CN116794518A - Method and system for predicting charge state of retired lithium battery - Google Patents
Method and system for predicting charge state of retired lithium battery Download PDFInfo
- Publication number
- CN116794518A CN116794518A CN202310752959.XA CN202310752959A CN116794518A CN 116794518 A CN116794518 A CN 116794518A CN 202310752959 A CN202310752959 A CN 202310752959A CN 116794518 A CN116794518 A CN 116794518A
- Authority
- CN
- China
- Prior art keywords
- value
- charge
- battery
- lithium battery
- support vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 title claims abstract description 87
- 229910052744 lithium Inorganic materials 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 75
- 238000012706 support-vector machine Methods 0.000 claims abstract description 59
- 238000012795 verification Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 51
- 230000006870 function Effects 0.000 claims description 50
- 238000012360 testing method Methods 0.000 claims description 20
- 230000009977 dual effect Effects 0.000 claims description 13
- 238000012417 linear regression Methods 0.000 claims description 13
- 238000002790 cross-validation Methods 0.000 claims description 10
- 238000005259 measurement Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 description 13
- 238000007599 discharging Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 6
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 229910001416 lithium ion Inorganic materials 0.000 description 5
- 238000010276 construction Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000032683 aging Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/385—Arrangements for measuring battery or accumulator variables
- G01R31/387—Determining ampere-hour charge capacity or SoC
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Secondary Cells (AREA)
Abstract
The application discloses a charge state prediction method and a charge state prediction system for retired lithium batteries, wherein the method comprises the following steps: acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value of a retired lithium battery at the current moment; inputting the acquired data into a support vector machine model, and predicting to obtain a state of charge (SOC) value of the retired lithium battery at the current moment; the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method. According to the application, the retired lithium battery is predicted by using a support vector machine, a sample space is mapped to a high-dimensional characteristic space, and a nonlinear problem is converted into a linear problem, so that the accurate prediction of the SOC of the retired lithium battery is realized.
Description
Technical Field
The application relates to the technical field of batteries of electric automobiles, in particular to a charge state prediction method and a charge state prediction system for retired lithium batteries.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
With the vigorous development of new energy electric automobile markets, how to properly process retired power batteries becomes an important consideration for guaranteeing the healthy development of industries. The recycling of the power battery has important practical significance for promoting the continuous healthy development of the new energy automobile industry, protecting the ecological environment and the social safety and guaranteeing the strategic resource supply.
Due to the fact that the power battery is different in factors such as self-discharging degree and environment temperature in the using process, inconsistency exists in capacity, internal resistance and voltage of the retired power battery, and further differences exist in aging degree of the power battery. Therefore, it is necessary to predict the State of Charge (SOC) of each retired battery to achieve efficient use of the retired battery.
State of Charge (SOC) is one of the key states of battery monitoring, defined as the percentage of the remaining capacity of the battery to its maximum capacity. The reliable SOC estimation can accurately judge the current state of the battery, prevent possible dangers and ensure the safe and stable operation of the battery. However, due to the non-linear and time-varying characteristics of the lithium ion battery SOC, SOC values cannot be directly observed, and the battery discharge characteristics are easily affected by factors such as battery aging, temperature variation, and the like, making SOC estimation challenging.
Currently, methods for detecting and predicting the SOC of the lithium battery comprise an open circuit voltage method, an ampere-hour integration method, a Kalman filtering method, a neural network method and the like. The method comprises the steps of determining an initial value of the SOC, calculating the SOC value at any moment by utilizing accumulation summation of currents and rated capacity of a battery, wherein the accumulation error exists in an integration process, and when the current fluctuation is large, the accumulation error can influence the SOC estimation precision to cause error estimation, and the initial value of the SOC is difficult to obtain generally; the open-circuit voltage method estimates through the relation between the open-circuit voltage and the SOC of the battery, but the battery needs to be fully stood and cannot be estimated in the running working state of the battery, so that the reliability in practical application is not high; kalman filtering (KF, kalman Filter) can model and estimate a linear system, so that the accuracy of the EKF estimation SOC depends on a constructed battery model, and the algorithm complexity is high; the neural network method estimates the SOC by carrying out network training on a large amount of battery working condition operation data to form a network with nonlinear prediction capability, has higher accuracy, is commonly used for predicting nonlinear problems, but has the accuracy related to the setting of network parameters, and the incorrect setting of the network parameters can affect the accuracy of a final prediction result.
