CN117755154A

CN117755154A - Power battery health state assessment method based on new energy automobile real vehicle working condition

Info

Publication number: CN117755154A
Application number: CN202311493606.9A
Authority: CN
Inventors: 刘骞; 高宏力
Original assignee: Southwest Jiaotong University
Current assignee: Southwest Jiaotong University
Priority date: 2023-11-10
Filing date: 2023-11-10
Publication date: 2024-03-26

Abstract

The invention discloses a power battery health state assessment method based on a new energy automobile real vehicle working condition, and relates to the technical field of automobile batteries. S1: determining health state attenuation coefficients alpha corresponding to different working condition intervals; s2: calculating the characteristic parameters of the actual vehicle conditions of the new energy vehicle; s3: constructing a characteristic parameter vector of a real vehicle working condition, and calculating a characteristic parameter vector similarity matrix; s4: classifying actual vehicle conditions; according to the similarity matrix, using a spectral clustering algorithm to divide all new energy automobiles into a plurality of actual automobile working condition categories; s5: sampling and testing the available capacity of the power battery of each real vehicle working condition type; s6: evaluation of classification effect; s7: evaluating the power battery health state values of different real vehicle working condition categories; the state of health value of the power battery of each real vehicle working condition type is evaluated by two methods of point estimation and interval estimation. The invention classifies the real vehicle working conditions of the new energy automobile into a plurality of categories, and evaluates the health state of the power battery of each category.

Description

Power battery health state assessment method based on new energy automobile real vehicle working condition

Technical Field

The invention relates to the technical field of automobile batteries, in particular to a power battery health state assessment method based on the actual vehicle working condition of a new energy automobile.

Background

The state of health of a power cell is generally defined as the ratio of the current power cell's available capacity after full charge to the full new power cell's rated capacity. The health state has important reference value in the whole life cycle of the new energy automobile. For example, in after-market maintenance of a vehicle, the state of health of the power battery is a core indicator for determining whether the power battery needs replacement; during the transaction of a second-hand vehicle, the health state of the power battery is a key factor for determining the residual value of the new energy vehicle; when the vehicle is scrapped, the health state of the power battery is a main basis for determining the gradient utilization scene of the power battery.

Regarding the evaluation of the state of health of the power battery, numerous researchers have conducted a large number of experiments and proposed various types of algorithms. These algorithms perform well on the power cell experimental condition data set, but the evaluation effect on the real vehicle condition data set is not satisfactory. This is because under experimental conditions, researchers generally develop power battery charge-discharge cycle aging experiments at constant current, constant temperature, and constant state of charge intervals. In the experimental process, the accurate capacity value of the battery after the specific cycle times can be obtained through an ampere-hour integration method, and then the health state of the power battery can be accurately calculated. The accurate health state values can also be used as machine learning label values for machine learning of the health state of the power battery under experimental working conditions, so that a mature algorithm is formed.

However, the state of health assessment in real vehicle conditions is much more complex. The actual vehicle working conditions of the new energy automobile, such as slow charge or fast charge, the frequency and depth of a brake pedal, the ratio of high-speed and low-speed running, the frequency and depth of an accelerator pedal, the charge and discharge depth, the battery temperature, the accumulated running mileage and the like, cause the change of the charging multiplying power (the multiplying power is based on the capacity and reflects the difference of the current), the discharging multiplying power, the charge state, the working temperature, the charge and discharge cycle times and the like of the power battery at any time, and further cause that the accurate capacity value of the power battery is difficult to obtain. This is because the capacity needs to be obtained by ampere-hour integration at constant current, constant temperature, and constant state of charge intervals to obtain an accurate value. Even if the capacity of the power battery is estimated by some algorithm, it is difficult to determine how much the error is between the estimated value and the actual value. In fact, once the power battery pack is assembled on the vehicle, it is impossible to obtain the exact capacity of the power battery pack unless the offline capacity test is performed under the rated conditions, and thus it is impossible to accurately evaluate the health status, and it is also difficult to determine the tag value of the health status for machine learning.

In summary, the state of health evaluation of the power battery is generally performed under constant experimental conditions. Under the actual vehicle condition, the charging rate, discharging rate, state of charge, working temperature, charge-discharge cycle times and the like of the power battery change at any time, and the state of health of the power battery is difficult to accurately evaluate. Meanwhile, the health state of the real vehicle is difficult to evaluate, so that the machine learning of the health state of the real vehicle lacks a tag value, and the development of a machine learning algorithm of the health state of the real vehicle is further restricted.

Therefore, the invention provides a power battery health state assessment method based on the actual vehicle working condition of the new energy automobile.

Disclosure of Invention

The invention aims to provide a power battery health state assessment method based on a new energy automobile real vehicle working condition. According to the invention, the real vehicle conditions of the new energy automobile are classified into different categories, the health state of the power battery is estimated according to the real vehicle conditions, meanwhile, more accurate label values can be assigned to the health states of the power battery under different conditions, and the label values can be used for machine learning of the health states, so that the development of a power battery health state estimation algorithm under the real vehicle conditions is promoted.

The technical aim of the invention is realized by the following technical scheme: the power battery health state assessment method based on the real vehicle working condition of the new energy automobile comprises the following steps:

s1: determining health state attenuation coefficients alpha corresponding to different working condition intervals;

s2: calculating the characteristic parameters of the actual vehicle conditions of the new energy vehicle; the calculation of the characteristic parameters comprises calculation of the characteristic parameters of the charging rate interval, calculation of the characteristic parameters of the discharging rate interval, calculation of the characteristic parameters of the charge state interval, calculation of the characteristic parameters of the working temperature interval and calculation of the characteristic parameters of the charge and discharge cycle frequency interval;

s3: constructing a characteristic parameter vector of a real vehicle working condition, and calculating a characteristic parameter vector similarity matrix;

s4: classifying actual vehicle conditions; according to the similarity matrix, using a spectral clustering algorithm to divide all new energy automobiles into a plurality of actual automobile working condition categories;

s5: testing the available capacity of the power battery of each real vehicle working condition type;

s6: evaluation of classification effect;

s7: evaluating the power battery health state values of different real vehicle working condition categories; the state of health value of the power battery of each real vehicle working condition type is evaluated by two methods of point estimation and interval estimation.

The invention is further provided with: step S1, carrying out a charge-discharge cycle aging experiment of a power battery pack with single working condition change, respectively fitting the relationship between the capacity attenuation rate and the charge-discharge cycle times under the same working condition by using straight lines, then fitting the functional relationship between the working condition and the single cycle capacity attenuation rate by using curves, finally dividing the working condition into a plurality of sections at specific intervals, calculating the average value of the single cycle capacity attenuation rate corresponding to each working condition section, and defining the average value as a health state attenuation coefficient alpha corresponding to the working condition section, thereby determining health state attenuation coefficients alpha corresponding to different charging multiplying power sections, different discharging multiplying power sections and different working temperature sections; fitting the relationship between the capacity attenuation rate and the equivalent charge-discharge cycle times of the same charge state interval by using straight lines respectively, and obtaining the slope of each fitting straight line to serve as a health state attenuation coefficient alpha corresponding to each charge state interval; and fitting the capacity attenuation rates corresponding to different charge and discharge cycle times by using a straight line to obtain the single cycle capacity attenuation rate. Dividing the charge-discharge cycle times into a plurality of intervals at specific intervals, calculating the average number of the charge-discharge cycle times in each interval, and multiplying the average number of the charge-discharge cycle times in each interval by the single cycle capacity attenuation rate to obtain the health state attenuation coefficient alpha corresponding to the charge-discharge cycle times interval. The health state attenuation coefficient alpha is a parameter for measuring the attenuation speed of the health state, and the greater the value is, the faster the attenuation speed of the health state is, and the smaller the attenuation speed of the health state is otherwise.

