Electric vehicle road section energy consumption distribution and prediction method

Abstract

The invention discloses an electric vehicle road section energy consumption distribution and prediction method. The method comprises the following steps: constructing a vehicle travel information data set; performing similarity classification on the vehicle tracks to obtain a vehicle track similarity classification result, selecting a prediction path according to the vehicle track similarity classification result, and obtaining all data on the prediction path; the prediction path is divided into a plurality of branch road sections, elevation information of each branch road section is extracted, and an energy consumption distribution value of each branch road section is determined and analyzed; based on the energy consumption distribution value, key influence factors of road section energy consumption are explored, a quantifiable feature data set is constructed, an energy consumption prediction model based on a machine learning algorithm is constructed according to the quantifiable feature data set, the precision of the energy consumption prediction model is evaluated, and the energy consumption prediction model with the highest precision is selected for electric vehicle road section energy consumption distribution and prediction. Accurate energy consumption prediction can be provided for an energy management system of the electric vehicle, and a driver is helped to reasonably plan a route and save energy.

Description

Technical Field

[0001] The present invention relates to the fields of machine learning, data mining and intelligent transportation technology, and in particular to a method for allocating and predicting energy consumption of electric vehicle sections. Background Art

[0002] The environmental problems caused by the current energy crisis are becoming increasingly prominent. With the increasing tension of energy resources and the improvement of environmental protection awareness, various industries have put forward higher requirements for energy management and utilization efficiency.

[0003] In the automotive industry, the sales of new energy vehicles are increasing day by day. With the popularization of electric vehicles, energy consumption prediction plays a vital role in the process. Accurate driving energy consumption prediction can solve the problem of "mileage anxiety" for drivers, and is also the basis for reasonably arranging the charging time of pure electric vehicles and achieving the optimal distribution of charging facilities. It is of great significance to promote the development of intelligent transportation systems and promote the further development and popularization of the electric vehicle industry. Therefore, how to predict the energy consumption of electric vehicles with high accuracy is also a new challenge. In addition, how to accurately distribute the energy consumption in each actual road section is also an important part of energy management.

[0004] In the prior art, energy consumption prediction of pure electric vehicles is mainly carried out by establishing physical models and data models for energy consumption prediction. The research on methods based on physical models mainly focuses on the calibration of various parameters in the energy consumption model and the generation of speed models. The physical model can be roughly divided into longitudinal dynamic model and specific power model. A random speed curve is used to offset the difference in energy consumption caused by different driving behaviors, and then a multiple linear regression model is used to predict energy consumption. However, since there are many and relatively complex factors affecting the energy consumption of pure electric vehicles, there are many uncertainties between multiple influencing factors, and it is difficult for physical models to describe the relationship between various influencing factors. In recent years, with the rapid development of big data and machine learning algorithms, data-driven methods are used to predict the energy consumption of electric vehicles. While avoiding the shortcomings of physical models, it also pushes the energy consumption prediction method to a new level. Among them, the more classic machine learning algorithms for pure electric vehicle energy consumption prediction include gradient boosting trees, neural networks, random forests, and support vector machines. The main process is to extract energy consumption influencing characteristic factors from traffic environment, vehicle status, and driving behavior, and then use machine learning algorithms to establish a prediction model to predict the energy consumption of electric vehicles.

[0005] Although the research on energy consumption prediction models has achieved remarkable results, in actual application, no efficient and accurate energy consumption prediction model has been established for specific travel route sections. Summary of the invention

[0006] In order to solve the above technical problems, the present invention proposes a method for allocating and predicting energy consumption of electric vehicles on a road section, which accurately allocates the energy consumption of the road section and then establishes a model for each road section and makes predictions.

[0007] To achieve the above object, the present invention provides a method for allocating and predicting energy consumption of electric vehicles on a road section, comprising:

[0008] Acquire target driving data of several electric vehicles and construct a vehicle travel information data set, wherein the target driving data complies with the GB / T 32960.3 data standard;

[0009] Based on the vehicle travel information data set, vehicle trajectory fitting is performed, vehicle trajectory similarity classification is performed, vehicle trajectory similarity classification results are obtained, and a predicted path is selected according to the vehicle trajectory similarity classification results, and all data on the predicted path are obtained;

[0010] Dividing the predicted path into a number of sub-sections, extracting elevation information of each sub-section, determining energy consumption allocation values ​​of each sub-section and analyzing them;

[0011] Based on the energy consumption allocation value, explore the key influencing factors of road section energy consumption, construct a quantifiable feature data set, construct an energy consumption prediction model based on a machine learning algorithm according to the quantifiable feature data set, and evaluate the accuracy of the energy consumption prediction model;

[0012] Select the energy consumption prediction model with the highest accuracy to allocate and predict energy consumption of electric vehicle sections.

[0013] Preferably, constructing the vehicle travel information data set includes grouping and preprocessing the travel data;

[0014] The grouping process is as follows:

[0015] Data is grouped according to vehicle numbers, that is, the travel trajectory data of all vehicles is separated into units of each vehicle, and the "time stamp daq_time" in all vehicle data is analyzed to check the time interval of data information return of each vehicle, and the preset interval is used as the basis for judging the single travel trajectory of the vehicle to group the data;

[0016] The pre-treatment process includes:

[0017] Analyze the abnormal data in the grouped data, remove the abnormal data or fill it with data using a data filling method, and construct the vehicle travel information data set that can be used for research.