Disclosure of Invention
In order to solve the problems, the application provides a charge state prediction method and a charge state prediction system for a retired lithium battery, which predict the retired lithium battery by using a support vector machine in machine learning, map a sample space to a high-dimensional characteristic space, convert a nonlinear problem into a linear problem, and further realize accurate prediction of the charge state SOC of the retired lithium battery.
In a first aspect, the present disclosure provides a state of charge prediction method for retired lithium batteries.
A state of charge prediction method for retired lithium batteries, comprising:
acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value of a retired lithium battery at the current moment;
inputting the acquired data into a support vector machine model, and predicting to obtain a state of charge (SOC) value of the retired lithium battery at the current moment;
the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method.
Further technical scheme, support vector machine model's construction and training includes:
performing offline charge and discharge test on the retired lithium battery to obtain measured data of the retired lithium battery under different working conditions, wherein the measured data comprise a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value;
sampling the obtained actual measurement data to construct a training sample set;
and training a support vector machine model by using the constructed training sample set, and constructing a regression model.
According to the further technical scheme, after a training sample set is obtained, preprocessing is carried out on data in the training sample set; the pretreatment comprises the following steps: and carrying out standardization processing on the sampled battery voltage value, battery current value, battery surface temperature value and state of charge (SOC) value.
According to a further technical scheme, the normalized sample value is the ratio of the difference between the sample value and the minimum value of the sample to the difference between the maximum value and the minimum value of the sample value.
According to a further technical scheme, the construction of the regression model comprises the following steps:
constructing a linear regression function, constructing a linear loss function according to the linear regression function, introducing a relaxation variable, and converting the linear loss function into a constraint relation;
converting the constraint relation into a dual problem by utilizing a Lagrange dual theory, solving to obtain Lagrange multipliers in the dual problem, and further obtaining regression coefficients and threshold values of the linear regression function;
based on the obtained linear regression function, adding nonlinear transformation, and mapping an input feature space of the model into a high-dimensional feature space;
constructing a kernel function, combining a training sample set, solving in an input feature space to obtain an inner product in a high-dimensional feature space, and training to obtain a final support vector machine model.
Further, the method for determining the optimal parameter combination of the support vector machine model by using a grid search method and a cross validation method comprises the following steps:
cross-verifying various possible radial basis function parameter combination values, and determining a combination pair with highest cross-verification accuracy;
each kernel function parameter combination is used for training a Support Vector Machine (SVM) model, and cross verification is used for evaluating performance;
after the fitting function attempts all the parameter combinations, an optimal classification surface is returned, and the optimal parameter combinations are automatically adjusted, so that the errors of all training samples from the optimal classification surface are minimized.
According to a further technical scheme, the parameter combination comprises gamma parameters of the radial basis function and penalty parameters C in the regression model.
In a second aspect, the present disclosure provides a state of charge prediction system for retired lithium batteries.
A state of charge prediction system for retired lithium batteries, comprising:
the data acquisition module is used for acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value at the current moment of the retired lithium battery;
the charge state prediction module is used for inputting the acquired data into the support vector machine model, and predicting to obtain a charge state SOC value of the retired lithium battery at the current moment;
the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method.
In a third aspect, the present disclosure also provides an electronic device comprising a memory and a processor, and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the steps of the method of the first aspect.
In a fourth aspect, the present disclosure also provides a computer readable storage medium storing computer instructions which, when executed by a processor, perform the steps of the method of the first aspect.
Compared with the prior art, the application has the beneficial effects that:
(1) The application provides a charge state prediction method and a charge state prediction system for a retired lithium battery, wherein a support vector machine in machine learning is used for predicting the retired lithium battery, a sample space is mapped to a high-dimensional characteristic space, a nonlinear problem is converted into a linear problem, the problem that regression cannot be predicted by the nonlinear problem is solved, the linear problem is converted, the convergence speed of a model is improved, and the accurate prediction of the charge state SOC of the retired lithium battery is realized.
