation module receives information from the data preprocessing module and estimates the vehicle quality for short trips and classifies the quality level for long trips.

[0011] The working condition identification and classification module receives information from the data preprocessing module to determine the traffic category of micro-trips and the traffic condition classification results of long-trips.

[0012] The scoring module receives and inputs the quantitative results from the quality estimation module and the operating condition identification and classification module, establishes a correlation between energy consumption and driver skill level, and evaluates the driver's ecological driving level.

[0013] Preferably, it is characterized in that,

[0014] The data preprocessing module includes: a data cleaning section, a data completion section, a map matching section, and a trip division section;

[0015] The quality estimation module includes: a micro-travel quality estimation part and a long-travel quality level classification part;

[0016] The operating condition identification and classification module includes: a historical data clustering analysis part and a long-distance traffic condition classification part.

[0017] Preferably, the aforementioned big data-based bus driver ecological driving evaluation system includes:

[0018] Step 1: Obtain the raw operating data of the bus and use the data preprocessing module to obtain micro-trips and long-trips;

[0019] Step 2: Based on the micro-travel and long-travel outputs from the data preprocessing results, estimate the quality of the micro-travel in the quality estimation module and classify the quality level of the long-travel.

[0020] Step 3: Output the traffic category of the micro-trip in the working condition identification and classification module, and classify the traffic conditions of the long-trip.

[0021] Step 4: Group the long-distance quality level and long-distance traffic condition classification results according to the grouping principle. Utilize the energy consumption data from the bus big data to score the driver's ecological driving level within each group. The scoring formula is as follows:

[0022]

[0023] In the formula, i represents the case where temperature, vehicle weight class, and traffic category are the same, j represents the long distance of any bus in i, and x ij Let Sore be the energy consumption per unit distance for any bus in i with a long journey j. ij This indicates the score for the level of eco-driving during the long journey.

[0024] Preferably, step one includes:

[0025] Processing abnormal data where the sampling point speed value exceeds the vehicle's maximum design speed and where the sampling point latitude and longitude values ​​are abnormal:

[0026] Data completion is performed to address two types of data defects: missing fields in sample point data and missing sample point data.

[0027] Map matching was performed to determine the actual latitude and longitude coordinates of bus stops. Long trips were defined based on the date and the charging status of the buses. The dataset was then divided into micro-trips between adjacent stops by comparing the spatial distance between the coordinates of sampled points in the data segment and the actual station coordinates.

[0028] The latitude and longitude values ​​of each station are obtained by querying the map and combining it with the dataset.

[0029] Preferably, the method for calculating the spatial distance between the coordinates of the sampling points and the actual station coordinates in the data segment includes:

[0030] Set latitude and longitude coordinates as X A (Long A ,Lat A ) and X B (Long B ,Lat B The spatial distance between two coordinates can be calculated using the distance formula:

[0031]

[0032] In the formula, L x Let L be the distance between points A and B along the east-west direction. yR is the distance between points A and B along the north-south direction, where rad is the unit of radians and its value is 180 / π; R is the Earth's radius and its value is 6371004m.

[0033] Calculate the spatial distance between the coordinates of the sampling points in the dataset and the actual station coordinates. The sampling point with the smallest distance is identified as the boundary point between data segments between adjacent stations. The objective function is:

[0034] L j =min[Distance(X) i , X′ j )], i=1, 2,..., Nj=1, 2,..., M;

[0035] In the formula: X i X′ represents the latitude and longitude coordinates of the sampling points in the dataset. j The extracted station latitude and longitude coordinates; N is the sample size of the dataset; M is the number of bus stops;

[0036] If road segments overlap, a penalty-inducing objective function is introduced when calculating the distance:

[0037] L j =min[Distance(X) i X j ′)+α|S i -S1-β|], i=1, 2,..., Nj=8, 9;

[0038] In the formula: α is the penalty coefficient; S i β represents the cumulative mileage at each sampling point; β is the average driving distance from the starting point to each sampling point.

[0039] The optimal solution of the objective function is the boundary point between data segments between adjacent stations, which can then divide the dataset into micro-travels between adjacent stations; the divided micro-travels can be further divided into effective micro-travels and undesirable micro-travels.

[0040] The criteria for poor micro-travel are: continuous gaps or loss of sampling point data within a data segment, and the inability to fill in the missing data.

[0041] Preferably, the minimum spatial distance between each sampling point and the bus stop is less than or equal to 65m.

[0042] Preferably, in step two, the quality estimation module estimates the micro-stroke quality, including:

[0043] Based on big data from buses, the mass of a bus is obtained through power balancing. This method is used for pure electric buses.

[0044]

[0045] In the formula, T is the motor torque, n is the motor speed, and u is the motor speed. a For vehicle speed, obtained from big data on buses, η T Let f be the mechanical efficiency of the transmission system, f be the rolling resistance coefficient, and C be the rolling resistance coefficient. D Where A is the air resistance coefficient, δ is the frontal area, and δ is the rotational mass conversion factor.

[0046] The formula for acceleration 'a' is:

[0047]

[0048] In the formula, i represents the i-th collected point, and v i Let a be the velocity corresponding to the i-th point. i Let Δt = t be the acceleration corresponding to the i-th point. i -t i-1 Δt is the time interval between the i-th point and the (i-1)-th point.

[0049] Preferably, in step two, the classification of long-stroke quality into four levels includes:

[0050] Level 1 has the lowest average mass for long-distance travel, and most short-distance travel is empty when carrying passengers.