[0018] Preferably, performing the vehicle trajectory fitting includes:

[0019] The vehicle trajectory data in the vehicle travel information data set is extracted, and the vehicle trajectory is fitted according to the vehicle GPS trajectory point data in the vehicle trajectory data, that is, the vehicle trajectory is fitted in the map, and the vehicle travel pattern and the distribution characteristics of the vehicle starting and ending points, departure time and driving route characteristics are mined through the visualization of the vehicle trajectory.

[0020] Preferably, obtaining the vehicle trajectory similarity classification result includes:

[0021] The DTW algorithm is used to classify vehicle trajectories by similarity, specifically:

[0022] Construct a distance matrix: First, calculate the distances of all pairs of points in the two time series to form a matrix;

[0023] Finding the best path: Starting from the upper left corner of the matrix and ending at the lower right corner, a path is found through dynamic programming so that the sum of the distances between the points on the path is minimized. The path represents the best alignment between the two sequences.

[0024] Calculate the total distance: the sum of the distances of the point pairs on the path is the DTW distance between the two sequences, and the DTW distance is the similarity;

[0025] The vehicle trajectory similarities are classified according to the similarities.

[0026] Preferably, extracting the elevation information of each road section includes:

[0027] Based on the longitude and latitude range of the predicted path, the predicted path is divided into several sub-sections according to the longitude and latitude, the longitude and latitude range of each sub-section is recorded, and the corresponding section is found according to the longitude and latitude range of each section to extract the section elevation information, the section elevation information includes the slope of the section and the plane length of the section.

[0028] Preferably, the method for determining the energy consumption allocation value of each sub-section includes:

[0029] Use SUMO simulation to obtain the influence of slope on vehicle energy consumption. Use the vehicle energy consumption value under the slope of the real road section to simulate and obtain the distance traveled on each sub-section slope when the same energy consumption is equivalent to the distance traveled on the flat section to obtain the conversion relationship. Record the conversion relationship of each section, convert the driving distance in the actual driving data according to the conversion relationship, obtain the flattened distance considering the slope, and reasonably distribute the energy consumption value in the sub-sections of the specific route.

[0030] Alternatively, based on the impact of the vehicle's driving speed, vehicle gravity, frictional resistance, and air resistance on vehicle energy consumption, the energy value consumed by the vehicle trajectory in each sub-section is calculated, and then the total energy consumption of the trajectory is allocated according to the proportion of the energy consumed in each sub-section to obtain the energy consumption allocation value on each sub-section.

[0031] Preferably, analyzing the energy consumption distribution value of each sub-section includes:

[0032] According to the distribution interval of the data longitude and latitude points, the sub-sections passed by the trajectory are determined and the distance passed through each sub-section is calculated. The vehicle's driving distance in each sub-section is converted according to the conversion relationship of each section to obtain the planarized distance of each sub-section considering the slope. The total energy consumption is divided by the total planarized distance passed through the sub-section to obtain the average energy consumption rate. Then, the energy consumption is allocated according to the planarized distance traveled in each sub-section, that is, the average energy consumption rate is multiplied by the planarized distance of each sub-section to obtain the energy consumption allocation value in each sub-section.

[0033] Preferably, analyzing the energy consumption distribution value of each sub-section further includes:

[0034] Given a preset average speed value, the relationship between slope and power is obtained. Without considering acceleration, the power ratio between each sub-section is estimated. Combined with the time interval of the car passing through the straight section, the energy value ratio between each sub-section is obtained, and the total energy consumption is distributed to obtain the energy consumption distribution value of each sub-section.

[0035] If the car has obvious acceleration and deceleration stages on the road section, the road length is set to L, and the vehicle accelerates from the initial speed V with an average acceleration a. 0 Accelerate to average speed V 1 , with an average speed V 1 After driving the remaining distance, considering the slope of the road section, calculate the power W in the acceleration stage respectively. 1 and the power W during the driving phase at average speed 2 , the power W 1 , Power W 2 The total energy W consumed by the vehicle is obtained by adding up the values ​​of W calculated under different conditions on different road sections, and the proportion is calculated to distribute the total energy consumption and determine the energy consumption distribution value of each sub-section. Among them, the power W in the acceleration stage is 1 Obtained by calculating the integral of the power over a given time.

[0036] Preferably, constructing the quantifiable feature data set includes:

[0037] Calculate the average energy consumption rate per kilometer in the entire journey, distribute energy consumption according to the planarized distance of each sub-section, determine the specific energy consumption value of each trajectory segment of each vehicle in each sub-section, divide the specific energy consumption value of each trajectory in each sub-section by the planarized distance of the driving in the section to obtain the energy consumption rate per kilometer and record it as the target variable to complete the construction of a quantifiable characteristic energy consumption data set;

[0038] Alternatively, the calculated energy consumption distribution value of each road section is divided by the distance traveled by the vehicle in each road section to obtain the energy consumption rate per kilometer and record it as the target variable to complete the construction of a quantifiable characteristic energy consumption data set.

[0039] Preferably, constructing the energy consumption prediction model based on the machine learning algorithm includes:

[0040] Selecting feature variables and target variables from the quantifiable feature data set, using KNN algorithm, neural network, support vector machine, and random forest algorithm to establish energy consumption prediction models respectively, and evaluating the accuracy of energy consumption prediction models under different algorithms through cross-validation method, calculating the mean absolute percentage error value MAPE, and selecting the algorithm model with the lowest MAPE value, that is, the energy consumption prediction model based on machine learning algorithm;

[0041] The calculation formula of MAPE is:

[0042] MAPE=(1 / n)*Σ(|(predicted_value-actual_value)| / actual_value)*100%

[0043] In the formula, n is the number of samples, predicted_value is the predicted value of the model, and actual_value is the actual value of the model.