(2) In the application, the SOC value of the last moment of the retired lithium battery is used as an input variable to be added into the support vector machine model, so that the prediction precision of the model for predicting the SOC value of the next moment of the model is improved.
Additional features and advantages of the application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the application.
Drawings
FIG. 1 is a flow chart of a state of charge prediction method for retired lithium batteries in an embodiment of the present application;
fig. 2 is an experimental flow chart of state of charge prediction for retired lithium batteries in an embodiment of the application;
FIG. 3 is a schematic diagram showing the results of experiment 1 in the embodiment of the present application;
FIG. 4 is a schematic diagram showing the results of experiment 2 in the example of the present application;
FIG. 5 is a schematic diagram showing the results of experiment 3 in the embodiment of the present application.
Detailed Description
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the application. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present application. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
Example 1
Aiming at the problems of complex algorithm and low accuracy of the existing lithium battery SOC prediction method, the embodiment discloses a charge state prediction method of a retired lithium battery, which predicts the retired lithium battery by using a support vector machine, maps a sample space to a high-dimensional characteristic space, converts a nonlinear problem into a linear problem, and realizes accurate prediction of the charge state SOC of the retired lithium battery, as shown in fig. 1, the method specifically comprises the following steps:
acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value of a retired lithium battery at the current moment;
inputting the acquired data into a support vector machine model, and predicting to obtain a state of charge (SOC) value of the retired lithium battery at the current moment;
the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method.
Considering that the SOC of the lithium battery is predicted to be nonlinear, and the effect of the support vector machine in machine learning on solving the nonlinear problem is better, the embodiment builds and trains a support vector machine model to realize accurate prediction of the SOC of the retired lithium battery. Furthermore, the support vector machine model is suitable for lithium battery SOC prediction after retirement, and data adopted by the construction and training of the support vector machine model are obtained based on retired lithium battery tests, so that the support vector machine model is suitable for SOC estimation during secondary utilization of retired batteries on electric vehicles, such as SOC prediction of echelon lithium battery energy storage. For the retired lithium battery, because the difference between the retired lithium battery and the lithium battery being used on the electric vehicle mainly lies in the internal resistance and the actual capacity of the battery, the SOC estimation of the retired lithium battery cannot be performed by adopting a traditional ampere-hour integration method and the like, but the big data fusion algorithm of the support vector machine provided by the embodiment is needed to be adopted, the model of the retired lithium battery is built based on the algorithm, and the SOC estimation of the lithium battery is further realized.
Specifically, as shown in fig. 2, the construction and training of the support vector machine model described above is as follows.
Firstly, performing offline charge and discharge test on a retired lithium battery to obtain actual measurement data of the retired lithium battery under different working conditions, including a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value, sampling the obtained actual measurement data, and constructing a training sample set.
Specifically, a lithium battery charging and discharging platform is used for offline charging and discharging of the retired lithium battery, and actual measurement data of the retired lithium battery under different working conditions are obtained. In this embodiment, a data set of the retired lithium battery is obtained by performing a charge-discharge test on the retired lithium battery. It should be noted that SOC estimation and SOH (State of Health) estimation are two different concepts, and SOH estimation is actually an estimation of battery life, and when estimating SOH, multiple cycles of charge and discharge are often required, and for SOC estimation, multiple cycles of charge and discharge are not required, and for a certain retired lithium battery, only one charge and discharge test is performed.
The data set comprises data collected at three temperature conditions of 0 ℃, 25 ℃ and 45 ℃, and loads of different simulated automobile driving states are respectively applied to the lithium ion battery under the same temperature conditions. Each data set comprises a full charge and full discharge process of the lithium ion battery, and is obtained through two experiments, wherein one data set is used for applying a simulated load from 80% of the remaining lithium ion battery capacity to the end of discharge, and the other data set is used for applying a simulated load from 50% of the remaining lithium ion battery capacity to the end of discharge. It should be noted that, the data adopted in this embodiment are all experimental data of the lithium battery after retired on the electric automobile on the charging and discharging cabinet, and the experimental data are accurate.