[0051] Level 2 refers to vehicles that are mostly short-haul and not empty, with a relatively small passenger capacity.

[0052] Level 3 refers to short-distance trips with a relatively large passenger capacity, but not at full capacity.

[0053] Level 4 has the highest average mass for long-distance trips, while most short-distance trips are fully loaded, corresponding to passenger volume during peak hours.

[0054] Preferably, step three includes:

[0055] K-means clustering is used to group micro-travels with similar characteristics into one class. The average value of the feature parameters of each class represents the parameter characteristics of the corresponding working condition. The traffic categories are defined as: smooth, normal, and congested.

[0056] The classification of traffic conditions for long-distance trips is defined as the proportion of different traffic categories for micro-trips, including:

[0057] Smoothness ratio:

[0058] In the formula, num1 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k1 represents the proportion of smooth micro-travels to the total number of micro-travels;

[0059] Normal proportion:

[0060] In the formula, num2 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in long strokes, and k2 represents the proportion of normal micro-travels to the total number of micro-travels;

[0061] Congestion ratio:

[0062] In the formula, num3 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k3 represents the proportion of smooth micro-travels to the total number of micro-travels.

[0063] num1+num2+num3=Num, k1+k2+k3=100%;

[0064] The values ​​of k1, k2, and k3 for each long-distance trip are obtained as a description of the traffic conditions of long-distance trips and as one of the grouping criteria.

[0065] Among them, long-distance traffic conditions are described by the values ​​of k1, k2, and k3. Different long distances are classified as the same traffic condition if the values ​​of k1, k2, and k3 do not exceed 4%.

[0066] Preferably, the grouping principle in step four is as follows:

[0067] Vehicles in the same group with the same weight rating indicate that vehicle weight has a similar impact on energy consumption.

[0068] When the traffic conditions in each group are approximately the same, the judgment is made based on the difference of 4% between k1, k2, and k3 in the long journey. At the same time, the scoring module standardizes the temperature for each group, dividing the temperature into 5 intervals: below -10℃, -10℃ to -1℃, 0℃ to 12℃, 13℃ to 26℃, and above 26℃. The temperature intervals in each group are kept consistent, and the same temperature interval is considered to have the same degree of influence on energy consumption.

[0069] The scoring module outputs the driver's economic driving score in units of long-distance travel, and records the start and end times and start and end SOC of the long-distance travel, thus completing the evaluation of the driver's eco-driving performance.

[0070] The beneficial effects of this invention are as follows:

[0071] 1. The present invention provides a big data-based ecological driving evaluation system for bus drivers, in which the data preprocessing module divides micro-trips by adjacent stops, effectively utilizing the operational characteristics of buses.

[0072] 2. The bus driver ecological driving evaluation system based on big data described in this invention has a working condition identification and classification module that divides road traffic conditions through cluster analysis and quantifies the road factors that generate differences in energy consumption.

[0073] 3. The present invention provides a big data-based evaluation system for the ecological driving of bus drivers. Its scoring module utilizes the concept of controlled variables, comprehensively considers the impact of vehicle weight and road conditions on energy consumption, and establishes a correspondence between energy consumption and the driver's ecological driving level.

[0074] 4. The big data-based ecological driving evaluation system for bus drivers described in this invention can provide feedback to bus companies and drivers, helping bus companies establish reasonable reward and punishment mechanisms and prompting drivers to improve their ecological driving skills, thereby reducing bus energy consumption and bus company operating costs. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the bus driver ecological driving scoring system described in this invention.

[0076] Figure 2 This is a flowchart illustrating the division of long and short strokes as described in this invention.

[0077] Figure 3 This is a schematic diagram of the lever method described in this invention.

[0078] Figure 4 This is a comparison chart of vehicle speed curves after data preprocessing by the data preprocessing module of this invention.

[0079] Figure 5 This is a flowchart of the quality estimation module described in this invention.

[0080] Figure 6 The clustering effect described in this invention Figure 1 .

[0081] Figure 7 The clustering effect described in this invention Figure 2 .

[0082] Figure 8 This is an example diagram of the ecological driving score for bus drivers described in this invention. Detailed Implementation

[0083] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0084] like Figure 1-8As shown, this invention provides a big data-based evaluation system for the ecological driving of bus drivers. This system comprehensively considers the impact of vehicle weight changes, road traffic conditions, and weather on bus energy consumption, establishes a correlation between energy consumption and the driver's ecological driving level, and evaluates the driver's ecological driving level using a scoring method. The system first preprocesses the bus big data, dividing the bus trips into long-distance and short-distance trips. In short-distance trips, the bus mass is estimated using a mass calculation formula, and a traffic category is defined based on operating conditions. Based on the mass estimation results of short-distance trips and the traffic category definitions, the system classifies the vehicle mass level and traffic conditions for long-distance trips. The system uses a controlled variable method to ensure that factors other than the driver have approximately the same impact on energy consumption. Due to the differences in bus mileage, the energy consumption per unit mile is substituted into the ecological driving scoring formula to calculate the driver's score, thus achieving the purpose of evaluating the bus driver's ecological driving.