[0044] Compared with the prior art, the present invention has the following advantages and technical effects:

[0045] The existing prediction methods only consider the energy consumption influencing factors of traffic environment, vehicle status, driving behavior, etc. to establish the model. The present invention is a segmented energy consumption prediction model established based on actual driving data under real road sections. While considering the existing influencing factors, it also considers the differences between different sections in the real road sections (mainly the influence of the differences in the slope of the sections). By providing two different schemes to determine the energy consumption distribution of vehicles under the influence of different section slopes, an accurate energy consumption prediction model is established to predict energy consumption. According to the driving data, the actual section is accurately allocated for energy consumption. When planning multiple travel route plans, the energy consumption prediction model can be built and the energy consumption can be predicted with high precision for the real section in any travel route according to this method. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0047] Figure 1 A flow chart of a method for allocating and predicting energy consumption of an electric vehicle road section according to an embodiment of the present invention;

[0048] Figure 2 This is a flow chart of trajectory similarity judgment according to an embodiment of the present invention;

[0049] Figure 3 This is a flow chart of obtaining the conversion relationship α of each road section by using SUMO simulation in an embodiment of the present invention;

[0050] Figure 4 A schematic diagram of energy consumption distribution of a road section according to an embodiment of the present invention;

[0051] Figure 5 This is a flow chart of energy consumption prediction for each road section according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0053] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0054] The present invention proposes a method for allocating and predicting energy consumption of electric vehicles on a road section. Figure 1 ,include:

[0055] Obtain target driving data of several electric vehicles and construct a vehicle travel information dataset. The target driving data complies with the GB / T 32960.3 data standard.

[0056] Based on the vehicle travel information dataset, vehicle trajectory fitting is performed, vehicle trajectory similarity classification is performed, and vehicle trajectory similarity classification results are obtained. The predicted path is selected based on the vehicle trajectory similarity classification results, and all data on the predicted path are obtained.

[0057] Divide the predicted path into several sub-sections, extract the elevation information of each sub-section, determine the energy consumption allocation value of each sub-section and analyze it;

[0058] Based on the energy consumption distribution value, the key influencing factors of road section energy consumption are explored, a quantifiable feature data set is constructed, an energy consumption prediction model based on a machine learning algorithm is constructed based on the quantifiable feature data set, and the accuracy of the energy consumption prediction model is evaluated;

[0059] Select the energy consumption prediction model with the highest accuracy to allocate and predict energy consumption of electric vehicle sections.

[0060] This embodiment provides a method for allocating energy according to each section of a specific travel route, and then establishing an energy consumption prediction model to predict energy consumption, which can better carry out travel planning and energy management. The DTW algorithm is used to classify similar trajectories for the actual driving data trajectory, and the section trajectory for energy consumption prediction is selected. For the travel trajectory of this section, the entire trip is divided into several sub-sections, and the slope of each sub-section is extracted. It is proposed to determine the influence of the slope of the section on energy consumption through traffic simulation or energy calculation, so as to accurately allocate the energy consumption of electric vehicles to each sub-section, and then a high-precision energy consumption prediction model based on a machine learning algorithm can be established for the actual section and evaluated. According to this method, accurate energy consumption prediction can be made for any travel route.

[0061] In order to enable electric vehicle energy managers and travelers to obtain accurate energy consumption predictions based on travel routes, this embodiment selects route trajectory data required for energy consumption prediction from the actual driving data of multiple pure electric vehicles, and through data grouping, data preprocessing, trajectory fitting, trajectory classification, energy consumption allocation, construction of quantifiable feature data sets, building energy consumption prediction models, model verification and evaluation and other research work, the energy consumption of road sections is accurately allocated, and then a model is established and predicted for each section.

[0062] Further, constructing a vehicle travel information dataset, including grouping and preprocessing the travel data;

[0063] The grouping process is as follows:

[0064] Data is grouped according to vehicle numbers, that is, the travel trajectory data of all vehicles is separated into units of each vehicle, and the "time stamp daq_time" in all vehicle data is analyzed to check the time interval of data information return of each vehicle, and the preset interval is used as the basis for judging the single travel trajectory of the vehicle to group the data;

[0065] The preprocessing process includes:

[0066] Analyze the abnormal data in the grouped data, remove the abnormal data or fill it with data using a data filling method, and construct the vehicle travel information data set that can be used for research.

[0067] Specifically, the data used in this embodiment conforms to the GB / T 32960.3 data standard, specifically including "id, daq_time, soc, lng, lat, max_temp, min_temp" in the GB / T32960.3 data standard.

[0068] A large amount of pure electric vehicle driving data information is sorted out, the data structure and the information content are analyzed, and the information feedback interval of each vehicle is determined according to the timestamp information. Generally, the vehicle information feedback time interval is between 5 and 30 seconds. If there is a large time interval, it is considered to be a segmentation of the vehicle's running journey. Therefore, all vehicle travel data is segmented at a time interval of 5 minutes to study the vehicle's single trip. After segmentation, check the data to verify the segmentation effect (whether the vehicle's single trip is separated according to the set time interval). Filter segmentation can automatically process data in Excel by executing a python program (this method can also be used for subsequent data processing).

