Background art:
driven by market and policy, in recent years, the electric automobile industry is rapidly developed, the demand of the power battery is greatly increased, the service life of the power battery is generally about 3-5 years, the power battery needs to be replaced when the capacity is reduced to about 80%, and the retirement peak period of the power battery in China is coming. The retired power battery can still be continuously used in a scene with low energy and power requirements, is called as echelon utilization, and has important significance for fully utilizing the economic value of the battery and relieving the environmental protection pressure. Therefore, the research on the power battery health state estimation in the running state of the electric automobile and the screening and classifying technology after the power battery is retired is very important for improving the safety and deep and reasonable gradient utilization of the power battery. English of battery health state: state of health, abbreviation: SOH.
Currently, for the health assessment of the power battery, there are mainly experimental analysis methods such as a direct discharge method, a method for measuring internal impedance, an electrochemical impedance analysis method and the like, modern estimation methods such as kalman filtering, fuzzy logic and the like, and data-driven regression prediction methods such as a support vector machine, a neural network and the like. The experimental analysis method is complex to operate, needs offline test and has low evaluation timeliness; the modern estimation method can carry out qualitative and quantitative analysis on the power battery, but accumulated errors may exist and are influenced by artificial experience; in the regression method based on data driving, the function form influence of the model is large, and meanwhile, a large amount of data support is needed by more model parameters and the optimization process so as to prevent the problems of over-fitting, under-fitting and the like. A new evaluation method is needed.
The invention content is as follows:
the invention provides a method for quickly estimating SOH (state of health) of a power battery based on kmeans clustering, which aims to simplify the evaluation process of the state of health of the energy storage battery, reduce the influence of human factors, get rid of the dependence of a regression model on a function form and data capacity, improve the timeliness of the state of health estimation technology of the power battery to be detected and realize the quick estimation of the state of health of the power battery. The technical scheme adopted by the invention is as follows:
a method for quickly estimating SOH of a power battery based on kmeans clustering comprises the following steps:
step 1: constructing a sample feature set; the method comprises the following specific steps:
step 1.1: measuring and analyzing the health state values of the power batteries of n different models in different health states by using an experimental analysis method, wherein the health state value of the ith power battery is
Step 1.2: carrying out one-time charge and discharge experiment on n power batteries, measuring voltage, current and temperature signals in the experiment process, wherein the voltage, the current and the temperature of the ith power battery are respectively V(i)、I(i)And T(i),i=1,2,…,n;
Step 1.3: extracting key features of the voltage, current and temperature signals and normalizing to form an m-dimensional feature vector, wherein the feature vector of the ith power battery
Forming a sample characteristic set by n power battery characteristics and state of health values
Step 2: clustering a sample set by using kmenas to form a plurality of ethnic groups; setting a clustering quantity parameter K, utilizing a kmeans clustering method and an m-dimensional characteristic vector X
(i)The samples in set A are grouped into K groups, each group being centered on
And step 3: calculating the distance from the sample to be detected to the center of each group, and judging the group to which the sample belongs; the method comprises the following specific steps:
step 3.1: measuring voltage V of power battery to be detected in one-time charging and discharging experiment process
(test)Current I
(test)And a temperature signal T
(test)Extracting the characteristic vector of the power battery to be detected
Step 3.2: calculating characteristic vector X of power battery to be detected
(test)To the center of each group
European distance of
Judging that the power battery to be detected belongs to a group closest to the center of each group as R, wherein R belongs to {1, 2.., K };
and 4, step 4: calculating the weight of the health state value of each sample in the group to which the power battery to be tested belongs; the method comprises the following specific steps:
step 4.1: calculating characteristic vector X of power battery to be detected(test)Euclidean distances to all samples in the population R, wherein the distance to the s-th sample in the population RRDistance of individual samples:
s
R=1,2,…,S
R,S
Rrepresenting movement in the group RThe number of force cell samples;
step 4.2: calculating to obtain the s
RPower battery sample state of health value
Normalized weight of
And 5: calculating the health state value of the power battery to be measured to form rapid estimation;
calculating the health state value of the power battery to be detected based on a weighted average mode:
and forming a health state estimation result of the power battery to be detected, and finishing test estimation.
Preferably, the step 2 comprises the following specific steps:
step 2.1: setting a clustering quantity parameter K, and randomly selecting K power battery samples from n power battery samples to form initialized K ethnic group centers;
step 2.2: calculating Euclidean distances between the feature vectors of the rest power battery samples and the centers of the K groups, wherein the Euclidean distance from the r-th sample to the center of the K group is
Judging the distance from each sample to the center of the Kth population, and assigning the distance to the nearest population;
step 2.3: updating the population centers according to the sample conditions in each population:
wherein the jth feature of the kth population center:
S
krepresents the number of samples in the kth population;
step 2.4: and (3) judging whether the clustering process meets the requirement, if the change degree of the center of each group is smaller than a set threshold value, if so, entering the step 3, and if not, returning to the step 2.2.
