CN113536518A

CN113536518A - Method for estimating remaining driving range of pure electric vehicle

Info

Publication number: CN113536518A
Application number: CN202010319512.XA
Authority: CN
Inventors: 田慧欣; 李昂
Original assignee: Tianjin Polytechnic University
Current assignee: Tianjin Polytechnic University
Priority date: 2020-04-22
Filing date: 2020-04-22
Publication date: 2021-10-22

Abstract

A method for estimating the remaining driving range of a pure electric vehicle. The invention belongs to the technical field of electric automobiles. The method aims to provide a method for estimating the remaining driving range of the pure electric vehicle based on map information and an iterative SVR model. The method comprises the following steps: (1) and analyzing influence factors influencing the driving range according to the automobile physical model and the actual data. (2) And carrying out segment division on the running condition data, and extracting the characteristic parameters of the segments. (3) And classifying the working condition segments by a clustering method based on the characteristic parameters. (4) And establishing an SVR model for calculating the SOC (state of charge) reduction value of the unit mileage. (5) And acquiring road information by utilizing a Baidu map API path planning function, and predicting the future driving condition. (6) And predicting the remaining driving range of the automobile by circularly iterating the SVR model and combining the current SOC value, the working condition parameters and the temperature value.

Description

Method for estimating remaining driving range of pure electric vehicle

Technical Field

The method is applied to the technical field of pure electric vehicles, and is a pure electric vehicle remaining driving range estimation method based on map information and an iterative SVR model, and particularly aims at the current situation that the driving range prediction precision is difficult to improve due to unknown driving conditions.

Background

With the aggravation of environmental pollution, the pure electric vehicle is supported by the great force of the country with the advantages of energy conservation and environmental protection, and the purchasing power of consumers on the electric vehicle is improved by releasing policies such as subsidy, tax deduction and exemption. However, due to the limitation of battery technology, the problems of low battery energy density and the like are not effectively solved, so that the battery capacity of the pure electric vehicle cannot reach the endurance level of the traditional vehicle. Meanwhile, due to the influence of road conditions and environment, the actual driving range of the pure electric vehicle is often different from the driving range given by the BMS, so that the phenomenon of 'range anxiety' of a driver is caused. Therefore, the driving range of the automobile needs to be accurately estimated in combination with the actual driving range, so that the worry of a driver is eliminated, and the method plays an important role in promoting the popularization and development of the pure electric automobile.

Driving range is divided into two categories: and calibrating the driving range and the residual driving range. The calibrated driving range refers to the longest distance that the electric automobile can drive under a certain cycle working condition. The remaining driving range refers to the maximum distance that the vehicle can travel when the vehicle runs to the battery cut-off voltage in a state that the battery is not fully charged in the actual running process.

Because the energy consumed by the electric automobile is different under different driving conditions and different temperatures, the driving range of the automobile is different under different conditions and different temperatures. The method predicts unknown driving conditions by collecting map information, and then adopts the iterative SVR model to accurately calculate the driving range, so that the prediction accuracy is improved.

Disclosure of Invention

The invention provides a driving range estimation method based on map information and an iterative SVR model, which comprises the following specific steps:

1. and analyzing factors influencing the driving range of the pure electric vehicle, wherein the factors comprise the factors of the pure electric vehicle and external environment factors.

1) Based on a physical mechanical model of vehicle motion, a calculation formula of mechanical energy dE required by a vehicle driving unit mileage ds is as follows:

in the formula, the unit of ds is km; dE is in kWh; m is the total mass of the vehicle in kg; m is_fIs equivalent mass of rolling inertia, unit kg; g is gravity acceleration in m/s²(ii) a f is the vehicle mass rolling coefficient;

is the road surface inclination angle in units of degrees (°); rho is air density in kg/m³；C_xIs the vehicle drag coefficient; a is the frontal area of the vehicle in m²(ii) a v is the running speed of the vehicle, and the unit is km/h; v. of_wIs the wind speed, the direction of which is defined as the unit km/h, opposite to the speed of the vehicle. As can be seen from the formula, the energy consumption of the mileage of the automobile can be influenced by the parameters of the automobile, such as the total mass of the automobile and the frontal area of the automobile. In general, the resistance which needs to be overcome by an automobile in the driving process comes from four parts, namely rolling resistance, acceleration resistance, ramp resistance and air resistance, except for the air resistance, the resistance and the total mass of the automobile, so that the influence of the total mass of the automobile on energy consumption is most obvious, and the driving range of the automobile is further influenced.

2) Besides the factors of the automobile itself, many external factors also affect the energy consumption of the automobile, such as the driving acceleration, the driving stability, the environmental temperature and the like, and the external factors can be summarized into the driving condition and the temperature of the automobile on the whole.