[0069] The segmented vehicle trajectory data is preprocessed as follows:

[0070] The segmented data is fitted with trajectory visualization. The main operation is to use Amap, Baidu Map, Google Map, etc. as the base map, and then mark the GPS latitude and longitude points of the vehicle trajectory data on the base map and mark the starting and ending points, and then connect all the points with lines to obtain the vehicle trajectory. After visualizing the trajectory, it will be found that there are abnormal situations, such as the loss of the starting and ending points of some trajectories, and the presence of abnormal points in the vehicle trajectory. This is because the data has not been cleaned, so the data must be preprocessed, and the vehicle data must be analyzed to find out the possible reasons for the abnormal situation. In this embodiment, several possible reasons for the abnormality are given: the speed of some vehicle data is 0 or the speed is too high, and there are abnormal points with latitude and longitude of 0. Therefore, it is necessary to remove these abnormal point data. Here, speed, latitude and longitude can be used as limiting conditions to remove abnormal points according to the abnormal situation of the data. After preprocessing all vehicle travel data, refit to check the fitting effect. After the abnormal situation is eliminated, the trajectory file after data preprocessing is used as the vehicle travel information data set for research.

[0071] Furthermore, vehicle trajectory fitting is performed, including:

[0072] The vehicle trajectory data in the vehicle travel information dataset is extracted, and the vehicle trajectory is fitted according to the vehicle GPS trajectory point data in the vehicle trajectory data, that is, the vehicle trajectory is fitted in the map, and the vehicle travel rules and the distribution characteristics of the vehicle's starting and ending points, departure time and driving route characteristics are mined through the visualization of the vehicle trajectory.

[0073] Specifically, based on the constructed vehicle travel information dataset, vehicle trajectory fitting is performed according to the vehicle GPS trajectory point data (lng, lat) in each segment of vehicle trajectory data, that is, the vehicle trajectory is fitted in the map (generally Amap, Baidu Map, OSM Map can be used), and the travel patterns of vehicles, the distribution characteristics of vehicle starting and ending points, departure time and driving route characteristics are mined through the visualization of vehicle trajectories.

[0074] This example uses the OSM map as the base map, and connects all the vehicle trajectory points on the base map using the folium.PolyLine method to obtain a single trip trajectory map of each vehicle; and marks the OD points of the trajectory on the base map using the folium.Marker method. All vehicle trajectories are fitted and generated into an html file for viewing (which can be done by executing the py program), and the travel rules of vehicles, the distribution characteristics of the vehicle's starting and ending points, and the characteristics of the driving route are mined through the visualization of vehicle trajectories.

[0075] Furthermore, the vehicle trajectory similarity classification results are obtained, including:

[0076] The DTW algorithm is used to classify vehicle trajectories by similarity, specifically:

[0077] Construct a distance matrix: First, calculate the distances of all pairs of points in the two time series to form a matrix;

[0078] Finding the best path: Starting from the upper left corner of the matrix and ending at the lower right corner, a path is found through dynamic programming so that the sum of the distances between the points on the path is minimized. The path represents the best alignment between the two sequences.

[0079] Calculate the total distance: the sum of the distances of the point pairs on the path is the DTW distance between the two sequences, and the DTW distance is the similarity;

[0080] The vehicle trajectory similarities are classified according to the similarities.

[0081] Specifically, the travel trajectories of all vehicles are similarly classified in order to find the main predicted path trajectory that needs to be predicted. In this embodiment, the DTW algorithm is used to classify the vehicle trajectories by similarity. The classification process of the DTW algorithm is shown in the attached Figure 2 The main idea of ​​trajectory classification is to find several trajectories with high fitting accuracy as standard trajectories in all generated vehicle visualization trajectory files, and then classify all other trajectories according to the DTW algorithm. (Classification can be performed through Python program) After classification, find all the data on the main path that needs to be predicted in the trajectory.

[0082] Furthermore, the elevation information of each road section is extracted, including:

[0083] Based on the longitude and latitude range of the predicted path, the predicted path is divided into several sub-sections according to the longitude and latitude, the longitude and latitude range of each sub-section is recorded, and the corresponding section is found according to the longitude and latitude range of each section to extract the section elevation information, the section elevation information includes the slope of the section and the plane length of the section.

[0084] Specifically, the main predicted path is divided into several sub-sections according to longitude and latitude. The main path can be divided into several sub-sections according to the needs of travel route energy consumption prediction. Baidu Maps, Amap, and OSM Maps can be used to divide the sections and determine the longitude and latitude ranges, record the longitude and latitude range of each sub-section, find the corresponding section in Google Earth according to the longitude and latitude range of each section, and extract the elevation information of the section (each point on the road has a corresponding elevation, and the slope of the road is obtained by the elevation difference), extract the slope θ of the section and the plane length of the section and record them. Because the actual energy consumption data is driven on a real road section, there will be a slope impact. The slopes of different sections affect the energy consumption allocation value of the vehicle in each sub-section, and this value generally cannot be directly obtained from the data (the minimum unit of the energy consumption value in the data is 1). Therefore, two schemes are provided in this embodiment to determine the energy consumption allocation value of each sub-section, so that an accurate energy consumption prediction model can be established in each sub-section later.

[0085] Furthermore, the method for determining the energy consumption allocation value of each sub-section includes:

[0086] Use SUMO simulation to obtain the influence of slope on vehicle energy consumption. Use the vehicle energy consumption value under the slope of the real road section to simulate and obtain the distance traveled on each sub-section slope when the same energy consumption is equivalent to the distance traveled on the flat section to obtain the conversion relationship. Record the conversion relationship of each section, convert the driving distance in the actual driving data according to the conversion relationship, obtain the flattened distance considering the slope, and reasonably distribute the energy consumption value in the sub-sections of the specific route.

[0087] Alternatively, based on the impact of the vehicle's driving speed, vehicle gravity, frictional resistance, and air resistance on vehicle energy consumption, the energy value consumed by the vehicle trajectory in each sub-section is calculated, and then the total energy consumption of the trajectory is allocated according to the proportion of the energy consumed in each sub-section to obtain the energy consumption allocation value on each sub-section.