[0085] like Figure 1 As shown, the big data-based bus driver ecological driving evaluation system includes: a data preprocessing module, a quality estimation module, a working condition identification and classification module, and a scoring module. The quality estimation module estimates the vehicle quality for micro-trips based on a quality calculation formula; the difference in vehicle quality between different micro-trips reflects changes in passenger load during bus operation. The working condition identification and classification module defines the traffic categories for micro-trips, which represent the degree of road congestion. The outputs from the quality estimation and working condition identification and classification modules are input into the scoring module, where a correlation is established between energy consumption and the driver's ecological driving level, thus evaluating the driver's ecological driving ability.

[0086] The data preprocessing module includes: data cleaning, data completion, map matching, and trip segmentation. This module addresses noise caused by external environmental interference and fluctuations, extracts valid data from bus operations, and divides the data into trip segments. Data cleaning, data completion, map matching, and trip segmentation may be performed intermittently as needed.

[0087] The data cleaning section is used to delete outlier data values. Outliers refer to speed outliers that exceed the maximum speed limit of buses and latitude and longitude outliers that deviate from the normal bus operating route. Among them, speed outliers that exceed the maximum speed limit of buses are replaced with the maximum speed limit of buses. Latitude and longitude outliers that deviate from the normal bus operating route are directly deleted, which means deleting all data fields in the row containing the outlier.

[0088] The data completion section fills in missing data and partially missing data fields. Missing data refers to data segments that do not conform to the normal sampling frequency. If there is a long-term data gap, the missing data segment is deleted; otherwise, interpolation is used to fill in the missing data segment. Interpolation methods can include linear interpolation, cubic spline interpolation, etc. Partially missing data fields refer to data segments that conform to the normal sampling frequency but have incomplete information. For example, speed is filled in by dividing the spatial distance between the sampling points before and after the missing data by the sampling time difference. If the vehicle state is different before and after the missing data, the missing or lost data is filled in by data from adjacent sampling points where the vehicle is off. Since the accelerator pedal and brake pedal states are generally not under force simultaneously during vehicle operation, the accelerator pedal and brake pedal openings are filled in using a lever method.

[0089] The map matching component identifies normal bus operating routes, eliminating route deviations caused by bus charging, maintenance, or unforeseen circumstances. A normal bus operating route refers to the one-way trajectory of a bus from its origin to its destination, passing through all stops on its designated route. Latitude and longitude values ​​are the primary reference values, while vehicle speed is a secondary reference value. Because there is a discrepancy between the latitude and longitude displayed in the map software and the actual collected latitude and longitude, reasonable deviation values ​​are set based on vehicle length, the number and width of road lanes, and the length of bus stop signs. Within this reasonable deviation range, the actual sampling points are matched to their positions on the map.

[0090] like Figure 1 As shown, the trip segmentation section divides bus big data into long trips and micro trips. Long trips refer to one-way segments from the originating station to the terminal station, primarily determined by identifying the latitude and longitude of the originating and terminal stations and extracting all valid data between them. Micro trips refer to segments between adjacent bus stops, also determined by identifying the latitude and longitude of the stops and extracting all valid data between them. Long trips consist of a number of micro trips, and their order within the long trip is not considered. Typically, a single long trip is driven by the same driver; therefore, the driver's driving skills are scored using the long trip as the smallest unit. In the quality estimation and traffic condition identification / classification modules, quality estimation and traffic category definition are performed using micro trips as the smallest unit, followed by quality level classification and traffic condition identification for long trips.

[0091] This invention also provides a method for evaluating the ecological driving of bus drivers based on big data, which uses the aforementioned big data-based bus driver ecological driving evaluation system to evaluate the ecological driving of drivers, including:

[0092] Step 1: Obtain the raw operating data of the bus and use the data preprocessing module to obtain micro-trips and long-trips;

[0093] Processing of abnormal data, including sampling point speed values ​​exceeding the vehicle's maximum design speed and abnormal sampling point latitude and longitude values:

[0094] To address two types of data defects—missing data fields at sampling points and missing data at sampling points—data supplementation is performed.

[0095] Map matching was performed to determine the actual latitude and longitude coordinates of bus stops. Long trips were defined based on the date and the charging status of the buses. The dataset was then divided into micro-trips between adjacent stops by comparing the spatial distance between the coordinates of sampled points in the data segment and the actual station coordinates.

[0096] The latitude and longitude values ​​of each station are obtained by querying the map and combining it with the dataset.

[0097] The methods for calculating the spatial distance between the coordinates of sampling points and the actual station coordinates in the data segment include:

[0098] Using latitude and longitude coordinates X A (Long A Lat A ) and latitude and longitude coordinates X B (Long B Lat B For example, the spatial distance between two coordinates can be calculated using the distance formula:

[0099]

[0100] In the formula, L x Let L be the distance between points A and B along the east-west direction. y R is the distance between points A and B along the north-south direction, where rad is the unit of radians and its value is 180 / π; R is the Earth's radius and its value is 6371004m.

[0101] Calculate the spatial distance between the coordinates of the sampling points in the dataset and the actual station coordinates. The sampling point with the smallest distance is identified as the boundary point between data segments between adjacent stations. The objective function is:

[0102] L j =min[Distance(X) i , X′ j )], i=1, 2,..., Nj=1, 2,..., M;

[0103] In the formula: X i X′ represents the latitude and longitude coordinates of the sampling points in the dataset. j The extracted station latitude and longitude coordinates; N is the sample size of the dataset; M is the number of bus stops;

[0104] If road segments overlap, a penalty-inducing objective function is introduced when calculating the distance:

[0105] L j =min[Distance(X) i X j ′)+α|S i -S1-β|], i=1,2,...,Nj=8,9; where: α is the penalty coefficient; S i β represents the cumulative mileage at each sampling point; β is the average driving distance from the starting point to each sampling point.