Under different driving conditions, the driving range of the automobile is different. Comparing the driving range of the automobile under different working conditions, as shown in fig. 1, wherein CON _60 is a constant speed working condition with a vehicle speed of 60 miles per hour; CON _45 is a constant speed working condition with the vehicle speed of 45 miles per hour; the WVYCITY working condition is an urban congestion working condition; the EUDC working condition is a suburb driving working condition; the ECE working condition is an urban driving working condition; the NEDC consists of 4 ECE conditions and 1 EUDC condition, which may be considered a hybrid condition.

The temperature also can obviously influence the driving range of the automobile, mainly reflected in the influence on the performance of the battery, the energy release rate is reduced in a low-temperature environment, the driving range of the same automobile under the same working condition and different temperatures is compared, as shown in fig. 2, the driving range of the electric automobile is reduced along with the reduction of the temperature.

2. Segmenting the original data of the running condition of the automobile, dividing the data into a plurality of segments by a fixed step length division method according to the step length of 120 seconds, and extracting characteristic parameters of the operating condition segments including average speed through a speed curve as shown in figure 3

Maximum driving speed v_maxIdle time ratio P_dStandard deviation of velocity v_stdSOC value and battery temperature.

3. Classifying the driving condition segments, and dividing the driving condition segments into four classes by adopting a fast search and discovery density peak clustering algorithm (CBFSAFODP), wherein the four clustering classes shown in FIG. 4 can correspond to 4 typical working conditions in the actual driving process of an automobile, C1 is an urban congestion working condition, the idle time is long, and the speed is low; c2 is an urban low-speed working condition, the idling time is medium, and the vehicle speed is low; c3 is suburb/urban smooth working condition, the idling time is short, and the vehicle speed is high; c4 is high speed, very short idle time and high speed.

4. And constructing an estimation model of the SOC (state of charge) reduction value of the driving unit mileage, wherein the variation of the SOC of the automobile unit mileage is related to the driving working condition, the current battery SOC value and the battery temperature. Based on the SVR model, the unit mileage SOC variation (delta SOC) is taken as a target output, and the working condition type, the battery pack temperature and the current battery SOC are taken as inputs. The estimation result of the model is shown in FIG. 5, the model error is shown in FIG. 6, and the model precision meets the requirement of estimating the SOC reduction value of the unit mileage in the driving range prediction.

5. And predicting the unknown running condition by combining the map information. The method comprises the following steps of obtaining 10 road types through a Baidu map API, summarizing the road types with small differences into four categories: high speed, primary, secondary and other roads. The road condition information is divided into five categories: no road condition, smooth, slow running, congestion and very congestion are respectively represented by the numbers 0-4. And determining the corresponding working condition type according to the road type and the road condition index. In the actual driving process of the automobile, the road type and road condition information is obtained through the path planning function of the Baidu map API, the prediction of the future driving condition is realized, and the method has important significance for improving the prediction precision of the driving range.

6. And (3) constructing an iterative SVR driving range prediction model, wherein the flow chart is shown in figure 7, and the overall block diagram is shown in figure 8.

1) And acquiring the current SOC value and the current battery temperature of the battery, and predicting the unknown running condition type through a Baidu map path planning function and map information. And predicting the change delta SOC of the SOC running for one kilometer in the future by using the SOC decline value model of the unit mileage and taking the current SOC value of the battery, the battery temperature and the working condition type as input.

2) And calculating the remaining SOC value of the battery according to the SOC-delta SOC.

3) And (5) circulating the step 1 and the step 2 until the SOC of the battery is equal to 0 or less than a set value. The electric automobile runs for one kilometer after one cycle, so the total cycle number R is the driving range of the electric automobile.

4) And when the cycle R times meets the residual mileage of the planned path, the SOC value of the battery still does not reach 0 or the minimum value is set, then a simple SVR model is adopted to estimate the residual driving mileage, the current SOC value of the battery, the average working condition parameter of the previous road section and the temperature are used as input, the residual driving mileage R 'of the automobile is roughly estimated, and the final driving mileage is R + R'.

Drawings

FIG. 1 shows the driving range of the electric vehicle under different working conditions

FIG. 2 driving range at different ambient temperatures

FIG. 3 fragmentation example

FIG. 4 clustering results of the condition segments

FIG. 5 model estimation Effect

FIG. 6 model error

FIG. 7 iterative SVR model flow chart

FIG. 8 is an overall flowchart of a driving range prediction method based on map information and an iterative SVR model

FIG. 9 automobile driving route

FIG. 10 vehicle travel speed Curve

FIG. 11 predicted result and predicted accuracy of driving range

Examples of the applications

And designing a simulation experiment by combining ADVISOR software and actual vehicle running data, extracting the actual running speed on the basis of the vehicle running data of a certain transportation enterprise, and loading the speed into the ADVISOR so that the simulation electric vehicle runs according to the speed to simulate the running process under the actual condition.