In the process of the charge-discharge experiment, the battery voltage value, the battery current value, the battery surface temperature value and the state of charge (SOC) value of the retired lithium battery are obtained, the obtained data are sampled under different working conditions, and the data interval is set to be 1s to generate a data point, so that a data set is constructed. The battery voltage value, the battery current value and the battery surface temperature value are obtained through sensors in the charging and discharging cabinet; the SOC value is calculated and obtained by the formula soc=q_domain/q_rate, q_rate is the nominal (rated) charge capacity of the battery, q_domain is the residual charge in the battery, the unit is Ah, and the two energy parameter values can be directly read through the charging and discharging cabinet. Dividing the acquired data set into a training sample set and a test sample set, training the built support vector machine model through the training set, and testing the trained support vector machine model through the test set. In this embodiment, the effect of this embodiment is tested using the previous data set as a training sample set and the next data set as a test sample set.
As another embodiment, the acquired data is preprocessed after the training sample set is acquired and constructed. In this embodiment, the sampled battery voltage value, battery current value, battery surface temperature value and state of charge SOC value are subjected to standardization processing, so as to facilitate subsequent training and establishment of the support vector machine model.
Further, the normalized sample value is the ratio of the difference between the sample value and the minimum value of the sample to the difference between the maximum value and the minimum value of the sample value.
And secondly, training a support vector machine model by using the constructed training sample set, and constructing a regression model.
In the step, firstly, a linear regression function is constructed, a linear loss function is constructed according to the linear regression function, a relaxation variable is introduced, the linear loss function is converted into a constraint relation, the constraint relation is converted into a dual problem by utilizing a Lagrange dual theory, lagrange multipliers in the dual problem are obtained by solving, and then regression coefficients and thresholds of the linear regression function are obtained.
The support vector machine is a supervised learning method in machine learning, and is called Support Vector Regression (SVR) when used for regression problem, and the decision function (i.e. regression function) of support vector regression is:
f(x)＝w T x+b (1)
wherein w is T And b is a threshold value.
For a given training sample set d= { (x) 1 ,y 1 ),(x 2 ,y 2 ),...,(x l ,y l )}，y i E R, training to obtain a regression model, so that the model predictive value f (x i ) And true value y i As identical as possible. The traditional regression model is to predict the model predictive value f (x i ) And true value y i The difference in (2) is noted as loss if and only if f (x i )＝y i When the loss is noted as 0; while support vector regression allows the predicted value f (x i ) And true value y i With a deviation epsilon between, when |f (x i )-y i When the I is less than or equal to epsilon, the prediction is considered to be correct, and the loss is not considered; when |f (x i )-y i |>At ε, the loss was calculated. The above can be expressed as:
wherein C is a penalty constant, l ε As a loss function, defined as:
to minimize the error, a relaxation variable ζ is introduced i And xi i The problem is converted into:
solving the constraint relation, firstly, introducing a pull in the constraint relationGelang's multiplier, mu 1 ≥0，α i ≥0，/>And further obtaining a Lagrangian function formula, which is:
the original problem is a very small and very large problem, and in order to solve the problem conveniently, under the condition that a KKT condition (a necessary condition for non-linear programming of optimal solution) is satisfied, the Lagrange dual theory is adopted to convert a constraint relation into a dual problem, namely, the formula (5) is converted into the dual problem, and the dual problem is changed into the very large and very small problem. According toFor w, b, ζ, ++>The partial derivative is zero, which can be obtained:
C＝a i +μ i (8)
substituting the formulas (6) to (9) into the formula (5) can obtain:
the above process needs to satisfy the KKT expression, i.e. the requirement
If and only if f (x i )-y i -ε-ξ i When=0, α i Can take a non-zero value if and only ifWhen (I)>Can take non-zero value, constraint f (x i )-y i -ε-ξ i =0 and +.>Cannot be established at the same time, thus alpha i And->At least one of which is zero.
Substituting the formula (10) into the formula (11) to obtain the regression coefficient w of the regression model T And a threshold b, and further obtaining an expression of the SVR model, wherein the expression is as follows:
however, in practical engineering, many problems are linear inseparable, and if the SOC prediction problem of the lithium battery in this embodiment is described above, the linear regression SVR cannot fit the data. Assuming that the input space is European space, the linearity is not separable but is transformed by some non-linearityMapping the input feature space to a high-dimensional feature space,the high-dimensional feature space is an European space. The original space is of finite dimensions, then there must be a high dimensional feature space such that the sample is separable.