[0106] The optimal solution of the objective function is the boundary point between data segments between adjacent stations, which can then divide the dataset into micro-travels between adjacent stations; the divided micro-travels can be further divided into effective micro-travels and undesirable micro-travels.

[0107] The criteria for poor micro-travel are: continuous gaps or loss of sampling point data within the data segment, and the missing data cannot be filled in; the minimum spatial distance between each sampling point and the bus stop is less than or equal to 65m, which is insufficient to meet the requirements for boundary point identification.

[0108] Step 2: Based on the micro-travel and long-travel outputs from the data preprocessing results, estimate the quality of the micro-travel in the quality estimation module and classify the quality level of the long-travel.

[0109] Based on big data from buses, the bus mass is obtained through power balancing. The mass estimation module estimates the bus mass based on the micro-travel mass. The vehicle power equation is as follows:

[0110]

[0111] In the formula, η T The mechanical efficiency of the transmission system is represented by m, the mass of the vehicle is represented by f, and the rolling resistance coefficient is represented by u. a Indicates vehicle speed, i represents gradient, and C represents speed. D δ is the air drag coefficient, A is the frontal area, δ is the rotational mass conversion factor, and a is the acceleration.

[0112] In this invention, as a preferred embodiment, a pure electric bus is used, with power provided by a motor. Then:

[0113]

[0114] In the formula, T is the motor torque and n is the motor speed.

[0115] The mass of the bus is obtained through power balancing.

[0116]

[0117] In the formula, T is the motor torque, n is the motor speed, and u is the motor speed. a Vehicle speed, obtained from big data analysis of public transportation; η T Let f be the mechanical efficiency of the transmission system, f be the rolling resistance coefficient, and C be the rolling resistance coefficient. D Here, A is the air drag coefficient, δ is the frontal area, and δ is the rotational mass conversion factor.

[0118] The formula for acceleration 'a' is:

[0119]

[0120] In the formula, i represents the i-th collected point, and v i Let a be the velocity corresponding to the i-th point. i Let Δt = t be the acceleration corresponding to the i-th point. i -t i-1 Δt is the time interval between the i-th point and the (i-1)-th point;

[0121] The long-stroke quality rating is divided into four levels, including:

[0122] Level 1 has the lowest average mass for long-distance travel, and most short-distance travel is empty when carrying passengers.

[0123] Level 2 refers to vehicles that are mostly short-haul and not empty, with a relatively small passenger capacity.

[0124] Level 3 refers to short-distance trips with a relatively large passenger capacity, but not at full capacity.

[0125] Level 4 has the highest average mass for long-distance trips, while most short-distance trips are fully loaded, corresponding to passenger volume during peak hours.

[0126] Step 3: Output the traffic category of the micro-trip in the working condition identification and classification module, and classify the traffic conditions of the long-trip.

[0127] K-means clustering is used to group micro-travels with similar characteristics into one class. The average value of the typical characteristic parameter of each class represents the parameter characteristics of that class of working conditions. The traffic categories are defined as: smooth, normal, and congested.

[0128] Traffic conditions for long-distance trips are categorized into the proportions of different traffic types for micro-trips, including:

[0129] Smoothness ratio:

[0130] In the formula, num1 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k1 represents the proportion of smooth micro-travels to the total number of micro-travels;

[0131] Normal proportion:

[0132] In the formula, num2 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in long strokes, and k2 represents the proportion of normal micro-travels to the total number of micro-travels;

[0133] Congestion ratio:

[0134] In the formula, num3 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k3 represents the proportion of smooth micro-travels to the total number of micro-travels.

[0135] num1+num2+num3=Num, k1+k2+k3=100%;

[0136] Among them, long-distance traffic conditions are described by the values ​​of k1, k2, and k3. Different long distances are classified as the same traffic condition if the values ​​of k1, k2, and k3 do not exceed 4%.

[0137] Step 4: Group the long-distance quality level and long-distance traffic condition classification results according to the grouping principle. Utilize the energy consumption data from the bus big data to score the driver's ecological driving level within each group. The scoring formula is as follows:

[0138] In the formula, i represents the case where temperature, vehicle weight class, and traffic category are the same, j represents the long distance of any bus in i, and x ij Let Sore be the energy consumption per unit distance for any bus in i with a long journey j. ij This indicates the score for the level of eco-driving performance during this long journey;

[0139] The grouping principle is as follows:

[0140] Vehicles in the same group with the same weight rating indicate that vehicle weight has a similar impact on energy consumption.

[0141] When the traffic conditions in each group are approximately the same, the judgment is made based on the difference of 4% between k1, k2, and k3 in the long journey. At the same time, the scoring module standardizes the temperature for each group, dividing the temperature into 5 intervals: below -10℃, -10℃ to -1℃, 0℃ to 12℃, 13℃ to 26℃, and above 26℃. The temperature intervals in each group are kept consistent, and the same temperature interval is considered to have the same degree of influence on energy consumption.

[0142] The scoring module outputs the driver's economic driving score in units of long-distance travel, and records the start and end times and start and end SOC of the long-distance travel, thus completing the evaluation of the driver's eco-driving performance.