The invention is further provided with: in the step S2, the characteristic parameter of the charging rate interval is the sum of products of the health state attenuation coefficient α corresponding to each charging rate interval and the duration time of the corresponding charging rate interval accounting for the proportion of all charging time, and the calculation formula is as follows:

in the above formula, i refers to an i-th charging rate interval; n refers to the total number of charging rate intervals.

The characteristic parameter of the discharge rate interval is the sum of the products of the health state attenuation coefficient alpha corresponding to each discharge rate interval and the total discharge time proportion of the duration time of the corresponding discharge rate interval, and the calculation formula is as follows:

in the above formula, j refers to a j-th discharge rate interval; r denotes the total number of discharge rate intervals.

The state of charge interval characteristic parameter is the sum of the products of the state of health attenuation coefficient alpha corresponding to each state of charge interval and the proportion of the duration time of the corresponding state of charge interval to all charge and discharge time, and the calculation formula is as follows:

in the above formula, k refers to a kth charge state interval; s refers to the total number of state of charge intervals.

The characteristic parameter of the working temperature interval is the sum of products of the health state attenuation coefficient alpha corresponding to each working temperature interval and the proportion of the duration time of the corresponding working temperature interval to the whole charge and discharge time, and the calculation formula is as follows:

In the above formula, q refers to a q-th working temperature interval; t refers to the total number of operating temperature intervals.

Taking the ratio of the accumulated charge-discharge capacity to the rated capacity as the charge-discharge cycle times; finding out the interval of the charge-discharge cycle times in the charge-discharge cycle aging result of the power battery pack to obtain a health state attenuation coefficient alpha, wherein the health state attenuation coefficient alpha is used as a characteristic parameter of the charge-discharge cycle times interval, and the calculation formula is as follows:

characteristic parameter of charge-discharge cycle number interval = health state attenuation coefficient alpha corresponding to interval where charge-discharge cycle number is located

The invention is further provided with: step S3 is to construct a characteristic parameter vector of the working condition of the real vehicle and calculate a characteristic parameter vector similarity matrix, and the steps are as follows:

(1) Calculating various characteristic parameters and median of the actual vehicle working conditions of each vehicle;

(2) Constructing a characteristic parameter vector of each automobile real-vehicle working condition; subtracting the corresponding median from each characteristic parameter of the actual vehicle working condition of each vehicle to be used as an element in the characteristic parameter vector of the actual vehicle working condition of the new energy vehicle;

(3) Calculating the similarity AS in the vector direction of the two characteristic parameters; firstly, the similarity COS in two vector directions needs to be calculated, and the formula is as follows:

Mapping the value of COS to between [0,1] to obtain AS, wherein the calculation formula is AS follows:

wherein a and b are characteristic parameter vectors of two automobile real vehicle working conditions respectively, a.b represents the dot product of the vector a and the vector b, and a and b represent norms (namely the length of the vector) of the vector a and the vector b. COS value range is between [ -1,1 ]; AS values range between [0,1] and gradually tend towards 1 AS the angle between the two vectors decreases, whereas tend towards 0;

(4) Calculating the similarity LS on the lengths of the two characteristic parameter vectors; the similarity LS of the vector a and the vector b in length is calculated, and the calculation formula is as follows:

wherein a and b are absolute values of differences between two vector norms a and b, and max (a and b) represents a larger value of a and b; LS values range between [0,1] and gradually tend to 1 as the difference between the two vector lengths decreases, whereas tend to 0;

(5) Calculating the similarity SIM of the two characteristic parameter vectors; the similarity SIM of the feature parameter vector is the product of the similarity AS of two feature parameters in the vector direction and the similarity LS in length, i.e

The SIM value ranges between 0,1 and gradually goes to 1 with increasing similarity of the two vectors, whereas it goes to 0;

(6) Constructing a similarity matrix; and calculating the similarity between the characteristic parameter vectors of the real vehicle working conditions of any two automobiles to obtain a similarity matrix.

The invention is further provided with: and S5, regularly extracting a certain number of automobiles from the new energy automobiles in each real automobile working condition category as test samples, and carrying out complete offline charge-discharge circulation in a 0-100% charge state interval under the rated working condition of specific charge-discharge current and working temperature to obtain an accurate available capacity value.

The invention is further provided with: step S6, because the power batteries of the same real vehicle working condition category are similar in health status and are obviously different from the available capacities of the power batteries of the new energy vehicles of other real vehicle working condition categories; therefore, the available capacity distribution of the power batteries of the new energy automobiles in various real automobile working condition categories is counted, whether the available capacity distribution of the same real automobile working condition category is concentrated in a certain section or not is judged, and whether the available capacity distribution of the same real automobile working condition category is remarkably different from the available capacity distribution of other working condition categories or not is judged, so that the classification effect is evaluated; the steps flow as follows:

A. determining the effective interval of the available capacity of the power battery of each real vehicle working condition type obtained by sampling test; the available capacity of the power battery of the new energy automobile of each real automobile working condition category obtained by sampling test is arranged in sequence from small to large, the beta percent quantile and the (1-beta percent) quantile are determined, and the interval between the beta percent quantile and the (1-beta percent) quantile is called an effective interval. The interval is used for representing the length of the effective interval;

B. Calculating the sum of maximum values among the overlapping lengths of each effective interval and other effective intervals; after classification, there are m real vehicle working condition categories, that is, m effective intervals, and two effective intervals are taken, wherein the first effective interval is [ X1, X2], and the second effective interval is [ Y1, Y2].

Under the condition that two effective sections are overlapped, subtracting a larger value of a left end point from a smaller value of a right end point of the two effective sections to obtain an overlapping length A, wherein the calculation formula is as follows:

in the case where the two effective sections do not overlap, the overlap length of the two effective sections is 0. Therefore, the calculation formula of the overlap length of the two effective sections is:

and traversing and calculating the overlapping length of each effective interval and other m-1 effective intervals to obtain the maximum value among the overlapping lengths of the effective interval and other m-1 effective intervals. Then, the sum of the maximum values among the overlapping lengths of all m effective sections and other effective sections is calculated, namely

Wherein, overlap _p Is the maximum value among the overlapping lengths of the p-th effective interval and other m-1 effective intervals;

C. calculating the sum of the lengths of all the effective intervals; the sum of the lengths of the m effective intervals is calculated as:

Wherein, interval _p A length interval of the p-th effective interval;

D. calculating a classification result verification parameter CRVP; defining a classification result verification parameter CRVP as a parameter for evaluating classification effect, wherein the calculation formula of the CRVP is as follows:

setting a certain CRVP threshold value when the value of CRVP is between [0,1], and judging the classification effect; if the classification effect cannot meet the requirement, the classification algorithm is adjusted again; if the classification effect meets the requirements, further estimating the power battery health state value of each real vehicle working condition class.

