CN108198425A - A kind of construction method of Electric Vehicles Driving Cycle - Google Patents

Abstract

The invention discloses a kind of construction methods of Electric Vehicles Driving Cycle, the construction method can not accurate evaluation electric vehicle property indices and the problem of energy consumption for existing driving cycle, on the basis of city road network information and the daily magnitude of traffic flow, establish the best trip mode model of driver, the computational methods and allocation rule of sample size needed for experiment are formulated, and test course has been planned according to this, it is determined that test period.A large amount of test data is obtained using GPS/IMU equipment, working characteristics and its operation characteristic in urban road for motor in electric automobile, it has formulated data prediction, the rule that data parse, has been finally based on the driving cycle that Markov static state Monte Carlo Analogue Method constructs electric vehicle.

Description

A method for constructing driving conditions of electric vehicles

technical field

The invention belongs to the field of development of actual road driving conditions of electric vehicles, and in particular relates to a construction method of driving conditions of electric vehicles.

Background technique

Vehicle driving conditions are a comprehensive reflection of many factors such as roads, climate environment, and driving habits in a country (region). One of the key factors restricting the development of the automobile industry.

At present, the existing driving conditions cannot meet the actual national conditions of our country, including road traffic characteristics, environmental conditions, etc.; nor can they truly reflect the performance of new energy vehicles, because the traditional energy conditions cannot evaluate the air conditioning and braking energy of new energy vehicles The impact of recycling and other situations on power consumption; at the same time, it also facilitates the entry of foreign new energy vehicles into the Chinese market, which is not conducive to the development of my country's independent new energy vehicles. In addition, the current research on driving conditions is mostly based on diesel locomotives, and mainly focuses on the analysis of test data and the construction of driving conditions in the later stage. There is not much research on the early stage of test planning, which will affect The authenticity of the test data and the reliability of the constructed driving conditions.

Contents of the invention

In order to make up for the above deficiencies, the present invention proposes a method for constructing driving conditions of electric vehicles for new energy vehicles represented by electric vehicles. The construction method clarifies in detail the specific content from the initial test planning stage of the construction of driving conditions, to the mid-term data collection stage, to the later stage of data analysis and construction of operating conditions. Comparing and analyzing the method with the existing short-stroke method, fixed-step interception method and V-A matrix analysis method, it can be seen that the accuracy of the driving conditions constructed by the method of the present invention is higher, and can truly reflect the actual road traffic conditions in the city . To sum up, the method for constructing driving conditions in the present invention has the characteristics of perfect theory, easy implementation, and high accuracy of constructed driving conditions, and is suitable for the development of driving conditions in large and medium-sized cities.

The present invention adopts following technical scheme to realize:

A method for constructing a driving condition of an electric vehicle, comprising the following steps:

1) Conduct research on the city's traffic conditions to obtain the city's basic road network information and daily traffic flow;

2) According to the obtained road network information, calculate the sample size and distribution ratio required for the test route, and plan the test route accordingly; according to the daily traffic flow of the city, analyze the peak period, off-peak period and low-peak period of the city's daily traffic conditions. Peak period, thus determine the test time;

3) According to the test route and test time, the road data collection test is carried out to obtain the speed-time information of the vehicle;

4) Analyze the speed-time data of the vehicle, obtain the different driving states of the vehicle and the transition probabilities between states, and construct the driving conditions.

The further improvement of the present invention is that, in the step 1), the road network information of the city is obtained by means of literature review and data collection; by reading the monitoring video, the main expressways, main roads, secondary roads and roads in the city are obtained Traffic flow on branch roads in one month.