[0143] Example

[0144] The data in this embodiment comes from the operation data of pure electric buses on the same route in a certain city. The data collection time span is one year. The on-board terminal samples at 15-second intervals and saves the collected real-time data. The data fields include running time, vehicle speed, vehicle status, charging status, total voltage, total current, cumulative mileage, battery SOC, drive motor speed, drive motor torque, drive motor speed, drive motor controller temperature, motor controller input voltage, motor controller DC bus voltage, motor controller DC bus current, latitude and longitude, accelerator pedal status, and brake pedal status.

[0145] Step 1: Obtain the original bus operation data, and obtain the micro-trip and long-trip data through the data preprocessing module;

[0146] To address the issues with the data, the following cleaning and completion processes were performed on the original data in the data cleaning and completion section:

[0147] (1) The following processing was performed on the abnormal data, including some sampling point speed values ​​exceeding the vehicle's maximum design speed and abnormal sampling point latitude and longitude values:

[0148] ① The speed values ​​at some sampling points exceeded the vehicle's designed maximum speed of 69 km / h. These abnormal speed values ​​were uniformly changed to the vehicle's maximum speed of 69 km / h.

[0149] ② Data segments deviating from the normal bus operating routes were filtered out based on latitude and longitude coordinates, and these abnormal data points were removed. The data after removal only contains valid data of buses operating on normal routes.

[0150] (2) To address the issues of missing data fields and lost sampling points, data supplementation is performed based on the following two data defect scenarios:

[0151] ① If a sampling point has a missing or incomplete data point: Determine if the vehicle state is the same before and after the defective data point. If so, use linear interpolation to complete the data. Specifically, speed is filled using the average speed calculated by dividing the spatial distance between the sampling points before and after the defective data point by the sampling time difference. The accelerator pedal and brake pedal states are filled using a lever method, as these are generally not simultaneously under force during vehicle operation. If the vehicle state is different before and after the defective data point, the missing or incomplete data is filled using data from adjacent sampling points where the vehicle is in an off state. Figure 3 As shown.

[0152] ② Multiple consecutive sampling points have missing data or multiple sampling points are lost consecutively: Determine if the spatial distance (latitude and longitude) between sampling points before and after the defective data segment is less than 65m. If it is less than 65m, supplement the defective data every 15 seconds using the same data from the previous sampling point, with speed values ​​supplemented as 0km / h. If the spatial distance is greater than or equal to 65m, no defective data supplementation operation will be performed. The reason for selecting an upper limit of 65m for the spatial distance is that analysis of the original dataset revealed that the average distance traveled by a vehicle in 15 seconds is 61.75m. Considering that the accuracy of GPS equipment is affected by environmental factors, the allowable range for the supplementation operation is expanded to 65m.

[0153] In the map matching section, the actual latitude and longitude coordinates of bus stops are determined. Then, in the trip segmentation section, long trips are divided based on the date and the vehicle's charging status. Subsequently, by comparing the spatial distance between the coordinates of the sampling points in the data segment and the actual station coordinates, the sampling point with the smallest distance and not exceeding 65m is identified as the boundary point of the micro-trip, and the dataset is divided into micro-trips between adjacent stations. The specific steps are as follows:

[0154] (1) Extracting station latitude and longitude

[0155] The latitude and longitude values ​​of each station are obtained by querying the map and combining it with the dataset.

[0156] The method for calculating the spatial distance between the coordinates of the sampling points and the actual station coordinates in the data segment is as follows. Using latitude and longitude coordinates X... A (Long A Lat A ) and latitude and longitude coordinates X B (Long B Lat B For example, the spatial distance between two coordinates can be calculated using the distance formula:

[0157]

[0158] In the formula, L x Let L be the distance between points A and B along the east-west direction. y R is the distance between points A and B along the north-south direction, where rad is the unit of radians and its value is 180 / π; R is the Earth's radius and its value is 6371004 meters.

[0159] Calculate the spatial distance between the coordinates of the sampled points in the dataset and the actual station coordinates, and identify the sampled point with the smallest distance as the boundary point between data segments between adjacent stations. The objective function is as follows:

[0160] L j =min[Distance(X) i , X′ j)], i=1, 2,..., Nj=1, 2,..., M;

[0161] In the formula: X i X′ represents the latitude and longitude coordinates of the sampling points in the dataset. j The extracted station latitude and longitude coordinates; N is the sample size of the dataset; M is the number of bus stops;

[0162] If road segments overlap, a penalty-inducing objective function is introduced when calculating the distance:

[0163] L j =min[Distance(X) i X j ′)+α|S i -S1-β|], i=1, 2,..., Nj=8, 9

[0164] In the formula: α is the penalty coefficient; S i β represents the cumulative mileage at each sampling point; β represents the average driving distance from the starting point to each sampling point.

[0165] The optimal solution of the objective function is the boundary point between data segments of adjacent stations. The dataset can then be divided into micro-travels between adjacent stations. These micro-travels can be classified as valid micro-travels and defective micro-travels. The criteria for defective micro-travels are: continuous gaps or loss of sampling point data within the data segment, and the inability to fill in the missing data; and the minimum spatial distance between each sampling point and the bus station is greater than 65m, which is insufficient to meet the boundary point identification requirements.

[0166] Step 2: Based on the micro-travel and long-travel outputs from the data preprocessing results, estimate the micro-travel quality and classify the long-travel quality level in the quality estimation module;

[0167] Since Δt typically ranges from 0 to 30 seconds, the average acceleration at a sampling point may not be the true acceleration at that point. Furthermore, there are sampling points that do not conform to the equation for vehicle drag. Therefore, sampling points must be screened before mass estimation.