The invention is further provided with: the point estimation method in the step S7 is that the expectation of the available capacity value of the power battery in each real vehicle working condition class is the average value of the available capacities of the power batteries in all sampling tests in the class; the desire for health status is then calculated from the ratio of the desire for available capacity to the rated capacity.

The invention is further provided with: the interval estimation method in step S7 is that the confidence interval with the expected confidence of the available capacity of the power battery of each real vehicle working condition type is 1-gamma:

wherein,the average value of the available capacity of the power battery samples for the class sampling test is represented by S, which is the standard deviation of the samples, and n, which is the number of the samples. A confidence interval for the health expectations is then calculated from the ratio of the expected available capacity to the rated capacity.

The invention is further provided with: and (3) taking the power battery health state values of different real vehicle working condition categories as tag values, calibrating the power battery health states of the corresponding working condition categories, and using the power battery health state values for machine learning of real vehicle working condition health state evaluation.

In summary, the invention has the following beneficial effects: the invention provides a calculation method of a health state attenuation coefficient alpha, and a calculation formula of a charging rate interval characteristic parameter, a discharging rate interval characteristic parameter, a charging state interval characteristic parameter, a working temperature interval characteristic parameter and a charging and discharging cycle frequency interval characteristic parameter, wherein a characteristic parameter vector of each new energy automobile working condition is constructed through the characteristic parameter, and meanwhile, a formula for calculating the similarity of the characteristic parameter vector is provided, and the formula not only considers the direction of the characteristic parameter vector, but also considers the length of the characteristic parameter vector. In addition, the classification result verification parameter CRVP is used for evaluating the classification effect of the real vehicle working conditions, and finally, the power battery health state values of different real vehicle working condition categories are obtained through point estimation or interval estimation. According to the invention, the real vehicle working conditions of the new energy automobile are classified into different categories, the health state of the power battery is determined according to the real vehicle working conditions, meanwhile, more accurate label values can be given to the health state of the power battery under different working conditions, and the label values can be used for machine learning of the health state to promote the development of a power battery health state assessment algorithm under the real vehicle working conditions.

Drawings

FIG. 1 is a general flow of a power battery health state evaluation method based on the actual vehicle condition of a new energy vehicle according to an embodiment of the invention;

FIG. 2 is a fitted straight line of capacity fade rate with cycle number for different charge rates according to an embodiment of the present invention;

FIG. 3 is a graph showing the fit of different charge rates to the single cycle capacity decay rate in accordance with an embodiment of the present invention;

fig. 4 is a diagram showing a health state attenuation coefficient α corresponding to different charging rate intervals according to an embodiment of the present invention;

FIG. 5 is a fitted straight line of capacity fade rate with cycle number for different discharge rates according to an embodiment of the present invention;

FIG. 6 is a graph showing the fit of different discharge rates to the single cycle capacity decay rate in accordance with an embodiment of the present invention;

FIG. 7 shows the health state attenuation coefficient α corresponding to different discharge rate intervals according to the embodiment of the present invention;

FIG. 8 is a fitted straight line of the capacity decay rate of different state of charge intervals according to the change of the equivalent charge-discharge cycle times in the embodiment of the invention;

FIG. 9 is a graph showing the state of health decay coefficients α corresponding to different state of charge intervals according to an embodiment of the present invention;

FIG. 10 is a graph showing the fit of the capacity fade rate with cycle number at different operating temperatures for an embodiment of the present invention;

FIG. 11 is a graph showing the fit of different operating temperatures to the rate of capacity decay for a single cycle in accordance with an embodiment of the present invention;

FIG. 12 is a graph showing the health state attenuation coefficient α corresponding to different operating temperature ranges according to an embodiment of the present invention;

FIG. 13 is a fitted straight line of capacity fade rate as a function of number of cycles for an embodiment of the present invention;

FIG. 14 is a block diagram of a hardware portion of an embodiment of the invention;

FIG. 15 is a graph showing a change of a charging rate with time and a division of a charging rate interval according to an embodiment of the present invention;

FIG. 16 is a graph showing the discharge rate over time and the division of the discharge rate intervals according to an embodiment of the present invention;

FIG. 17 is a plot of state of charge over time and a division of state of charge intervals for an embodiment of the present invention;

FIG. 18 is a graph showing the change of the operating temperature with time and the division of the operating temperature intervals according to the embodiment of the present invention.

Detailed Description

The invention is described in further detail below with reference to fig. 1-18.

Examples: as shown in FIG. 1, firstly, a power battery pack aging experiment with single working condition change is carried out, and health state attenuation coefficients alpha corresponding to different charging rate intervals, different discharging rate intervals, different state of charge intervals, different working temperature intervals and different charging and discharging cycle times intervals are determined. And secondly, calculating the characteristic parameters of the charging rate interval, the characteristic parameters of the discharging rate interval, the characteristic parameters of the charge state interval, the characteristic parameters of the working temperature interval and the characteristic parameters of the charging and discharging cycle times interval according to the actual vehicle working condition of each new energy automobile. And thirdly, constructing a characteristic parameter vector of each new energy automobile real vehicle working condition, calculating the similarity between the characteristic parameter vectors of any two new energy automobile real vehicle working conditions to obtain a similarity matrix, and classifying all the new energy automobile real vehicle working conditions into a plurality of categories by using a spectral clustering algorithm. Then, the available capacity of the power battery of each real vehicle working condition category is sampled and tested, and the classification effect is evaluated by using a classification result verification parameter CRVP. And finally, evaluating the power battery health state values of different real vehicle working condition categories through point estimation or interval estimation. Meanwhile, the power battery health status values of different real vehicle working condition categories can also be used as tag values to calibrate the power battery health status of the corresponding working condition categories, and the method is used for machine learning of real vehicle working condition health status assessment.

1. And (3) carrying out a charge-discharge cycle aging experiment of the power battery pack with single working condition change, and determining health state attenuation coefficients alpha corresponding to different charge rate intervals, different discharge rate intervals, different charge state intervals, different working temperature intervals and different charge-discharge cycle times intervals. The health state attenuation coefficient alpha is a parameter for measuring the attenuation speed of the health state, and the greater the value is, the faster the attenuation speed of the health state is, and the smaller the attenuation speed of the health state is otherwise.

And (I) respectively carrying out charge-discharge cycle aging experiments on different power battery packs at different charge multiplying powers and the same discharge multiplying power in a charge state interval of 0% to 100% at a constant working temperature, and determining health state attenuation coefficients alpha corresponding to the different charge multiplying power intervals. The specific implementation method is as follows:

at a constant operating temperature (e.g., room temperature 25 ℃) different power battery packs are charged at different charge rates (the maximum charge rate in the aging test is higher than the maximum charge rate in the real vehicle, and the minimum charge rate in the real vehicle is lower than the minimum charge rate in the real vehicle, e.g., the charge rates in the aging test are respectively selected from 0.1C, 0.5C, 1.0C, 1.5C, 2.0C, etc.) in the state of charge interval of 0% to 100%. After each charge is completed, the power battery pack is discharged at the same discharge rate (e.g., 0.1C). And calculating the actual discharge capacity of the power battery pack in a charge state interval of 0-100% by using an ampere-hour integration method every several charge-discharge cycles, and calculating the capacity attenuation rate according to the actual discharge capacity and the rated capacity. And carrying out charge and discharge cycles for a plurality of times until the actual discharge capacity of the power battery pack is attenuated to 80% of the rated capacity.