[0088] Specifically, Scheme 1: Use SUMO simulation to obtain the impact of slope on vehicle energy consumption, and obtain the energy consumption of each sub-section under the slope equivalent to the distance conversion coefficient α of flat driving. Figure 3 . Use the NetEdit software in the SUMO simulation to select one of the sub-sections, draw a simple straight line section, and then calculate the height and length information L 0Determine the OD point and save the road network file. Then write the traffic flow rou file (only one electric car needs to be set) and sumocfg file for simulation, and determine the vehicle energy consumption through the simulation result Battery.out.xml file. Then adjust the road network file, change the road section to a flat section (that is, the OD point elevation is consistent), and adjust the section length L 1 The energy consumption is made consistent with the previous simulation. At this time, the driving distance equivalent to the flat road section under the consideration of the road section slope is obtained, and the distance conversion coefficient of each road section is calculated as α = L 1 / L 0 , perform this operation on all road sections that need to be predicted and record the conversion coefficient α of each road section, so that the energy consumption value can be reasonably allocated to the road sections of the specific route later.

[0089] Solution 2: Taking into account the impact of vehicle speed, vehicle gravity, friction resistance and air resistance on vehicle energy consumption under different road slopes, the invention provides a method for calculating the energy value consumed by the vehicle trajectory in each sub-section. The power P of the vehicle trajectory in each sub-section is calculated, and then the energy value consumed by each sub-section is calculated using the formula W=Pt (W is energy, t is the travel time interval of the sub-section). According to the proportion of the energy value consumed by each sub-section, the total energy consumption of the trajectory is distributed, thereby obtaining the energy consumption distribution value on each sub-section.

[0090] Furthermore, the energy consumption distribution value of each sub-section is analyzed, including:

[0091] According to the distribution interval of the data longitude and latitude points, the sub-sections passed by the trajectory are determined and the distance passed through each sub-section is calculated. The vehicle's driving distance in each sub-section is converted according to the conversion relationship of each section to obtain the planarized distance of each sub-section considering the slope. The total energy consumption is divided by the total planarized distance passed through the sub-section to obtain the average energy consumption rate. Then, the energy consumption is allocated according to the planarized distance traveled in each sub-section, that is, the average energy consumption rate is multiplied by the planarized distance of each sub-section to obtain the energy consumption allocation value in each sub-section.

[0092] Specifically, since the electric vehicle power consumption in the data is in units of 1, and the energy consumption of 1 may be divided into more than one sub-section, it is obviously not enough to find the specific energy consumption of any section in the path. Therefore, it is necessary to separate the trajectory with a complete energy consumption of 1 in each vehicle trajectory, and separate the data segment with a complete energy consumption value of 1 from the travel information data set to obtain a new data file (subsequent prediction work is carried out based on the separated data file). In response to the two solutions mentioned above, this embodiment provides two solutions to determine the energy consumption distribution value of the sub-section:

[0093] Solution 3: Determine which sub-sections the trajectory has passed through based on the distribution interval of the data longitude and latitude points and calculate the distance through each sub-section (the calculated distance can be calculated based on the longitude and latitude of the trajectory OD point. The vehicle's travel distance in each sub-section is converted according to the conversion coefficient α of each sub-section in the above solution 1 to obtain the planarization distance that the vehicle actually travels in each sub-section considering the slope. The total energy consumption 1 is divided by the total planarization distance through the sub-section to obtain the average energy consumption rate. Then, the energy consumption is distributed according to the planarization distance traveled in each sub-section, that is, the average energy consumption rate is multiplied by the planarization distance of each sub-section to obtain the specific energy consumption value in each sub-section and record it. Refer to the attached Figure 4 , which is used to subsequently establish a quantifiable feature dataset.

[0094] Solution 4: Determine which sub-segments the track has passed through based on the distribution interval of the data longitude and latitude points and calculate the distance through each sub-segment (the distance can be calculated by calculating the great circle distance based on the longitude and latitude of the track OD point) and record it. Considering the impact of the vehicle's speed, vehicle gravity, friction resistance and air resistance on vehicle energy consumption under different road slopes, as well as some energy losses, the power of the vehicle on each sub-segment can be calculated according to the following formula:

[0095]

[0096] In the formula, the upslope θ is positive, the downslope θ is negative, r is the conductor resistance, R t is the effective electric vehicle tire radius, K is the product of the armature constant and the magnetic flux, m is the vehicle mass, a is the acceleration, k is the aerodynamic drag coefficient, v is the vehicle speed, and f r is the rolling resistance coefficient, g is the acceleration of gravity, and θ is the road slope. The first half of the formula represents the fixed energy loss. For the same vehicle, r, R t , K remain unchanged (generally: r = 0.11Ω, R t =0.5m, K=10.08Vs); the second half considers the influence of the force F.

[0097] Among them, the vehicle mass m is given according to the vehicle data, the variable road slope θ has been extracted in the above steps, the speed v can be extracted from the trajectory data to calculate the average speed, and the acceleration a can be calculated according to the formula V 2 =V 1 +a 12 *t 12 (t 12 is the time interval between two adjacent data points of the trajectory, V 1 、V 2The acceleration between every two adjacent data points is calculated, and the average value is added to get the average acceleration a; other parameters are generally taken as: k = 1.3kg / m 3 、f r =0.006, g = 9.81 m / s 2 .

[0098] After obtaining the vehicle power P value of each sub-section, extract the OD point time interval t of each sub-section. Assuming that the sub-section L 1 and the subsection L 2 The power values ​​are P 1 and P 2 , use the formula W = Pt (W is energy, t is the time interval between the OD points of the sub-section) to get the ratio of the energy value of each sub-section, that is, W1 / W2 = P 1 *t 1 / P 2 *t 2 , based on which the total energy consumption 1 is distributed in proportion to obtain the energy consumption distribution value of each sub-section.