[0168] (1) Screen all driving points. According to the equation of motion of a car, the driving resistance considered when calculating the mass is rolling resistance, air resistance and acceleration resistance. When a pure electric vehicle brakes, the motor will perform braking energy recovery. When the recovery strategy of an unknown model is not known, the vehicle power during regenerative braking cannot be accurately calculated. Therefore, the braking point can be ignored at this time. At the same time, the coasting point is also a deceleration condition, so it is also excluded. The data used to determine the driving state are the accelerator pedal opening and the brake pedal opening. The state can be directly obtained from the big data of buses.

[0169] The driving point is determined by the following principles: when the accelerator pedal opening is greater than 0 and the brake pedal opening is equal to 0, it is the driving point; when both pedal openings are 0, it is the coasting point; when the accelerator pedal opening is equal to 0 and the brake pedal opening is greater than 0, it is the braking point. All driving points are thus selected.

[0170] (2) Screening possible uniform acceleration points: According to the acceleration calculation formula, the acceleration of each point is actually the average acceleration between two sampling points. There may be data where the acceleration does not match the state, as shown in the table below:

[0171] Table 1 Examples of Acceleration-State Mismatch Data

[0172]

[0173] As in the example above, if the velocity of the previous sampling point is higher than that of the next sampling point, and the average acceleration is negative, it does not match the driving state. In this case, the condition v needs to be added during the filtering process. i >v i-1 To match acceleration with the corresponding state.

[0174] (3) Remove points that do not meet the car driving equation. The car driving equation is not met when the car is turning, the clutch is disengaged, the brake is applied, or the actual gear is in neutral. The method is to calculate the mass point by point and determine whether the mass is within the theoretical mass range. Since the sampling frequency of bus big data is low, the mass calculated by the uniform acceleration points selected in the above steps may not be within the theoretical mass range. Delete the sampling points that are not within the theoretical mass range, and then consider all the remaining points to be valid uniform acceleration points, and the mass of the bus can be calculated.

[0175] The data used to determine the driving status are the accelerator pedal opening and the brake pedal opening, as shown in the table below:

[0176] Table 2 Comparison of Driving Status and Pedal Opening

[0177]

[0178] The average vehicle weight for each long-distance journey is calculated and categorized into four levels from lowest to highest:

[0179] Table 3 Vehicle Weight Classification Table

[0180]

[0181] This completes the classification of long-distance bus weight levels.

[0182] Step 3: Based on the micro-trips and long-trips output from the data preprocessing results, the traffic conditions of long-trips are classified in the work condition identification and classification module; in the historical data clustering analysis part of the work condition identification and classification module, one week's worth of bus operation data is randomly selected as clustering samples, and the following 12 feature parameters can be extracted for each micro-trip:

[0183] Table 4 Feature Parameter Table

[0184]

[0185]

[0186] The calculation formulas for each parameter are shown below:

[0187] (1) Average vehicle speed

[0188]

[0189] In the formula, n is the total time of the working segment, and v i Let i represent the vehicle speed at time i.

[0190] (2) Standard deviation of vehicle speed v std ;

[0191]

[0192] (3) Average acceleration during the acceleration phase

[0193]

[0194] In the formula, a i Let be the acceleration at time i;

[0195] (4) Average deceleration during the deceleration phase

[0196]

[0197] (5) Acceleration time ratio T acc ;

[0198]

[0199] (6) Deceleration time ratio T dec ;

[0200]

[0201] (7) Uniform speed time ratio T cons ;

[0202]

[0203] (8) Idle time ratio T idle :

[0204] T idle =1-T acc -T dec -T cons ;

[0205] In the formula, t idle t represents the total idling time in the operating condition segment. acc t represents the total acceleration time in the test segment. dec t represents the total deceleration time in the driving condition segment. all This represents the total time of a single operating segment.

[0206] (9) Average speed of the driving section

[0207]

[0208] In the formula, n is the total time of the working condition segment, and u i This refers to the vehicle speed excluding the idling period.

[0209] Principal component analysis was performed on the above 12 characteristic parameters:

[0210] First use x1, x2, ..., x m Let m = 12 represent the feature parameters extracted above; let i = 1, 2...n represent the n samples contained in the cluster sample, and let [x] represent the feature parameters of the driving segment of the i-th sample. i1 x i2 ,....x im Its characteristic parameter matrix can be represented as:

[0211]

[0212] The calculation steps are as follows:

[0213] (1) Standardize and dimensionless the original feature parameter matrix X to eliminate the order-of-magnitude differences between the feature parameters due to their different dimensions:

[0214]

[0215]

[0216]

[0217]

[0218] Where i = 1, 2...n, j = 1, 2...m (m = 12), Let s be the mean of the j-th feature parameter. j The standard deviation of the j-th feature parameter

[0219] (2) Calculate the correlation coefficient matrix:

[0220]

[0221] In the formula, r ii =1, r ij =r ji r ij It is the correlation coefficient between the i-th feature parameter and the j-th feature parameter.

[0222]

[0223] In the formula, x ki x kj Representing two feature parameters, and The numerator represents the mean of the characteristic parameters. The numerator in this formula represents the covariance of the two parameters, and the denominator represents the square root of the product of the variances of the two parameters.