The further improvement of the present invention is that, in said step 2), on the basis of urban road network information, a sample is selected in the overall road network by sampling survey to characterize the overall characteristics, and formula (1) is used to calculate the required The sample size n:

Among them, S ² is the sample standard deviation, S ² ≈P(1-P), and Δ is the maximum allowable absolute error;

In order to reasonably allocate the proportion of roads of each grade in the sample capacity, based on the analytic hierarchy process, the length of the road network of expressways, trunk roads, secondary trunk roads and branch roads is used as the input, and the width of the road network and the speed limit level are used as the criteria to establish The driver’s best travel mode model, according to the ratio of roads of each grade output by the model, calculates the required length of roads of each grade, and plans the test route accordingly to ensure that the test route is connected end to end, and the error of the calculation result is within 5%;

According to the obtained traffic flow, analyze the peak period, low peak period and flat peak period of a day's traffic in the city, and select 2 hours as the test time; select electric vehicles with a large number of vehicles in the city as test vehicles.

The further improvement of the present invention is, described step 3) in, carry out a one-week road data collection test according to the test route and test time of formulating, utilize GPS and IMU inertial navigation system to collect the speed-time information of vehicle travel, sampling frequency Set to 1Hz.

A further improvement of the present invention is that, in the step 4), in the data analysis stage, aiming at the driving characteristics of the electric vehicle on urban roads, at first the speed-time data is denoised using the formula (2), and then the formula (3 ) for smoothing:

Among them, v _t represents the vehicle speed before processing, v′ _t represents the vehicle speed after processing, and k is a smoothing parameter;

Use the formula (4) to calculate the acceleration at each moment:

Among them, v _t is the vehicle speed at the current moment, v _t-1 is the vehicle speed at the previous second, and the unit is km/h, and _at is the acceleration at the current moment, and the unit is m/s ² ;

According to the speed and acceleration changes of the vehicle, the speed data is divided into acceleration, deceleration, constant speed and idle speed segments by formula (5):

Select the maximum speed, minimum speed, average speed, speed standard deviation, maximum acceleration, maximum deceleration, average acceleration, acceleration standard deviation, speed range and running time, a total of 10 characteristic parameters to describe the fragment characteristics; use the principal component analysis method Dimensionality reduction is performed on the characteristic parameters, and the principal component score is calculated as a new attribute variable of the segment; the segment is divided into 6 categories by using the K-Means clustering algorithm, and the velocity and acceleration characteristics of each segment are counted, and the obtained segment is defined as There are 6 states of strong acceleration, strong deceleration, weak acceleration, weak acceleration, high-speed constant speed and low-speed constant speed, and the transition probability between states is calculated according to formula (6):

In the formula, N _ij represents the frequency of the current state i and the next state j; p _ij represents the probability that the current state is i and the next state is j, and l is the number of categories.

The further improvement of the present invention is that in the step 4), in the construction stage of the driving condition, first randomly select an idle segment whose duration does not exceed 5s as the initial part, and then use MATLAB to generate a uniform distribution between [0, 1] A random number x, if the random number satisfies:

Then the next state is q. In state q, select a segment without replacement and connect it end to end with the previous segment, then assign q to i, and repeat the previous steps to select a segment until the cumulative duration of the driving condition reaches 1200s. The selected fragments should meet the following principles:

(1) The distance between the selected segment and the cluster center of this class should be within the first 15%;

(2) The difference between the starting speed of the selected segment and the ending speed of the previous segment should be within 1km/h;

(3) When the number of segments satisfying the above two principles is not unique, the segment with the closest distance to the cluster center is preferred;

After that, multiple sets of random numbers are continuously regenerated to construct multiple alternative working conditions;

Select average speed, acceleration time ratio, deceleration time ratio, constant speed time ratio, idle time ratio, 0~10km/h speed ratio, 10~20km/h speed ratio, 20~30km/h speed ratio, 30~40km /h speed range ratio, 40-50km/h speed range ratio, and 50km/h speed range ratio, a total of 11 characteristic parameters describing the statistical distribution are used as the evaluation criteria of driving conditions. According to formula (8), the alternative working conditions and The average relative error d between the characteristic parameters of the original test data:

In the formula, t _i is the i-th characteristic parameter of the alternative working condition, z _i is the i-th characteristic parameter of the original test data, and m is the number of characteristic parameters;

Finally, the alternative operating condition with the smallest average relative error with the test data is selected as the final fitted electric vehicle driving condition.