And marking the capacity attenuation rate corresponding to different charge and discharge cycle times under different charge rates in a coordinate system by taking the charge and discharge cycle times of the power battery pack as an X axis and the capacity attenuation rate as a Y axis. And fitting the capacity attenuation rates under the same charging multiplying power by using straight lines respectively to obtain fitting straight lines of which the capacity attenuation rates change along with the cycle times under different charging multiplying powers, as shown in fig. 2.

The slope of a fitted straight line, namely deltay/deltax, of the capacity fading rate with the change of the cycle times under different charging rates is obtained, and the physical meaning of the slope is the percentage of capacity fading after one charge and discharge cycle, and is called as the capacity fading rate of one cycle. The larger the single cycle capacity attenuation rate is, the faster the power battery health state is attenuated under the working condition; the smaller the opposite is.

And marking the single-cycle capacity attenuation rate corresponding to different charging rates in a coordinate system by taking the charging rate as an X axis and the single-cycle capacity attenuation rate as a Y axis, and fitting a function relationship between the different charging rates and the single-cycle capacity attenuation rate by using a curve, as shown in fig. 3.

The charging rate is divided into a plurality of sections at specific intervals (e.g., 0.25C), for example, [0C,0.25C, 0.5C ], and the like, an average value of the single-cycle capacity attenuation ratio for each section is calculated, and the average value is defined as a health state attenuation coefficient α for the charging rate section and is noted in a coordinate system, as shown in fig. 4.

And (II) respectively carrying out charge-discharge cycle aging experiments on different power battery packs at the same charge multiplying power and different discharge multiplying powers in a 0-100% charge state interval at a constant working temperature, and determining the health state attenuation coefficients alpha corresponding to the different discharge multiplying power intervals. The specific implementation method is as follows:

the power battery pack is charged at the same charge rate (e.g., 0.1C) in a state of charge interval of 0% to 100% at a constant operating temperature (e.g., room temperature 25 ℃). After each charge is completed, different power battery packs are discharged with different discharge rates (the maximum value of the discharge rate in the aging test is higher than the maximum value of the discharge rate in the real vehicle condition, and the minimum value is lower than the minimum value of the discharge rate in the real vehicle condition, for example, the discharge rates in the aging test are respectively selected from 0.1C, 0.5C, 1.0C, 1.5C, 2.0C and the like). And calculating the actual charge capacity of the power battery pack in a charge state interval of 0-100% by using an ampere-hour integration method every several charge and discharge cycles, and calculating the capacity attenuation rate according to the actual charge capacity and the rated capacity. The charge and discharge cycles are performed in this way until the actual charge capacity of the power battery pack decays to 80% of the rated capacity.

And marking the capacity attenuation rate corresponding to different charge and discharge cycle times in a coordinate system under different discharge multiplying factors by taking the charge and discharge cycle times of the power battery pack as an X axis and the capacity attenuation rate as a Y axis. And fitting the capacity attenuation rates under the same discharge multiplying power by using straight lines respectively to obtain fitting straight lines of which the capacity attenuation rates change along with the cycle times under different discharge multiplying powers, as shown in fig. 5.

The slope of a fitted straight line, namely deltay/deltax, of the capacity fading rate with the change of the cycle times under different discharge rates is obtained, and the physical meaning of the slope is the percentage of capacity fading after one charge-discharge cycle, and is called as the capacity fading rate of one cycle. The larger the single cycle capacity attenuation rate is, the faster the power battery health state is attenuated under the working condition; the smaller the opposite is.

And marking the single-cycle capacity attenuation rate corresponding to different discharge rates in a coordinate system by taking the discharge rate as an X axis and the single-cycle capacity attenuation rate as a Y axis, and fitting a function relationship between the different discharge rates and the single-cycle capacity attenuation rate by using a curve, as shown in fig. 6.

The discharge rate is divided into a plurality of sections at specific intervals (e.g., 0.25C), for example, [0C,0.25C, 0.5C ], and the like, and an average value of the single-cycle capacity attenuation ratio for each section is calculated, and the average value is defined as a health state attenuation coefficient α for the discharge rate section and is marked in a coordinate system as shown in fig. 7.

And thirdly, respectively carrying out charge-discharge cyclic aging experiments on different power battery packs in different charge state intervals at the same charge multiplying power and the same discharge multiplying power under the constant working temperature to determine the health state attenuation coefficient alpha corresponding to the different charge state intervals. The specific implementation method is as follows:

at a constant operating temperature (e.g., room temperature 25 ℃), different power cells are cycled between charge and discharge at a fixed charge rate (e.g., 0.1C) and a fixed discharge rate (e.g., 0.1C) over different state of charge intervals, respectively, [0%,20%, 40%, [40%, 60%, [60%, 80%, [80%, 100% ]. Every 100% charge-discharge cycle is accumulated, and the charge-discharge cycle is counted as an equivalent charge-discharge cycle. And calculating the equivalent discharge capacity (namely, the discharge capacity of 100% of accumulated discharge) of the power battery pack by using an ampere-hour integration method every several equivalent charge-discharge cycles, and calculating the capacity attenuation rate according to the equivalent discharge capacity and the rated capacity. And carrying out equivalent charge and discharge cycles for a plurality of times until the equivalent discharge capacity of the power battery pack is reduced to 80% of the rated capacity.

And marking the capacity attenuation rate corresponding to different equivalent charge and discharge cycle times in different charge state intervals in a coordinate system by taking the equivalent charge and discharge cycle times of the power battery pack as an X axis and the capacity attenuation rate as a Y axis. And fitting the capacity attenuation rates of the same charge state interval by using straight lines respectively to obtain fitting straight lines of the capacity attenuation rates of different charge state intervals changing along with the cycle times, as shown in fig. 8.

The slope of a fitted straight line, namely deltay/deltax, of the capacity-fade rate of different charge state intervals along with the change of the cycle times is obtained, and the physical meaning of the slope is the percentage of capacity-fade after one charge-discharge cycle, and is called as the capacity-fade rate of a single cycle. The larger the single cycle capacity attenuation rate is, the faster the power battery health state is attenuated under the working condition; the smaller the opposite is.

The single cycle capacity decay rate for each state of charge interval is a fixed value. The single-cycle capacity decay rate of each state of charge section is defined as the state of health decay coefficient alpha corresponding to that state of charge section. And the state of charge interval is taken as an X axis, the state of health attenuation coefficient alpha is taken as a Y axis, and the state of health attenuation coefficients alpha corresponding to different state of charge intervals are marked in a coordinate system, as shown in fig. 9.

And (IV) respectively carrying out charge-discharge cycle aging experiments on different power battery packs at different working temperatures in a state-of-charge interval of 0% to 100% with the same charge multiplying power and the same discharge multiplying power, and determining the health state attenuation coefficients alpha corresponding to different working temperature intervals. The specific implementation method is as follows:

at different operating temperatures (the maximum value of the operating temperature is higher than the maximum value of the operating temperature under the actual vehicle condition in the aging test, the minimum value is lower than the minimum value of the operating temperature under the actual vehicle condition in the aging test, for example, the operating temperature is respectively selected to be 25 ℃, 35 ℃, 45 ℃, 55 ℃ and the like), and different power battery packs are respectively charged at the same charging multiplying power (for example, 0.1C) and the same discharging multiplying power (for example, 0.1C) in the interval of 0% to 100% of the state of charge. And calculating the actual discharge capacity of the power battery pack in a charge state interval of 0-100% by using an ampere-hour integration method every several charge-discharge cycles, and calculating the capacity attenuation rate according to the actual discharge capacity and the rated capacity. And carrying out charge and discharge cycles for a plurality of times until the actual discharge capacity of the power battery pack is attenuated to 80% of the rated capacity.