[0224] (3) Calculate the eigenvalue λ and eigenvector l:

[0225] Solve for the eigenvalues ​​λ and eigenvectors l of the correlation coefficient matrix R, and sort the eigenvalues ​​λ in descending order. Then, the expressions for each principal component can be obtained, and the i-th principal component E... i The coefficients preceding the expression correspond to its eigenvalues ​​λ. i eigenvectors[l i1 l i2 l i3 … l im ].

[0226] (4) Calculate the contribution rate and determine the number of principal components using the obtained eigenvalues ​​λ:

[0227] The i-th principal component E i Contribution rate α i for:

[0228] The cumulative contribution rate α of the first q principal components q for:

[0229] Theoretically, if the cumulative contribution rate of principal components exceeds 85%, it can be considered that the principal components can preserve the original data information well. Therefore, the cumulative contribution rate α is chosen. qThe top q principal components, which are greater than 85%, are used as new feature parameters to represent the original feature parameter matrix. The eigenvalues ​​are then arranged in descending order, and λ... k Defined as the k-th eigenvalue.

[0230] (5) Calculate the new feature parameters:

[0231]

[0232] Based on the principal component analysis results, K-means clustering analysis was selected to classify and define the traffic categories of micro-trips.

[0233] The main steps are as follows:

[0234] (1) First, determine the number of categories, denoted by k1, and randomly select k1 data points from the original dataset as cluster centers;

[0235] (2) For each point in the dataset, calculate its Euclidean distance to each cluster center. If it is closer to a cluster center, assign it to the set to which that cluster center belongs.

[0236] The formula for calculating Euclidean distance is as follows:

[0237]

[0238] (3) After classifying all the data into sets, there are a total of k1 sets. Then, the center of each set is recalculated.

[0239] (4) For all k1 cluster centers, after updating using the iterative method of (2) and (3), the values ​​remain unchanged and the objective function tends to converge. It can be considered that the clustering has achieved the desired result and the algorithm terminates. If the distance between the new cluster center and the original cluster center changes greatly, the process of steps two to five is repeated.

[0240] To determine the traffic category for any microtrip, follow these steps:

[0241] (1) For any micro-travel, extract the feature parameters and perform principal component dimensionality reduction of the feature parameters;

[0242] (2) Compare the principal component scores of the micro-trip with the distances to the cluster centers of the three traffic categories, and determine its category based on the closest distance. The distance calculation formula is as follows:

[0243]

[0244] In the formula, x1 represents the sample data, x2 represents the cluster center of various driving conditions, and x 1k x 2k In this context, k represents the k-th dimension component;

[0245] The historical data clustering analysis defines traffic categories for all micro-trips. Based on the above analysis, the traffic conditions of long-trips are described by the proportion of different traffic categories in the micro-trips. In the long-trip traffic condition classification, the following proportional formula is used to obtain the description of different traffic conditions, that is, the proportion of smooth, normal, and congested traffic in the same long trip:

[0246] (1) Smoothness ratio:

[0247] In the formula, num1 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k1 represents the proportion of smooth micro-travels to the total number of micro-travels.

[0248] (2) Normal proportion:

[0249] In the formula, num2 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in long strokes, and k2 represents the proportion of normal micro-travels to the total number of micro-travels.

[0250] (3) Congestion ratio:

[0251] In the formula, num3 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k3 represents the proportion of smooth micro-travels to the total number of micro-travels.

[0252] num1+num2+num3=Num, k1+k2+k3=100%

[0253] The values ​​of k1, k2, and k3 for each long-distance trip are obtained as a description of the traffic conditions of long-distance trips and as one of the grouping criteria.

[0254] Step 4: In the scoring module, group the results based on the above results to ensure that variables other than the driver have the same impact on energy consumption within the group, and use the scoring formula to score the driver's ecological driving level within the group.

[0255] Vehicles within the same group are grouped based on similar temperature ranges, vehicle weight classes, and traffic conditions. A scoring formula is then used to assign scores within each group.

[0256] In the formula, i represents the case where the temperature, vehicle weight class, and traffic category are the same; j represents any long-distance bus journey in i; x ij Sore represents the energy consumption per unit distance of any bus in i traveling a long distance j; ij This indicates the score for the level of eco-driving during the long journey.

[0257] Ultimately, eco-driving scores were obtained for all drivers based on their travel distance.