Compared with the prior art, the present invention revolves around new energy vehicles represented by electric vehicles, on the basis of urban road network information and daily traffic flow big data, establishes the driver's optimal travel mode model, and plans the test accordingly. Route, determine the test time. A large amount of vehicle speed-time data is obtained through the actual road data collection experiment. Combined with the operation characteristics of electric vehicles on urban roads, the original test data are denoised and smoothed by using filtering algorithm. According to the characteristics of the speed and acceleration of the electric vehicle, the test data is divided into several segments of acceleration, deceleration, constant speed and idle speed, and the characteristics of the segments are described. The main information of the characteristic parameters was extracted by principal component analysis, and new segment attribute variables were calculated; six motion states of vehicles were divided by K-Means clustering algorithm, and the transition probabilities between states were counted. In order to solve the problem that the traditional Markov method will cause the loss of small probability events, the static Monte Carlo simulation method is used to construct the driving conditions of electric vehicles, which fully reflects the randomness and uncertainty of electric vehicles in urban road driving.

Description of drawings

Fig. 1 is a flowchart of the present invention.

Fig. 2 is the driver's optimal travel mode model established by the present invention.

Fig. 3 is the test route formulated by the present invention.

Fig. 4 is a flow chart of test data processing in the present invention.

Fig. 5 is a diagram of the driving conditions of the electric vehicle constructed in the present invention.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

Embodiment: Taking a typical large city in China as an example, the driving conditions of electric vehicles are constructed.

1. Traffic situation research and test planning

Referring to Figure 1, in order to make the constructed driving conditions fully reflect the actual traffic conditions of a certain city, the road network information of a certain city is first investigated, as shown in Table 1:

Table 1 Road network information of a large city

In order to reduce the test cost, a sampling method without replacement is adopted, a sample is selected from the road network population, and the characteristics of the population are reflected through the characteristics of the sample, so as to achieve the purpose of investigation. Assuming that the road network length X of a certain city obeys normal distribution, that is, X～N(μ,σ ² ), (x ₁ ,x ₂ ,…,x _n ) is a simple random sample from the overall road network, and is determined by statistics It can be seen from the principle that the drawn samples obey the normal distribution which is

Through calculation, it is found that when ^σ2 is unknown, the same confidence interval can be obtained no matter whether parameter estimation or hypothesis testing is performed on the overall mean Set the sample mean The maximum absolute error allowed when estimating or testing the population mean μ is On the premise of knowing the maximum absolute error Δ and the confidence level 1-α, the necessary sample size n at this time can be calculated, as shown in formula (1):

When the population variance σ ² is unknown, σ ₀ ² can be used instead of S ² . In the case of significance level α≤0.05, when the sample population is greater than 30, t _α (n-1)≈2, the sample size n can be calculated using the approximate formula (2):

In the formula, S ² ≈P(1-P), so the maximum value of sample standard deviation is 0.25. In the sampling survey, it is hoped to find an interval in which the confidence interval of the sample size is 95% and the limit error is 16%. According to the survey results in Table 1, the total length of urban roads in a certain city is 2562km, and the sample size to be selected is 39.04km according to formula (2).

In order to allocate the sample size of different grades of road tests, the overall planning is made in combination with the traffic flow and sample size of each grade of roads. Considering the differences in the north-south and east-west traffic of a certain city, as well as the different traffic intensities of roads of different grades, 11 monitoring points were selected, and the monitoring videos passing through the monitoring points from 7:00 a.m. to 20:00 p.m. were read, including 2 Expressway, 5 arterial roads, 4 secondary arterial roads and 1 branch road, the final statistics are the average number of vehicles passing the monitoring point in one day, as shown in Table 2:

Table 2 The average number of vehicles passing the monitoring point in one day

road type Freeway main road secondary road branch road Average value (vehicle) 47065 24414 14928 5588