And marking the capacity attenuation rate corresponding to different charge and discharge cycle times in a coordinate system at different working temperatures by taking the charge and discharge cycle times of the power battery pack as an X axis and the capacity attenuation rate as a Y axis. And fitting the capacity attenuation rates at the same working temperature by using straight lines respectively to obtain fitting straight lines of which the capacity attenuation rates change along with the cycle times at different working temperatures, as shown in fig. 10.

The slope of a fitted straight line, namely deltay/deltax, of the capacity attenuation rate with the change of the cycle times at different working temperatures is obtained, wherein the physical meaning is the percentage of capacity attenuation after one charge-discharge cycle, and the percentage is called as the capacity attenuation rate of one cycle. The larger the single cycle capacity attenuation rate is, the faster the power battery health state is attenuated under the working condition; the smaller the opposite is.

And marking the single-cycle capacity attenuation rate corresponding to different working temperatures in a coordinate system by taking the working temperature as an X axis and the single-cycle capacity attenuation rate as a Y axis, and fitting a function relation between the different working temperatures and the single-cycle capacity attenuation rate by using a curve. As shown in fig. 11.

The operating temperature is divided into a plurality of sections at specific intervals (e.g., 5 ℃) such as [25 ℃,30 ℃), [30 ℃,35 ℃), [35 ℃,40 ℃), [40 ℃,45 ℃), [45 ℃,50 ℃) and [50 ℃,55 ℃) and the like, the average value of the single-cycle capacity attenuation rate corresponding to each section is calculated, the average value is defined as the health state attenuation coefficient alpha corresponding to the operating temperature section, and the health state attenuation coefficient alpha is marked in a coordinate system, as shown in fig. 12.

And fifthly, carrying out a power battery pack charge-discharge cycle aging experiment at a constant charge rate and a constant discharge rate in a state of charge interval of 0% to 100% at a constant working temperature, and determining health state attenuation coefficients alpha corresponding to different charge-discharge cycle frequency intervals. The specific implementation method is as follows:

the power battery pack is subjected to charge-discharge cycles at a constant charge rate (e.g., 0.1C) and a constant discharge rate (e.g., 0.1C) in a region of 0% to 100% state of charge at a constant operating temperature (e.g., room temperature 25 ℃). And calculating the actual discharge capacity of the power battery pack in a charge state interval of 0-100% by using an ampere-hour integration method every several charge-discharge cycles, and calculating the capacity attenuation rate through the actual discharge capacity and the rated capacity. And carrying out charge and discharge cycles for a plurality of times until the actual discharge capacity of the power battery pack is reduced to 80% of the rated capacity.

And marking the capacity attenuation rate corresponding to different charge and discharge cycle times in a coordinate system by taking the charge and discharge cycle times of the power battery pack as an X axis and the capacity attenuation rate as a Y axis. Fitting the capacity attenuation rates corresponding to different charge and discharge cycle times by using a straight line to obtain a fitting straight line of the capacity attenuation rate changing along with the charge and discharge cycle times, as shown in fig. 13.

The slope of a fitted straight line of the capacity fade rate with the number of charge and discharge cycles, namely Δy/Δx, is obtained, and its physical meaning is the percentage of capacity fade after one charge and discharge cycle, and is referred to herein as the single-cycle capacity fade rate. The larger the single cycle capacity attenuation rate is, the faster the power battery health state is attenuated under the working condition; the smaller the opposite is.

The number of charge/discharge cycles is divided into a plurality of sections at specific intervals (for example, 50 times), for example, [0,50 ], [50,100 ], [1950,2000 ], etc., an average number of charge/discharge cycles in each section (for example, an average number of 0 to 50 sections is 25) is calculated, the average number of charge/discharge cycles in each section is multiplied by a single cycle capacity attenuation ratio, and the calculated average number of charge/discharge cycles is used as a health state attenuation coefficient α corresponding to the charge/discharge cycle section, and is shown in table 1.

TABLE 1 health State attenuation coefficient alpha corresponding to different charge and discharge cycle intervals

2. Real vehicle working condition analysis and characteristic parameter determination of new energy vehicle

The actual vehicle condition of the new energy automobile is complex, for example, whether the new energy automobile is charged by slow charge or fast charge, the frequency and depth of a brake pedal, the ratio of high-speed and low-speed running, the frequency and depth of an accelerator pedal, the charging and discharging depth, the battery temperature, the accumulated running mileage and the like. According to the influence on the health state of the power battery, the above real vehicle working conditions can be summarized into five problems:

First, a charging rate interval. The essential difference between slow charging and fast charging is that the charging currents are different in magnitude, namely the current is distributed differently in different charging rate intervals; the new energy automobile generally performs braking energy feedback in the braking process, and is also a charging process for the power battery. Therefore, the frequency and depth of selecting slow charge or fast charge for charging and the brake pedal can be summarized as the problem of the charging rate interval. Second, a discharge rate interval. The power battery runs at high speed and low speed, and the discharging current of the power battery is different in magnitude, namely the current is distributed differently in different discharging multiplying power intervals; the vehicle acceleration process is a high current discharge process for the power battery. Therefore, the ratio of high-speed and low-speed traveling, the frequency and depth of the accelerator pedal can be summarized as the problem of the discharge rate interval. Third, state of charge interval. The charge-discharge depth can be generalized to the problem of charge state distribution among different charge state intervals. Fourth, an operating temperature interval. The battery temperature is summarized as the distribution problem of the working temperature of the power battery in different working temperature ranges. Fifth, charge-discharge cycle number interval. The accumulated driving mileage of the vehicle can be summarized as the problem of the charge and discharge cycle time interval of the power battery.

Therefore, the method classifies the actual vehicle conditions of the new energy vehicles according to the charging rate interval characteristic parameter, the discharging rate interval characteristic parameter, the state of charge interval characteristic parameter, the working temperature interval characteristic parameter and the charging and discharging cycle number interval characteristic parameter of each new energy vehicle.

The hardware part of the new energy automobile real-vehicle working condition classification for the power battery health state is shown in fig. 14. The current sensor is used for monitoring the charging current and the discharging current. The values of the charge current and the discharge current may be used to determine a charge rate interval and a discharge rate interval, and to evaluate the state of charge. The voltage sensor is used for monitoring the working voltage. The value of the operating voltage may be used to evaluate the state of charge. The temperature sensor is used for monitoring the working temperature. The value of the operating temperature may be used to determine the operating temperature interval and to evaluate the state of charge. The networking module is used for uploading and downloading data. The control module is responsible for controlling data sampling and interaction with the cloud server. The cloud server is used for calculating characteristic parameters of the operation conditions, calculating a similarity matrix, classifying the operation conditions, evaluating classifying effects and the like.