[0099] Furthermore, the energy consumption distribution value of each sub-section is analyzed, and the following is also included:

[0100] Given a preset average speed value, the relationship between slope and power is obtained. Without considering acceleration, the power ratio between each sub-section is estimated. Combined with the time interval of the car passing through the straight section, the energy value ratio between each sub-section is obtained, and the total energy consumption is distributed to obtain the energy consumption distribution value of each sub-section.

[0101] If the car has obvious acceleration and deceleration stages on the road section, the road length is set to L, and the vehicle accelerates from the initial speed V with an average acceleration a. 0 Accelerate to average speed V 1 , with an average speed V 1 After driving the remaining distance, considering the slope of the road section, calculate the power W in the acceleration stage respectively. 1 and the power W during the driving phase at average speed 2 , the power W 1 , Power W 2 The total energy W consumed by the vehicle is obtained by adding up the values ​​of W calculated under different conditions on different road sections, and the proportion is calculated to distribute the total energy consumption and determine the energy consumption distribution value of each sub-section. Among them, the power W in the acceleration stage is 1 Obtained by calculating the integral of the power over a given time.

[0102] Specifically, in order to simplify the frequent power calculation process, this embodiment provides two cases according to the above solution 4:

[0103] Case 1: Given a preset average speed value, the relationship between the slope θ and the power is obtained. (Without considering the acceleration), the power ratio between each sub-section can be approximately estimated. Combined with the time interval t between the OD points of the sub-sections, the ratio of the energy value W between each sub-section can be obtained, thereby allocating the total energy consumption 1 to obtain the energy consumption distribution value of each sub-section.

[0104] Case 2: If there are obvious acceleration and deceleration stages on the road section, that is, acceleration a≠0, the road length is set to L, and the vehicle accelerates from an initial speed V with an average acceleration a. 0 Accelerate to an average speed value V 1 , and then use this average speed value V 1 Drive the remaining distance. In this process, considering the influence of the road slope θ, calculate the acceleration stage W 1 (If there is a significant deceleration phase, the calculation method is the same) and the power W of the average speed driving phase 2 The sum of the two is the total energy W consumed by the vehicle. According to the W values ​​calculated under different conditions on different road sections, the proportion is used to allocate the total energy consumption 1 to determine the energy consumption allocation value for each sub-section.

[0105] The work W in the acceleration phase 1 , can be obtained by calculating the integral of the power over a given time, as follows:

[0106] 1. Calculate the acceleration time: First, based on the given acceleration a and velocity change (from the initial velocity V 0 To the final average speed V 1 ), calculate the time required for the vehicle to accelerate. This is calculated using the following formula:

[0107]

[0108] t 0 The time taken for the acceleration phase.

[0109] 2. Speed ​​as a function of time, the formula is: V 1 (t) = V 0 +a*t 0

[0110] 3. Calculate power: Calculate according to the formula for calculating power P given in Scheme 4. Power P is calculated based on the current instantaneous speed V(t) and angle θ.

[0111] 4. Calculate work: Use the numerical integration method, the formula is:

[0112] The work W during the average speed phase (which can be regarded as a constant speed) 2 , the calculation steps are as follows:

[0113] 1. Calculate the uniform speed driving time:

[0114]

[0115] Where L is the total length of the road section, t 1 is the uniform speed driving time.

[0116] 2. Calculate the power in the uniform speed stage: 1 Substitute it into the power P calculation formula (at this time a=0), the power P calculation formula is:

[0117]

[0118] 3. Calculate the work W 2 , the formula is:

[0119] Finally, the total work W = W 1 +W 2 , according to different road section lengths L and slopes θ, calculate the W values ​​in different situations, calculate the W value ratio and distribute the total energy consumption 1 to obtain the energy consumption distribution value of each sub-section.

[0120] At this point, the electric vehicle road section energy consumption allocation method based on the GB / T 32960.3 data standard in this embodiment has been completed, and the subsequent energy consumption prediction step can be performed according to the energy consumption allocation result.

[0121] Furthermore, a quantifiable feature dataset is constructed, including:

[0122] In this embodiment, multiple factors such as vehicle speed, travel distance, temperature, travel time, etc. can be selected as characteristic variables, and the average energy consumption rate per kilometer in the entire journey is calculated. The energy consumption is allocated according to the planarized distance of each sub-section passed, and the specific energy consumption value of each track segment of each vehicle in each sub-section is determined. The specific energy consumption value of each track in each sub-section is divided by the planarized distance of the travel in the section to obtain the energy consumption rate per kilometer and record it as the target variable, thereby completing the construction of a quantifiable characteristic energy consumption data set;

[0123] Alternatively, the calculated energy consumption distribution value of each road section is divided by the distance traveled by the vehicle in each road section to obtain the energy consumption rate per kilometer and record it as the target variable to complete the construction of a quantifiable characteristic energy consumption data set.

[0124] Specifically, the present embodiment includes influencing factors such as the age of the vehicle, the type of vehicle battery, travel time, temperature, vehicle speed, and road section characteristics. By studying several segments of trajectory data with a complete energy consumption of 1, we find out the specific road sections that each segment of the trajectory passes through, determine the route of each segment of the trajectory, and then extract temperature (calculate the average temperature of the segment trajectory), travel time (time difference of OD points), calculate the total distance traveled (calculate the great circle distance based on the longitude and latitude of the trajectory OD points), average speed and other characteristic factors from the trajectory segment data and record them in Excel to establish a quantifiable feature data set, which will be subsequently input as a model feature variable. The quantifiable feature data set should also include the energy consumption rate per kilometer of each vehicle trajectory on each segment as the goal of subsequent model establishment, refer to the attached Figure 5 .