In addition, when traveling in traffic, you can choose express roads, main roads, secondary arterial roads, or branch roads, and no matter which road you choose, you can send travelers to their destinations. When making a choice, many factors such as road width, speed limit level, convenience of the road network and traffic conditions are often considered. Therefore, based on the analytic hierarchy process, taking the length of the road network of expressways, main roads, secondary arterial roads, and branch roads as input, and the width and level of the road network as criteria, the optimal travel mode model for drivers is established as shown in Figure 2 , the output of the model is the sampling ratio of roads of each grade in the test route, and combined with the sample size, the required length of roads of each grade is calculated, as shown in Table 3:

Table 3 Proportion and length of roads of each grade in the test sample

road grade Freeway main road secondary road branch road Proportion 0.2996 0.2480 0.2685 0.1839 Length (km) 11.69 9.68 10.48 7.18

Combined with the length of roads of each grade in Table 3, plan the test route. The test route must be able to reflect the overall comprehensive characteristics of urban roads. The overall route needs to cover the main commercial areas, industrial areas, cultural areas, and living areas. The final planned test route is shown in Figure 3.

2. Road data collection

In order to make the constructed driving conditions more representative, a certain type of electric vehicle with a large population in the city was selected as the test vehicle; in order to avoid the influence of driving behavior, an experienced driver was selected to drive at 7:00 in the morning rush hour of a certain city. -9:00, afternoon peak period 12:30-14:30, evening low peak period 19:00-21:00 cyclically drive on the test route to collect data. Considering that when the vehicle is driving on a boulevard or passing tall buildings, GPS may not be able to receive satellite signals stably, resulting in the loss of sampling points, speed curve glitches and other problems, during the test, in addition to GPS, it is also equipped with an IMU inertial measurement unit Synchronously collect the speed-time data of the vehicle. In order to collect as much data as possible, the sampling frequency is set to 1Hz.

3. Analysis of test data and construction of driving conditions

As shown in Figure 4, the test data analysis is mainly divided into three parts: data preprocessing, test data analysis and driving condition construction. The following three parts will be described separately.

1. Data preprocessing

In addition to the sampling equipment, the test data will be distorted due to the driver's operation error, sudden changes in road traffic conditions, etc., so the test data needs to be denoised and smoothed. Use the formula (3) to remove the abnormal points in the test data:

Among them, v _t represents the vehicle speed before processing, and v′ _t represents the vehicle speed after processing.

Due to the need to divide the original test data, in order to make the divided fragments not too messy and the length of the fragments not too small, use the formula (4) to smooth the data:

where k is a smoothing parameter.

2. Analysis of test data

Firstly, according to the change of vehicle motion state (acceleration, deceleration, constant speed, idle speed), the test data is divided into several small driving segments. Combined with the operating characteristics of electric vehicle motors, the division rules shown in formula (5) are adopted:

After data division, a total of 15682 driving segments were obtained. Select 10 characteristic parameters such as maximum speed, minimum speed, average speed, speed standard deviation, maximum acceleration, maximum deceleration, average acceleration, acceleration standard deviation, initial and final speed difference, and running time to describe the characteristics of the driving segment. Principal component analysis was used to reduce the dimensionality of the feature parameters of the fragments, and K-Means clustering analysis was used to classify the fragments with the same characteristics into one category. Finally, 15682 kinematic fragments were divided into 6 categories, as shown in Table 4. According to the characteristics of each type of speed and acceleration, these six types of segments are defined as: high-speed uniform speed driving, low-speed uniform speed driving, weak deceleration driving, strong deceleration driving, weak acceleration driving, and strong acceleration driving.

Table 4 Classification results and various features

According to the clustering results, the attributes of each driving segment can be obtained, and the state transition probability matrix between the 6 types of segments can be calculated using the formula (6):

Among them, N _ij represents the frequency that the current state is i and the next state is j; p _ij represents the probability that the current state is i and the next state is j, and l is the number of categories. The transition probability matrix between various types is shown in Table 5:

Table 5 State transition probability matrix

3. Construction of driving conditions

In the construction stage of driving conditions, in order to solve the problem that the traditional Markov method will cause the loss of small probability events, the static Monte Carlo simulation method is used to construct the driving conditions. The specific steps are as follows:

First, randomly select an idle segment whose duration does not exceed 5s as the starting part, and then use MATLAB to generate a uniformly distributed random number x between [0, 1]. If the random number satisfies:

Then the next state is q. In state q, select a segment without replacement and connect it end-to-end with the previous segment, then assign q to i, and repeat the previous steps to continuously select segments until the cumulative duration of the driving condition reaches 1200s. . The selected fragments should meet the following principles:

(1) The distance between the selected segment and the cluster center of this class should be within the first 15%;

(2) The difference between the starting speed of the selected segment and the ending speed of the previous segment should be within 1km/h;

(3) When the number of segments satisfying the above two principles is not unique, the segment with the closest distance to the cluster center is preferred.