3. Calculation of real vehicle working condition characteristic parameters of new energy vehicle

Calculation of charging Rate Interval characteristic parameters

The charging current of the power battery is monitored in real time by a current sensor by setting an appropriate current sampling frequency (e.g., 10 Hz). Meanwhile, the charging rate is calculated according to the ratio of the charging current to the rated capacity, and a change curve of the charging rate along with time is obtained. Wherein, the time in the horizontal axis of the curve only accumulates the charge time, and does not accumulate the discharge time. The charge rates were divided into different intervals, for example, [0C,0.25C ], [0.25C,0.5C ], [1.75C, 2C), etc., at specific intervals (consistent with the power battery charge-discharge cycle aging experiments at different charge rates), as shown in fig. 15.

The ratio of the duration of each charging rate interval to the total charging time was calculated as shown in table 2.

TABLE 2 ratio of duration of each Charge Rate interval to total Charge time

Charging rate interval	[0C,0.25C)	[0.25C,0.5C)	...	[1.75C,2C)	...
						Proportion of occupied	25％	8.3％	...	3.3％	...

The characteristic parameter of the charging rate interval is the sum of the products of the health state attenuation coefficient alpha corresponding to each charging rate interval and the duration time of the corresponding charging rate interval accounting for the whole charging time proportion, and the calculation formula is that

In the formula, i refers to an ith charging rate interval, and a health state attenuation coefficient alpha corresponding to the charging rate interval i can be queried through the charge-discharge cycle aging experimental results of the power battery pack under different charging rates, as shown in fig. 4. n refers to the total number of charging rate intervals.

(II) calculation of discharge multiplying power interval characteristic parameters

The discharge current of the power battery is monitored in real time by a current sensor by setting an appropriate current sampling frequency (e.g., 10 Hz). Meanwhile, the discharge rate is calculated according to the ratio of the discharge current to the rated capacity, and a curve of the change of the discharge rate along with time is obtained. Wherein, the time in the horizontal axis of the curve only accumulates the discharge time, and does not accumulate the charge time. The discharge rates were divided into different intervals, for example, [0C,0.25C ], [0.25C,0.5C ], [1.75C, 2C), etc., at specific intervals (consistent with the power battery pack charge-discharge cycle aging experiments at different discharge rates), as shown in fig. 16.

The ratio of the duration of each discharge rate interval to the total discharge time was calculated as shown in table 3.

TABLE 3 ratio of duration of each discharge Rate interval to total discharge time

Discharge rate interval	[0C,0.25C)	[0.25C,0.5C)	...	[1.75C,2C)	...
						Proportion of occupied	15％	23％	...	1％	...

The characteristic parameter of the discharge rate interval is the sum of the products of the health state attenuation coefficient alpha corresponding to each discharge rate interval and the ratio of the duration time of the corresponding discharge rate interval to the total discharge time, and the calculation formula is that

In the formula, j refers to a j-th discharge rate interval, and the health state attenuation coefficient alpha corresponding to the discharge rate interval j can be queried through the charge-discharge cycle aging experimental results of the power battery pack under different discharge rates, as shown in fig. 7. r denotes the total number of discharge rate intervals.

(III) calculation of State of Charge section characteristic parameters

The voltage, the current and the temperature of the power battery are monitored in real time through a voltage sensor, a current sensor and a temperature sensor by setting proper voltage sampling frequency (for example, 0.1 Hz), current sampling frequency (for example, 10 Hz) and temperature sampling frequency (for example, 0.1 Hz), and the state of charge of the power battery is estimated. Regarding the evaluation of the state of charge of the power battery, there have been many studies made by researchers, for example, the state of charge of the power battery can be evaluated by an ampere-hour integration method, an open-circuit voltage method, a kalman filter method, etc., and will not be described in detail herein. Finally, a curve of the change of the charge state along with time is obtained. The states of charge are divided into different intervals (e.g., [0%, 20%), [20% -40%), [40% -60%), [60% -80%), [80% -100% ] at specific intervals (consistent with the power battery charge-discharge cycle aging experiments for the different state of charge intervals), as shown in fig. 17.

The ratio of the duration of each state of charge interval to the total charge-discharge time was calculated as shown in table 4.

TABLE 4 ratio of duration of each State of charge interval to total charge/discharge time

State of charge interval	[0％,20％)	[20％—40％)	[40％—60％)	[60％—80％)	[80％—100％]
						Proportion of occupied	0％	0％	63.3％	36.7％	0％

The characteristic parameter of the state of charge interval is the sum of the products of the state of health attenuation coefficient alpha corresponding to each state of charge interval and the duration time of the corresponding state of charge interval accounting for the proportion of all charge and discharge time, and the calculation formula is

/>

In the formula, k refers to a kth state of charge section, and the state of health attenuation coefficient alpha corresponding to the state of charge section k can be queried through the charge-discharge cycle aging test results of the power battery packs in different state of charge sections, as shown in fig. 9. s refers to the total number of state of charge intervals.

(IV) calculation of characteristic parameters of working temperature interval

And setting a proper temperature sampling frequency (for example, 0.1 Hz), and monitoring the working temperature of the power battery in real time through a temperature sensor to obtain a curve of the change of the working temperature along with time. The operating temperatures are divided into different sections, for example, 25 c, 30 c, 35 c, 40 c, 45 c, 50 c, 55 c, etc., at specific intervals (consistent with the charge-discharge cycle aging experiments of the power battery pack at different operating temperatures, for example, 5 c), as shown in fig. 18.

The ratio of the duration of each working interval to the total charge-discharge time was calculated as shown in table 5.

TABLE 5 proportion of duration of each operating temperature interval to total charge/discharge time

Working temperature interval

[25℃,30℃)

[30℃,35℃)

...

[45℃,50℃)

[50℃,55℃)

...

Proportion of occupied

16.7％

18.3％

...

23.3％

3.3％

...

The characteristic parameter of the working temperature interval is the sum of the products of the health state attenuation coefficient alpha corresponding to each working temperature interval and the proportion of the duration time of the corresponding working temperature interval to the total charge and discharge time, and the calculation formula is that

In the formula, q refers to a q-th working temperature interval, and the health state attenuation coefficient alpha corresponding to the working temperature interval q can be queried through the results of the charge-discharge cycle aging experiments of the power battery pack at different working temperatures, as shown in fig. 12. t refers to the total number of operating temperature intervals.

(V) calculation of characteristic parameters of charge-discharge cycle number interval

And setting a proper current sampling frequency (for example, 10 Hz), monitoring the discharge current of the power battery in real time through a current sensor, and integrating the current in time to obtain the accumulated discharge capacity. The ratio of the accumulated charge-discharge capacity to the rated capacity is used as the charge-discharge cycle number. The interval of the charge-discharge cycle times is found in the charge-discharge cycle aging experimental result of the power battery pack (shown in table 1), the health state attenuation coefficient alpha is obtained and is used as the characteristic parameter of the charge-discharge cycle times interval, and the calculation formula is as follows

4. Calculation of characteristic parameter vector similarity matrix of actual vehicle working condition of new energy automobile

Calculating various characteristic parameters and median of the actual vehicle working conditions of the new energy vehicle

And calculating the characteristic parameters of the charging rate interval, the discharging rate interval, the charge state interval, the working temperature interval and the charge and discharge cycle number interval of each new energy automobile, and obtaining the median of each characteristic parameter, as shown in table 6.