[0125] For the calculation of energy consumption rate per kilometer, two methods are provided here according to the two solutions proposed in the above technical solution:

[0126] Method 1: Calculate the average energy consumption rate per kilometer in the entire journey (1 / total planarization distance of the section) (the total planarization distance of the section here is the sum of the distances after all the sub-sections are planarized after the simulation of Scheme 1), allocate energy consumption according to the planarization distance of each sub-section, determine the specific energy consumption value of each trajectory segment of each vehicle in each sub-section (mentioned in Scheme 3), divide the specific energy consumption value of each trajectory in each sub-section by the planarization distance of the driving in the section to obtain the energy consumption rate per kilometer and record it as the target variable for the subsequent model establishment, and the quantifiable characteristic energy consumption data set is constructed (including trajectory vehicle information, selected characteristic variables, and target variables).

[0127] Method 2: Divide the energy consumption allocation value of each sub-section calculated in Scheme 4 by the distance traveled in each sub-section in the vehicle trajectory to obtain the energy consumption rate per kilometer and record it as the target variable for subsequent model establishment. The quantifiable characteristic energy consumption data set is constructed (including trajectory vehicle information, selected characteristic variables, and target variables).

[0128] According to the actual situation, two schemes can be used to construct two different sets of quantifiable feature data sets. When the model is subsequently established, the two sets of data sets can be compared to select the best one.

[0129] Furthermore, an energy consumption prediction model based on machine learning algorithm is constructed, including:

[0130] Selecting feature variables and target variables from a quantifiable feature data set, using KNN algorithm, neural network, support vector machine, and random forest algorithm to establish energy consumption prediction models respectively, and evaluating the accuracy of energy consumption prediction models under different algorithms through a cross-validation method, calculating the mean absolute percentage error value MAPE, and selecting the algorithm model with the lowest MAPE value, that is, the energy consumption prediction model based on the machine learning algorithm;

[0131] The characteristic variables include travel time, temperature, and average vehicle speed, and the target variable is the vehicle energy consumption rate per kilometer.

[0132] The calculation formula of MAPE is:

[0133] MAPE=(1 / n)*Σ(|(predicted_value-actual_value)| / actual_value)*100%

[0134] In the formula, n is the number of samples, predicted_value is the predicted value of the model, and actual_value is the actual value of the model.

[0135] Specifically, based on the established quantifiable characteristic energy consumption data set, travel time, temperature, and average speed are used as characteristic variable inputs, and the vehicle energy consumption rate per kilometer (energy consumption of each sub-section / planarization distance of the sub-section) is used as the target variable. KNN algorithm, neural network, support vector machine, random forest and other algorithms are used to establish energy consumption prediction models respectively (in this embodiment, 80% of the data set is used as a training set for model training, and 20% is used as a test set to verify the model accuracy). The accuracy of the energy consumption prediction model under different algorithms can be evaluated under 5-fold to 10-fold cross-validation.

[0136] Finally, the algorithm model with the lowest MAPE value (highest accuracy) is selected for energy consumption prediction. The model building method in the present invention can be used to build models and perform energy consumption prediction for each real road section to better select and plan travel routes.

[0137] The present invention is a road section energy consumption prediction model established based on actual driving data on a real road section. While considering existing influencing factors, it also considers the differences between different sections in the real road section (mainly the influence of differences in road section slopes). The invention provides two different schemes to determine the vehicle energy consumption distribution under the influence of different road section slopes, thereby establishing an accurate energy consumption prediction model for energy consumption prediction. According to the driving data, accurate energy consumption is allocated to the actual road section. When planning multiple travel route plans, the energy consumption prediction model can be built and high-precision energy consumption prediction can be performed on the real road section in any travel route according to this method.

[0138] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for allocating and predicting energy consumption of electric vehicles on a road section, characterized in that: include: Acquire target driving data of several electric vehicles and construct a vehicle travel information data set, wherein the target driving data complies with the GB / T 32960.3 data standard; Based on the vehicle travel information data set, vehicle trajectory fitting is performed, vehicle trajectory similarity classification is performed, vehicle trajectory similarity classification results are obtained, and a predicted path is selected according to the vehicle trajectory similarity classification results, and all data on the predicted path are obtained; Dividing the predicted path into a number of sub-sections, extracting elevation information of each sub-section, determining and analyzing energy consumption allocation values ​​of each sub-section; Based on the energy consumption allocation value, explore the key influencing factors of road section energy consumption, construct a quantifiable feature data set, construct an energy consumption prediction model based on a machine learning algorithm according to the quantifiable feature data set, and evaluate the accuracy of the energy consumption prediction model; Select the energy consumption prediction model with the highest accuracy to allocate and predict energy consumption of electric vehicle sections.

2. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 1, characterized in that: Constructing the vehicle travel information data set, including grouping and preprocessing the travel data; The grouping process is as follows: Data is grouped according to vehicle numbers, that is, the travel trajectory data of all vehicles is separated into units of each vehicle, and the "time stamp daq_time" in all vehicle data is analyzed to check the time interval of data information return of each vehicle, and the data is grouped based on the preset interval as the basis for judging the vehicle's single travel trajectory; The pre-treatment process includes: Analyze the abnormal data in the grouped data, remove the abnormal data or fill it with data using a data filling method, and construct the vehicle travel information data set that can be used for research.

3. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 1, characterized in that: Performing the vehicle trajectory fitting includes: The vehicle trajectory data in the vehicle travel information data set is extracted, and the vehicle trajectory is fitted according to the vehicle GPS trajectory point data in the vehicle trajectory data, that is, the vehicle trajectory is fitted in the map, and the vehicle travel pattern and the distribution characteristics of the vehicle starting and ending points, departure time and driving route characteristics are mined through the visualization of the vehicle trajectory.

4. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 3 is characterized in that: Obtaining the vehicle trajectory similarity classification result, including: The DTW algorithm is used to classify vehicle trajectories by similarity, specifically: Construct a distance matrix: First, calculate the distances of all pairs of points in the two time series to form a matrix; Finding the best path: Starting from the upper left corner of the matrix and ending at the lower right corner, a path is found through dynamic programming so that the sum of the distances between the points on the path is minimized. The path represents the best alignment between the two sequences. Calculate the total distance: the sum of the distances of the point pairs on the path is the DTW distance between the two sequences, and the DTW distance is the similarity; The vehicle trajectory similarities are classified according to the similarities.

5. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 1, characterized in that: Extract elevation information for each road segment, including: Based on the longitude and latitude range of the predicted path, the predicted path is divided into several sub-sections according to the longitude and latitude, the longitude and latitude range of each sub-section is recorded, and the corresponding section is found according to the longitude and latitude range of each section to extract the section elevation information, the section elevation information includes the slope of the section and the plane length of the section.

6. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 5, characterized in that: The method for determining the energy consumption allocation value of each sub-section includes: Use SUMO simulation to obtain the influence of slope on vehicle energy consumption. Use the vehicle energy consumption value under the slope of the real road section to simulate and obtain the distance traveled on each sub-section slope when the same energy consumption is equivalent to the distance traveled on the flat section to obtain the conversion relationship. Record the conversion relationship of each section, convert the driving distance in the actual driving data according to the conversion relationship, obtain the flattened distance considering the slope, and reasonably distribute the energy consumption value in the sub-sections of the specific route. Alternatively, based on the impact of the vehicle's driving speed, vehicle gravity, frictional resistance, and air resistance on vehicle energy consumption, the energy value consumed by the vehicle trajectory in each sub-section is calculated, and then the total energy consumption of the trajectory is allocated according to the proportion of the energy consumed in each sub-section to obtain the energy consumption allocation value on each sub-section.

7. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 6, characterized in that: Analyzing the energy consumption distribution value of each sub-section, including: According to the distribution interval of the data longitude and latitude points, the sub-sections passed by the trajectory are determined and the distance passed through each sub-section is calculated. The vehicle's driving distance in each sub-section is converted according to the conversion relationship of each section to obtain the planarized distance of each sub-section considering the slope. The total energy consumption is divided by the total planarized distance passed through the sub-section to obtain the average energy consumption rate. Then, the energy consumption is allocated according to the planarized distance traveled in each sub-section, that is, the average energy consumption rate is multiplied by the planarized distance of each sub-section to obtain the energy consumption allocation value in each sub-section.

8. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 6, characterized in that: Analyzing the energy consumption distribution value of each sub-section also includes: Given a preset average speed value, the relationship between slope and power is obtained. Without considering acceleration, the power ratio between each sub-section is estimated. Combined with the time interval of the car passing through the straight section, the energy value ratio between each sub-section is obtained, and the total energy consumption is distributed to obtain the energy consumption distribution value of each sub-section. If the car has obvious acceleration and deceleration stages while driving on the road section, the road length is set to L, and the vehicle accelerates from the initial speed V0 to the average speed V1 at the average acceleration a, and travels the remaining distance at the average speed V1. At this time, considering the influence of the road section slope, the power W1 of the acceleration stage and the power W2 of the stage at the average speed are calculated respectively, and the power W1 and the power W2 are added to obtain the total energy W consumed by the vehicle. According to the W values ​​calculated under different conditions on different road sections, the proportion is calculated to distribute the total energy consumption, and the energy consumption distribution value of each sub-section is determined, among which the power W1 of the acceleration stage is obtained by calculating the integral of the power within a given time.

9. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 1, characterized in that: Constructing the quantifiable feature data set includes: Calculate the average energy consumption rate per kilometer in the entire journey, distribute energy consumption according to the planarized distance of each sub-section, determine the specific energy consumption value of each trajectory segment of each vehicle in each sub-section, divide the specific energy consumption value of each trajectory in each sub-section by the planarized distance of the driving in the section to obtain the energy consumption rate per kilometer and record it as the target variable to complete the construction of a quantifiable characteristic energy consumption data set; Alternatively, the calculated energy consumption distribution value of each road section is divided by the distance traveled by the vehicle in each road section to obtain the energy consumption rate per kilometer and record it as the target variable to complete the construction of a quantifiable characteristic energy consumption data set.

10. The method for allocating and predicting energy consumption of electric vehicles on a road section according to claim 9, characterized in that: Constructing the energy consumption prediction model based on the machine learning algorithm includes: Selecting feature variables and target variables from the quantifiable feature data set, using KNN algorithm, neural network, support vector machine, and random forest algorithm to establish energy consumption prediction models respectively, and evaluating the accuracy of energy consumption prediction models under different algorithms through cross-validation method, calculating the mean absolute percentage error value MAPE, and selecting the algorithm model with the lowest MAPE value, that is, the energy consumption prediction model based on machine learning algorithm; The calculation formula of MAPE is: MAPE=(1 / n)*Σ(|(predicted_value-actual_value)| / actual_value)*100% In the formula, n is the number of samples, predicted_value is the predicted value of the model, and actual_value is the actual value of the model.

Images