Compared with the closest prior art, the invention has the beneficial effects that:
in the technical scheme, the existing historical power battery samples similar to the power battery to be detected are screened out by means of a kmeans clustering mode, the similarity degree of the existing historical power battery samples is represented by the distance, the weight of the existing historical power battery health state is further obtained, and a basis is provided for evaluation and calculation of the power battery health state. Compared with the power battery health state evaluation method under the existing regression prediction model, the method provided by the invention has the advantages that the function form of the regression model is not required to be defined by prior information, the parameters required to be set are few, the requirement on the sample capacity is low, and the problems of over-fitting or under-fitting and the like in the power battery health state evaluation modeling process are favorably avoided.
In the technical scheme of the invention, the power battery health state estimation method is represented as a weighted average process, and the weight calculation only depends on key historical samples of similar groups in the kmeans clustering result, so that the influence of irrelevant historical samples on the evaluation result is avoided, the calculation amount of the power battery health state estimation is reduced, and the simple and effective calculation process is beneficial to the quick estimation of the power battery health state and the quick screening of the power battery gradient application.
The specific implementation mode is as follows:
example (b):
the technical solution in the embodiments of the present application will be clearly and completely described below with reference to fig. 1 to 3.
A method for quickly estimating SOH of a power battery based on kmeans clustering comprises the following steps:
step 1: acquiring historical power battery data, extracting key features and constructing a sample feature set; the method comprises the following specific steps:
step 1.1: measuring and analyzing the health state values of the power battery under n different health states by using an experimental analysis method such as direct discharge and the like, wherein the health state value of the ith power battery is
Step 1.2: carrying out one-time charge and discharge experiment on n power batteries, measuring voltage, current and temperature signals in the experiment process, wherein the voltage, the current and the temperature of the ith power battery are respectively V(i)、I(i)And T(i),(i=1,2,…,n);
Step 1.3: extracting key features of the voltage, current and temperature signals and normalizing to form an m-dimensional feature vector, wherein the feature vector of the ith power battery
Forming a sample characteristic set by n power battery characteristics and state of health values
Step 2: clustering a sample set by using kmenas to form a plurality of ethnic groups; the method comprises the following specific steps:
step 2.1: setting a clustering quantity parameter K, and randomly selecting K power battery samples from n power battery samples to form initialized K ethnic group centers;
step 2.2: calculating Euclidean distances between the feature vectors of the rest power battery samples and the centers of the K groups, wherein the Euclidean distance from the r-th sample to the center of the K group is
Judging the distance from each sample to the center of the Kth population, and assigning the distance to the nearest population;
step 2.3: updating the population centers based on the sample condition in each population
Wherein the jth feature of the kth population center is
S
kRepresents the number of samples in the kth population;
step 2.4: judging whether the clustering process meets the requirement, if the change degree of the center of each group is smaller than a set threshold value, if so, entering the step 3, otherwise, returning to the step 2.2;
and step 3: calculating the distance from the sample to be detected to the center of each group, and judging the group to which the sample belongs; the method comprises the following specific steps:
step 3.1: measuring voltage V of power battery to be detected in one-time charging and discharging experiment process
(test)Current I
(test)And a temperature signal T
(test)Extracting the characteristic vector of the power battery to be detected
Step 3.2: calculating characteristic vector X of power battery to be detected
(test)To the center of each group
European distance of
Judging that the power battery to be detected belongs to a group closest to the center of each group as R, wherein R belongs to {1, 2.., K };
and 4, step 4: calculating the weight of the health state value of each sample in the group to which the power battery to be tested belongs; the method comprises the following specific steps:
step 4.1: calculating characteristic vector X of power battery to be detected
(test)Euclidean distances to all samples in the population R, wherein the distance to the s-th sample in the population R
RA distance of one sample being
S
RRepresenting the number of power cell samples in the population R;
step 4.2: calculating to obtain the s
RPower battery sample state of health value
Normalized weight of
Forming weight vectors
And 5: calculating the health state value of the power battery to be measured to form rapid estimation;
calculating the state of health value of the power battery to be detected based on a weighted average mode
And forming a health state estimation result of the power battery to be detected, and finishing test estimation.
Finally, it should be noted that: the described embodiments are only some embodiments of the present application and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.