TABLE 6 real vehicle operating parameters and median for New energy vehicles

(II) constructing characteristic parameter vectors of actual vehicle working conditions of new energy vehicles

Subtracting the corresponding median from the characteristic parameters of the charging rate interval, the discharging rate interval, the charge state interval, the working temperature interval and the charge-discharge cycle number interval of each new energy automobile to serve as elements in the characteristic parameter vector of the actual automobile working condition of the new energy automobile. For example, if the respective characteristic parameters of the vehicle 1 are 0.6, 0.7, 0.9, 0.4, and 0.5, respectively, and the respective median values of all the respective characteristic parameters of the vehicle are 0.4, 0.5, 0.6, 0.5, and 0.5, respectively, the characteristic parameter vector of the vehicle 1 is [0.2,0.2,0.3, -0.1, and 0].

(III) calculating the similarity AS in the vector direction of the two characteristic parameters

And calculating the similarity between the characteristic parameter vectors of the actual vehicle working conditions of any two new energy vehicles. Let a and b be the characteristic parameter vectors of two new energy automobile working conditions respectively, first need to calculate the similarity COS in two vector directions, its formula is

Wherein a.b represents the dot product of vector a and vector b, the terms a and b represent norms of the vectors a and b (i.e., lengths of the vectors). The COS value range is between-1 and 1. Mapping the value of COS to between [0,1] to obtain AS, wherein the calculation formula is

AS values range between [0,1] and gradually tend towards 1 AS the angle between the two vectors decreases, and conversely towards 0.

(IV) calculating the similarity LS of the lengths of the two characteristic parameter vectors

Calculating the similarity LS of the vector a and the vector b in length, wherein the calculation formula is as follows

Wherein a and b are absolute values of differences between two vector norms a and b, and max (|a and b) represents a larger value of a and b.

LS values range between 0,1 and gradually tend to 1 as the difference between the two vector lengths decreases, and conversely tend to 0.

(V) calculating the similarity SIM of the two feature parameter vectors

The similarity SIM of the feature parameter vector is the product of the similarity AS of two feature parameters in the vector direction and the similarity LS in length, i.e

The SIM value ranges between 0,1 and gradually goes to 1 as the similarity of the two vectors increases, whereas it goes to 0.

(VI) constructing a similarity matrix

And calculating the similarity between the characteristic parameter vectors of the actual vehicle working conditions of any two new energy vehicles to obtain a similarity matrix, as shown in table 7.

TABLE 7 similarity matrix

	Vehicle 1	Vehicle 2	……	Vehicle w-1	Vehicle w
						Vehicle 1	1	0.7	……	0.6	0.3
Vehicle 2	--	1	……	0.5	0.9
						……	--	--	1	0.3	0.5
Vehicle w-1	--	--	--	1	0.6
						Vehicle w	--	--	--	--	1

5. Classifying real vehicle conditions using spectral clustering algorithms

And according to the similarity matrix, using a spectral clustering algorithm to divide all the new energy automobiles into a plurality of actual automobile working condition categories. The spectral clustering algorithm is a mature algorithm and is not described here in detail.

6. Sampling test of available capacity of power battery of each real vehicle working condition type

And regularly extracting a certain number of automobiles (during the period that the automobiles enter a 4S shop for maintenance or repair) from the new energy automobiles in each real automobile working condition category as test samples, and carrying out complete offline charge-discharge circulation in a 0-100% charge state interval under the rated working condition of specific charge-discharge current and working temperature on the premise of proving user consent to obtain accurate available capacity values.

7. Evaluation of Classification Effect

The health states of the power batteries of the same real vehicle working condition type should be relatively close, that is to say, the available capacities of the power batteries of the same real vehicle working condition type should be relatively close, and the available capacities of the power batteries of the new energy vehicles of other real vehicle working condition types are remarkably different. Therefore, the available capacity distribution of the power batteries of the new energy vehicles in various real vehicle working condition categories is counted, whether the available capacity distribution in the same real vehicle working condition category is concentrated in a certain section or not is judged, and whether the available capacity distribution in the same real vehicle working condition category is remarkably different from the available capacity distribution in other working condition categories or not is judged, so that the classification effect is evaluated.

Determining the effective interval of the available capacity of the power battery of each real vehicle working condition category obtained by sampling test

In order to remove outliers, the available capacity of the power battery of the new energy automobile of each real automobile working condition category obtained by sampling test is arranged in a sequence from small to large, the beta% quantile and the (1-beta%) quantile (for example, 10% quantile and 90% quantile) of the power battery are determined, and the interval between the beta% quantile and the (1-beta%) quantile is called as an effective interval. The length of the effective interval is denoted by interval.

(II) calculating the sum of maximum values among the overlapping lengths of each effective interval and other effective intervals

Assuming that after classification, m real vehicle working condition categories are provided, namely m effective intervals are provided, two effective intervals are taken from the m effective intervals, and the overlapping length overlap of the two effective intervals is calculated. Assuming that the first effective interval is [ X1, X2], the second effective interval is [ Y1, Y2], and the calculation formula of the overlap length overlap of the two effective intervals is

Wherein,

the derivation of this formula is briefly described as follows:

under the condition that two effective sections are overlapped, subtracting a larger value of a left end point from a smaller value of a right end point of the two effective sections to obtain an overlapping length A, wherein the calculation formula is as follows

In the case where the two effective sections do not overlap, the overlap length of the two effective sections should be 0, however, a obtained by the above formula is a negative value. Therefore, the formula for calculating the overlap length overlap of two effective sections is given by taking into consideration whether or not the two effective sections overlap

Wherein, overlap _p Is the maximum value among the overlapping lengths of the p-th effective interval and the other m-1 effective intervals.

(III) calculating the sum of all the effective interval lengths

Calculating the sum of the lengths of m effective intervals as

Wherein, interval _p Is the length of the p-th active interval.

(IV) calculating a classification result verification parameter CRVP

Defining a classification result verification parameter CRVP as a parameter for evaluating classification effect, wherein the calculation formula of the CRVP is as follows

The value of CRVP is between [0,1], the better the classification effect of working condition classification is, the closer the value of CRVP is to 1, and the closer to 0 is the other way around.

And setting a certain CRVP threshold value, and judging the classification effect. If the classification effect cannot meet the requirement, the classification algorithm is adjusted again; if the classification is good, the power battery state of health value of each real vehicle working condition class is further estimated.

8. Estimation of health state values

The power cell state of health value for each real vehicle condition category can be estimated by two methods:

(one) Point estimation

The expected value of the available capacity of the power battery in each real vehicle working condition category is the average value of the available capacities of all the power battery samples sampled and tested in the categoryThe desire for health status is then calculated from the ratio of the desire for available capacity to the rated capacity.

(two) section estimation

The confidence interval for the power battery available capacity of each real vehicle working condition category is 1-gamma

In addition, the power battery health status values of different real vehicle working condition categories can also be used as tag values to calibrate the power battery health status of the corresponding working condition categories, and the method is used for machine learning of real vehicle working condition health status assessment.

The invention classifies the real vehicle working conditions of the new energy automobile into a plurality of categories, and evaluates the health state of the power battery of each category. The following functions are realized:

(1) A power battery pack aging experiment with single working condition change is carried out, and a calculation method of a health state attenuation coefficient alpha is provided, and is used for determining the health state attenuation coefficient alpha corresponding to different charging multiplying power intervals, different discharging multiplying power intervals, different charging state intervals, different working temperature intervals and different charging and discharging cycle times intervals.

(2) And a formula for calculating the characteristic parameters of the charging rate interval, the discharging rate interval, the charge state interval, the working temperature interval and the charging and discharging cycle times according to the actual vehicle working condition of each new energy automobile is provided.