After that, multiple sets of random numbers are continuously regenerated to construct multiple alternative working conditions.

Select average speed, acceleration time ratio, deceleration time ratio, constant speed time ratio, idle time ratio, 0~10km/h speed ratio, 10~20km/h speed ratio, 20~30km/h speed ratio, 30~40km 11 characteristic parameters describing the statistical distribution, such as the ratio of the speed range of /h, the ratio of the speed range of 40-50km/h, and the ratio of the speed range of 50km/h or more, are used as the evaluation criteria of the driving conditions, and the alternative working conditions are calculated according to formula (8). The average relative error d between the characteristic parameters of the original test data:

In the formula, t _i is the i-th characteristic parameter of the alternative working condition, z _i is the i-th characteristic parameter of the original test data, and m is the number of characteristic parameters. Finally, the alternative operating condition with the smallest average relative error with the test data is selected as the final fitted electric vehicle driving condition.

4. Verification of driving conditions

In order to verify the accuracy of the driving conditions constructed by this method, this method is compared with the driving conditions constructed by the existing short-stroke method, V-A matrix method and fixed-step interception method. Table 6 compares the characteristic parameters of the driving conditions constructed by the four methods with the original test data, and Table 7 shows the average relative error between the driving conditions constructed by the four methods and the original test data. The result shows that the average error rate between the method of the present invention and the original test data is only 6.34%, while the average error rate between the short-stroke method, the V-A matrix method and the fixed-step interception method and the original test data is 12.62%, 9.52%, respectively, 15.02%. It can be seen that the driving conditions constructed by the method of the present invention have higher accuracy and can truly reflect the actual road traffic conditions in the city.

Table 6 The characteristic parameters of driving conditions and original test data constructed by four methods

Table 7 The average relative error between the driving conditions constructed by the four methods and the original test data

Method in this paper V-A matrix method short stroke method fixed-step truncation method average error(%) 6.34 9.52 12.62 15.02

Claims (6)