In order to evaluate the prediction accuracy of the model under different trips, a standardized index needs to be defined, and the prediction accuracy cannot accurately reflect the different lengths of the mileage simply by subtracting the change of the driving range from the actual driving distance, so that a model accuracy calculation formula is defined as follows:

where a is the prediction accuracy, Distance is the actual driving Distance, and Range _ use is the change in the remaining Range (the difference between the first prediction result and the current prediction result). The prediction precision is higher as A is closer to 0, and when A is larger than 0, the residual driving range given by the algorithm is more conservative, otherwise, if A is smaller than 0, the given driving range is higher.

In addition, in order to measure the average accuracy of the results of multiple tests, an absolute average value E of the prediction accuracy is defined, as shown in formula (5-2):

in the formula, A_iFor the prediction accuracy of the i-th test, N isThe smaller E represents the higher prediction accuracy of the model in the total number of tests.

Taking a trip extracted from the data as an example, the starting point of the trip is Jian city, Jinggang mountain, Jiangxi province, and the ending point is the west lake region of Nanchang city, Jiangxi province, along the great river, the vehicle driving route of the trip is shown in FIG. 9, and the speed curve is shown in FIG. 10. The speed of the journey is loaded into the ADVISOR, the set EV1 type electric vehicle is used for completing the mileage according to the speed, and the driving range is predicted in the simulation process. The result shows that the travel of the section totals 322.16km, the residual driving range is reduced from 517km to 189km in the simulation, the driving range is changed to 328km, and the prediction precision is-0.0186, as shown in FIG. 11.

According to the result, the pure electric vehicle driving range prediction method based on the map information and the iterative SVR model meets the precision requirement, and the pure electric vehicle driving range can be accurately estimated.

Claims

1. A pure electric vehicle remaining driving range estimation method based on map information and an iterative SVR model is characterized in that: the method for estimating the remaining driving range of the pure electric vehicle based on the map information and the iterative SVR model is characterized by comprising the following steps of: constructing a unit mileage SOC descending value calculation model based on the characteristic analysis of the driving segment; and on the basis of a vehicle navigation system, in combination with map information, a driving range prediction model is constructed on the basis of the iterative SVR, and the remaining driving range is calculated.

2. The method for estimating driving range of the pure electric vehicle based on the map information and the iterative SVR model according to claim 1, wherein the segment division is performed based on a fixed-step division method, the data unit is defined, and a plurality of characteristic parameters including average speed in the working condition segment are extracted

Maximum driving speed v_maxIdle time ratio P_dStandard deviation of velocity v_stdCurrent SOC value and battery temperature.

3. The pure electric vehicle driving range estimation method based on map information and iterative SVR model as claimed in claim 1, characterized in that fast search and discovery density peak Clustering (CBFSAFODP) algorithm is used to divide the working condition segments and construct the SOC decline value calculation model of unit mileage.

4. The pure electric vehicle driving range estimation method based on the map information and the iterative SVR model according to claim 1, wherein a road type and road condition information are obtained based on a vehicle navigation system in combination with the map information, a corresponding working condition type is determined according to the road type and the road condition index, prediction of a driving working condition of a next kilometer is realized, and prediction accuracy of a driving range is effectively improved.

5. The pure electric vehicle driving range estimation method based on map information and iterative SVR model as recited in claim 1, wherein constructing a driving range prediction model specifically comprises the steps of:

1) and acquiring the current SOC value of the battery, and predicting the driving condition of one kilometer in the future through map information.

2) And calculating the SOC decline value delta SOC of the future one kilometer by using a unit mileage SOC decline value calculation model, and calculating the residual SOC value of the battery according to a formula SOC-delta SOC.

3) And repeating the step 1 and the step 2 until the remaining SOC is equal to 0 or less than a set value. And recording the total cycle number R, wherein the remaining driving range of the electric automobile is the total cycle number R because the automobile drives one kilometer in one cycle.

4) And when the vehicle is circulated for R times, namely the remaining range of the planned path is met, the SOC value of the battery still does not reach 0 or the minimum value is set, then the remaining driving range is estimated by adopting a simple SVR model, the current SOC value of the battery, the average working condition parameter of the previous road section and the temperature are used as input, the remaining driving range R 'of the vehicle is roughly estimated, and the final driving range is R + R'.