(3) The characteristic parameter vector of each new energy automobile working condition is constructed through the characteristic parameters, and a formula for calculating the similarity of the characteristic parameter vector is provided, wherein the formula not only considers the direction of the characteristic parameter vector, but also considers the length of the characteristic parameter vector. And calculating the similarity between the characteristic parameter vectors of the actual vehicle working conditions of any two new energy vehicles by using a formula to obtain a similarity matrix, and classifying all the actual vehicle working conditions of the new energy vehicles into a plurality of categories by using a spectral clustering algorithm.

(4) And sampling and testing the available capacity of the power battery of each real vehicle working condition type, and evaluating the classification effect by using a classification result verification parameter CRVP.

(5) And obtaining the power battery health state values of different real vehicle working condition categories through point estimation or interval estimation.

(6) The power battery health status values of different real vehicle working condition categories can also be used as tag values to calibrate the power battery health status of the corresponding working condition categories, and the method is used for machine learning of real vehicle working condition health status assessment.

The present embodiment is only for explanation of the present invention and is not to be construed as limiting the present invention, and modifications to the present embodiment, which may not creatively contribute to the present invention as required by those skilled in the art after reading the present specification, are all protected by patent laws within the scope of claims of the present invention.

Claims

1. The power battery health state evaluation method based on the real vehicle working condition of the new energy automobile is characterized by comprising the following steps of:

s5: sampling and testing the available capacity of the power battery of each real vehicle working condition type;

s6: evaluation of classification effect;

2. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: step S1, carrying out a charge-discharge cycle aging experiment of a power battery pack with single working condition change, respectively fitting the relationship between the capacity attenuation rate and the charge-discharge cycle times under the same working condition by using straight lines, then fitting the functional relationship between the working condition and the single cycle capacity attenuation rate by using curves, finally dividing the working condition into a plurality of sections at specific intervals, calculating the average value of the single cycle capacity attenuation rate corresponding to each working condition section, and defining the average value as a health state attenuation coefficient alpha corresponding to the working condition section so as to determine health state attenuation coefficients alpha corresponding to different charging multiplying power sections, different discharging multiplying power sections and different working temperature sections; fitting the relationship between the capacity attenuation rate and the equivalent charge-discharge cycle times of the same charge state interval by using straight lines respectively, and obtaining the slope of each fitting straight line to serve as a health state attenuation coefficient alpha corresponding to each charge state interval; fitting capacity attenuation rates corresponding to different charge and discharge cycle times by using a straight line to obtain a single cycle capacity attenuation rate, dividing the charge and discharge cycle times into a plurality of sections at specific intervals, calculating the average number of the charge and discharge cycle times in each section, and multiplying the average number of the charge and discharge cycle times in each section by the single cycle capacity attenuation rate to obtain a health state attenuation coefficient alpha corresponding to the charge and discharge cycle time section; the health state attenuation coefficient alpha is a parameter for measuring the attenuation speed of the health state, and the greater the value is, the faster the attenuation speed of the health state is, and the smaller the attenuation speed of the health state is otherwise.

3. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: in the step S2, the characteristic parameter of the charging rate interval is the sum of products of the health state attenuation coefficient α corresponding to each charging rate interval and the duration time of the corresponding charging rate interval accounting for the proportion of all charging time, and the calculation formula is as follows:

in the above formula, i refers to the ith charging rate interval, and n refers to the total number of charging rate intervals;

in the above formula, j refers to the j-th discharge rate interval, and r refers to the total number of discharge rate intervals;

in the above formula, k refers to the kth state of charge interval, and s refers to the total number of the state of charge intervals;

In the above formula, q refers to the q-th working temperature interval, and t refers to the total number of working temperature intervals;

charge-discharge cycle number interval characteristic parameter = health state attenuation coefficient alpha corresponding to the interval in which the charge-discharge cycle number is located.

4. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: step S3 is to construct a characteristic parameter vector of the working condition of the real vehicle and calculate a characteristic parameter vector similarity matrix, and the steps are as follows:

wherein, a and b are absolute values of differences between two vector norms a and b, and max (a and b) represents a larger value of a and b; LS values range between [0,1] and gradually tend to 1 as the difference between the two vector lengths decreases, whereas tend to 0;

5. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: and S5, regularly extracting a certain number of automobiles from the new energy automobiles in each real automobile working condition category as test samples, and carrying out complete offline charge-discharge circulation in a 0-100% charge state interval under the rated working condition of specific charge-discharge current and working temperature to obtain an accurate available capacity value.

6. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: step S6, because the power batteries of the same real vehicle working condition category are similar in health status and are obviously different from the available capacities of the power batteries of the new energy vehicles of other real vehicle working condition categories; therefore, the available capacity distribution of the power batteries of the new energy automobiles in various real automobile working condition categories is counted, whether the available capacity distribution of the same real automobile working condition category is concentrated in a certain section or not is judged, and whether the available capacity distribution of the same real automobile working condition category is remarkably different from the available capacity distribution of other working condition categories or not is judged, so that the classification effect is evaluated; the steps flow as follows:

A. Determining the effective interval of the available capacity of the power battery of each real vehicle working condition type obtained by sampling test; the available capacity of the power battery of the new energy automobile of each real automobile working condition category obtained by sampling test is arranged in order from small to large, the beta percent fraction and the (1-beta percent) fraction are determined, the interval between the beta percent fraction and the (1-beta percent) fraction is called an effective interval, and the interval is used for representing the length of the effective interval;

B. calculating the sum of maximum values among the overlapping lengths of each effective interval and other effective intervals; after classification, m real vehicle working condition categories are provided, namely m effective intervals are provided, two effective intervals are selected, the first effective interval is [ X1, X2], and the second effective interval is [ Y1, Y2];

in the case that the two effective sections do not overlap, the overlapping length overlap of the two effective sections is 0;

therefore, the calculation formula of the overlap length of the two effective sections is:

for each effective interval, traversing and calculating the overlapping length of the effective interval and other m-1 effective intervals to obtain the maximum value among the overlapping lengths of the effective interval and other m-1 effective intervals; then, the sum of the maximum values among the overlapping lengths of all m effective sections and other effective sections is calculated, namely

wherein, interval _p A length interval of the p-th effective interval;

7. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: the point estimation method in the step S7 is that the expectation of the available capacity value of the power battery in each real vehicle working condition class is the average value of the available capacities of the power batteries in all sampling tests in the class; the desire for health status is then calculated from the ratio of the desire for available capacity to the rated capacity.

8. The power battery health state evaluation method based on the real vehicle working condition of the new energy vehicle according to claim 1, wherein the method is characterized by comprising the following steps: the interval estimation method in step S7 is that the confidence interval with the expected confidence of the available capacity of the power battery of each real vehicle working condition type is 1-gamma:

wherein,the average value of the available capacity of the power battery samples for the class sampling test is that S is the standard deviation of the samples, and n is the number of the samples; a confidence interval for the health expectations is then calculated from the ratio of the expected available capacity to the rated capacity.

9. The method for evaluating the health state of the power battery based on the actual vehicle working condition of the new energy automobile according to claim 8, wherein the method is characterized by comprising the following steps: and (3) taking the power battery health state values of different real vehicle working condition categories as tag values, calibrating the power battery health states of the corresponding working condition categories, and using the power battery health state values for machine learning of real vehicle working condition health state evaluation.