1. A construction method of electric vehicle driving condition, is characterized in that, comprises the following steps: 1) Conduct research on the city's traffic conditions to obtain the city's basic road network information and daily traffic flow; 2) According to the obtained road network information, calculate the sample size and distribution ratio required for the test route, and plan the test route accordingly; according to the daily traffic flow in the city, analyze the peak period, off-peak period and low-peak period of the city's daily traffic conditions. Peak period, thus determine the test time; 3) According to the test route and test time, the road data collection test is carried out to obtain the speed-time information of the vehicle; 4) Analyze the speed-time data of the vehicle, obtain the different driving states of the vehicle and the transition probabilities between states, and construct the driving conditions. 2. The construction method of a kind of electric vehicle driving condition according to claim 1, is characterized in that, in described step 1), obtain the road network information of city by the mode of literature review, data collection; By reading monitoring In the form of video, the traffic flow of major expressways, trunk roads, secondary trunk roads and branch roads in the city is obtained within a month. 3. The construction method of a kind of electric vehicle driving condition according to claim 2, characterized in that, in said step 2), on the basis of urban road network information, in the overall road network by sampling survey Select a sample to characterize the overall characteristics, and use the formula (1) to calculate the sample size n required for the test: Among them, S ² is the sample standard deviation, S ² ≈P(1-P), and Δ is the maximum allowable absolute error; In order to reasonably allocate the proportion of roads of each grade in the sample capacity, based on the analytic hierarchy process, the length of the road network of expressways, trunk roads, secondary trunk roads and branch roads is used as the input, and the width of the road network and the speed limit level are used as the criteria to establish The driver’s best travel mode model, according to the ratio of roads of each grade output by the model, calculates the required length of roads of each grade, and plans the test route accordingly to ensure that the test route is connected end to end, and the error of the calculation result is within 5%; According to the obtained traffic flow, analyze the peak period, low peak period and flat peak period of a day's traffic in the city, and select 2 hours as the test time; select electric vehicles with a large number of vehicles in the city as test vehicles. 4. the construction method of a kind of electric vehicle running condition according to claim 3, it is characterized in that, in described step 3), according to the test route and test time of formulating, carry out a one-week road data collection test, utilize The GPS and IMU inertial navigation system collect the speed-time information of the vehicle, and the sampling frequency is set to 1Hz. 5. The construction method of a kind of driving condition of an electric vehicle according to claim 4, characterized in that, in the step 4), in the data analysis stage, for the driving characteristics of the electric vehicle on urban roads, at first using the formula (2) Denoise the speed-time data, and then use formula (3) for smoothing: Among them, v _t represents the vehicle speed before processing, v _t ' the vehicle speed after processing, and k is a smoothing parameter; Use the formula (4) to calculate the acceleration at each moment: Among them, v _t is the vehicle speed at the current moment, v _t-1 is the vehicle speed at the previous second, and the unit is km/h, and _at is the acceleration at the current moment, and the unit is m/s ² ; According to the speed and acceleration changes of the vehicle, the speed data is divided into acceleration, deceleration, constant speed and idle speed segments by formula (5): Select 10 characteristic parameters including maximum speed, minimum speed, average speed, speed standard deviation, maximum acceleration, maximum deceleration, average acceleration, acceleration standard deviation, speed range and running time to describe the fragment characteristics; use principal component analysis method Dimensionality reduction is performed on the characteristic parameters, and the principal component score is calculated as a new attribute variable of the segment; the segment is divided into 6 categories by using the K-Means clustering algorithm, and the velocity and acceleration characteristics of each segment are counted, and the obtained segment is defined as There are 6 states of strong acceleration, strong deceleration, weak acceleration, weak acceleration, high-speed constant speed and low-speed constant speed, and the transition probability between states is calculated according to formula (6): In the formula, N _ij represents the frequency of the current state i and the next state j; p _ij represents the probability that the current state is i and the next state is j, and l is the number of categories. 6. The construction method of a kind of driving condition of an electric vehicle according to claim 5, characterized in that, in the step 4), in the construction stage of the driving condition, at first randomly select an idle section whose duration is no more than 5s as The initial part, and then use MATLAB to generate a uniformly distributed random number x between [0, 1], if the random number satisfies: Then the next state is q. In the state q, select the segment without replacement and connect it end to end with the previous segment, then assign q to i, and repeat the previous steps to select the segment until the cumulative duration of the driving condition reaches 1200s. The selected fragments should meet the following principles: (1) The distance between the selected segment and the cluster center of this class should be within the first 15%; (2) The difference between the starting speed of the selected segment and the ending speed of the previous segment should be within 1km/h; (3) When the number of segments satisfying the above two principles is not unique, the segment with the closest distance to the cluster center is preferred; After that, multiple sets of random numbers are continuously regenerated to construct multiple alternative working conditions; Select average speed, acceleration time ratio, deceleration time ratio, constant speed time ratio, idle time ratio, 0~10km/h speed ratio, 10~20km/h speed ratio, 20~30km/h speed ratio, 30~40km /h speed range ratio, 40-50km/h speed range ratio, and 50km/h speed range ratio, a total of 11 characteristic parameters describing the statistical distribution are used as the evaluation criteria of driving conditions. According to formula (8), the alternative working conditions and The average relative error d between the characteristic parameters of the original test data: In the formula, t _i is the i-th characteristic parameter of the alternative working condition, z _i is the i-th characteristic parameter of the original test data, and m is the number of characteristic parameters; Finally, the alternative working condition with the smallest average relative error with the test data is selected as the final fitted electric vehicle driving condition.