[0258] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for evaluating the ecological driving of bus drivers based on big data, characterized in that, include: Step 1: Obtain the raw operating data of the bus and obtain the micro-trip and long-trip data through data preprocessing; Step 2: Based on the micro-travel and long-travel outputs from the data preprocessing results, estimate the quality of the micro-travel and classify the quality level of the long-travel. The long-stroke quality level is divided into four levels, including: Level 1 has the lowest average mass for long-distance travel, and most short-distance travel is empty when carrying passengers. Level 2 refers to vehicles that are mostly short-haul and not empty, with a relatively small passenger capacity. Level 3 refers to short-distance trips with a relatively large passenger capacity, but not at full capacity. Level 4 has the highest average mass for long-distance trips, while most short-distance trips are fully loaded, corresponding to passenger volume during peak hours. Step 3: Output the traffic categories for micro-trips and classify long-trip traffic conditions, including: K-means clustering is used to group micro-travels with similar characteristics into one class. The average value of the feature parameters of each class represents the parameter characteristics of the corresponding working condition. The traffic categories are defined as: smooth, normal, and congested. The classification of traffic conditions for long-distance trips is defined as the proportion of different traffic categories for micro-trips, including: Smoothness ratio: In the formula, num1 represents the number of smooth micro-travels, Num represents the total number of micro-travels contained in the long stroke, and k1 represents the proportion of smooth micro-travels to the total number of micro-travels; Normal proportion: In the formula, num2 represents the number of normal micro-strokes, Num represents the total number of micro-strokes contained in the long stroke, and k2 represents the proportion of normal micro-strokes to the total number of micro-strokes; Congestion ratio: In the formula, num3 represents the number of congested micro-travels, Num represents the total number of micro-travels contained in a long trip, and k3 represents the proportion of congested micro-travels to the total number of micro-travels. num1+num2+num3=Num, k1+k2+k3=100%; The values ​​of k1, k2, and k3 for each long-distance trip are obtained as a description of the traffic conditions of long-distance trips and as one of the grouping criteria. Among them, long-distance traffic conditions are described by the values ​​of k1, k2, and k3. Different long distances are judged as the same traffic condition if the values ​​of k1, k2, and k3 do not exceed 4%. Step 4: Group the long-distance quality level and long-distance traffic condition classification results according to the grouping principle. Utilize the energy consumption data from the bus big data to score the driver's ecological driving level within each group. The scoring formula is as follows: In the formula, i represents the case where temperature, vehicle weight class, and traffic category are the same, j represents the long distance of any bus in i, and x ij Let Sore be the energy consumption per unit distance for any bus in i with a long journey j. ij This indicates the score for the level of eco-driving performance during this long journey; The grouping principle is as follows: Vehicles in the same group with the same weight rating indicate that vehicle weight has a similar impact on energy consumption. When the traffic conditions in each group are approximately the same, the distinction is made based on the 4% difference between k1, k2, and k3 in the long-distance travel. At the same time, the temperature in each group is standardized and divided into 5 intervals: below -10℃, -10℃ to -1℃, 0℃ to 12℃, 13℃ to 26℃, and above 26℃. The temperature intervals in each group are kept consistent, and the same temperature interval is considered to have the same degree of influence on energy consumption. Output the driver's economic driving score in units of long-distance travel, and record the start and end times and start and end SOC of the long-distance travel to complete the evaluation of the driver's eco-driving performance.

2. The method for evaluating the ecological driving of bus drivers based on big data according to claim 1, characterized in that, The evaluation was conducted using a big data-based ecological driving evaluation system for bus drivers, including: The data preprocessing module is used to clean and complete the big data of buses, and to perform map matching on the bus operation routes, and then divide the long trips and micro trips based on the matching results. The quality estimation module receives information from the data preprocessing module and estimates the vehicle quality for short trips and classifies the quality level for long trips. The working condition identification and classification module receives information from the data preprocessing module to determine the traffic category of micro-trips and the traffic condition classification results of long-trips. The scoring module receives and inputs the quantitative results from the quality estimation module and the operating condition identification and classification module, establishes a correlation between energy consumption and driver skill level, and evaluates the driver's ecological driving level.

3. The method for evaluating the ecological driving of bus drivers based on big data according to claim 2, characterized in that, Step one includes: Processing abnormal data where the sampling point speed value exceeds the vehicle's maximum design speed and where the sampling point latitude and longitude values ​​are abnormal: Data completion is performed to address two types of data defects: missing fields in sample point data and missing sample point data. Map matching was performed to determine the actual latitude and longitude coordinates of bus stops. Long trips were defined based on the date and the charging status of the buses. The dataset was then divided into micro-trips between adjacent stops by comparing the spatial distance between the coordinates of sampled points in the data segment and the actual station coordinates. The latitude and longitude values ​​of each station are obtained by querying the map and combining it with the dataset.

4. The method for evaluating the ecological driving of bus drivers based on big data according to claim 3, characterized in that, The method for calculating the spatial distance between the coordinates of the sampling points and the actual station coordinates in the data segment includes: Set latitude and longitude coordinates as X A (Long A ,Lat A ) and X B (Long B ,Lat B The spatial distance between two coordinates can be calculated using the distance formula: In the formula, L x Let L be the distance between points A and B along the east-west direction. y R is the distance between points A and B along the north-south direction, where rad is the unit of radians and its value is 180 / π; R is the Earth's radius and its value is 6371004m. Calculate the spatial distance between the coordinates of the sampling points in the dataset and the actual station coordinates. The sampling point with the smallest distance is identified as the boundary point between data segments between adjacent stations. The objective function is: L j =min[Distance(X i ,X′ j )],i=1,2,...,N.j=1,2,...,M; In the formula: X i X′ represents the latitude and longitude coordinates of the sampling points in the dataset. j The extracted station latitude and longitude coordinates; N is the sample size of the dataset; M is the number of bus stops; If road segments overlap, a penalty-inducing objective function is introduced when calculating the distance: L j =min[Distance(X i ,X j ′)+α|Sx-S1-β|], i=1, 2,..., Nj=8, 9; In the formula: α is the penalty coefficient; S i β represents the cumulative mileage at each sampling point; β is the average driving distance from the starting point to each sampling point. The optimal solution of the objective function is the boundary point between data segments between adjacent stations, which can then divide the dataset into micro-travels between adjacent stations; the divided micro-travels can be further divided into effective micro-travels and undesirable micro-travels. The criteria for poor micro-travel are: continuous gaps or loss of sampling point data within a data segment, and the inability to fill in the missing data.

5. The method for evaluating the ecological driving of bus drivers based on big data according to claim 4, characterized in that, The minimum spatial distance between each sampling point and the bus stop is less than or equal to 65m.

Images