Therefore, on the basis of the obtained linear regression function, adding nonlinear transformation, and mapping the input feature space of the model into a high-dimensional feature space;
constructing a kernel function, combining a training sample set, solving in an input feature space to obtain an inner product in a high-dimensional feature space, and training to obtain a final support vector machine model.
Specifically, on the basis of the original regression model, nonlinear transformation is added to map the input feature space of the model into the high-dimensional feature space, and the above formula (12) becomes:
in the above formula (13), phi (x) i ) T φ(x j ) Representing solution x i And x j When the inner product of (a) is high-dimensional in the face of the input space, it becomes difficult to directly calculate the inner product, and to reduce the calculation amount, a kernel function k (x i ,x j )＝φ(x i ) T (x j )＝<φ(x i ),φ(x j )>The kernel function κ can be used to determine the inner product in the high-dimensional feature space, i.e., κ (x) i ,x j )＝φ(x i ) T φ(x j ) Then equation (13) converts to:
the above formula (14) is the basic form of a support vector machine SVR model based on kernel functions, wherein alpha i ≥0、For the optimal solution of the lagrangian multiplier, κ (x, x i ) Representing a kernel function.
Currently, common kernel functions mainly include a linear kernel, a polynomial kernel, a radial basis kernel (gaussian kernel), a laplace kernel, a Sigmoid kernel, and the like. In this embodiment, a radial basis kernel is selected as the kernel function. Wherein, the formula of the radial basis function is:
the radial basis function comprises a parameter gamma, and the regression model comprises a penalty parameter C. The larger the C value is, the larger the punishment to the misclassified points is, so that the model is more complex, the fitting is easy, and the generalization capability is poor; the smaller the C value is, the smaller the point punishment to misclassification is, misclassification point parameters are not important, and the fitting is easy to be carried out under. Therefore, how to adjust these parameters to optimize the fitting effect is the key of the experiment. In this embodiment, a grid search method is used in combination with a cross-validation method to traverse all the candidate parameters, make a correct evaluation on the model, and determine an optimal regression model based on a radial basis function.
The grid search method is a violent solution, the numerical value of the parameter to be adjusted is traversed by adopting an exhaustion method, all parameters are arranged and combined to generate a grid, all results are used in SVR for training, the results after all the parameters are compared, and the optimal parameter combination is automatically returned.
The k-fold cross validation refers to dividing a data set into k subsets with similar sizes, training the model by k-1 subsets each time, testing the model by the rest subset, performing k training and testing, and finally returning the average value of k testing results to evaluate the model.
In this embodiment, determining the optimal parameter combination of the support vector machine model by using the grid search method in combination with the cross validation method includes:
various possible combination values are tried and then cross-validated to find the combination pair that maximizes the accuracy of the cross-validation. The possible values of all the parameters are arranged and combined, and all possible combined results are listed to generate a grid;
each combination was used for SVM training and performance was evaluated using cross-validation;
after the fitting function attempts all the parameter combinations, it returns a proper classification plane and automatically adjusts to the optimal parameter combination.
Further, after determining the optimal parameter combination of the model, the predictive model is trained. When the performance of the models is the same, a combination of parameters with a relatively small penalty factor C is preferred in order to reduce computation time.
And finally, inputting the data in the acquired test sample into a support vector machine model, wherein the data comprises a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value at the current moment of the retired lithium battery, and predicting to obtain the SOC value at the current moment of the retired lithium battery.
To further verify the superiority of the protocol described in this example, it was verified by the following experiment. Specifically, a lithium battery charging and discharging platform is used for offline charging and discharging of a certain retired lithium battery, 3 experiments are respectively carried out, actual measurement data of the retired lithium battery are obtained, the actual measurement data comprise a battery voltage value, a battery current value, a battery surface temperature value and a state of charge SOC value, and the obtained actual measurement data are sampled to form a data set. The data set is divided into a training set and a testing set, and a support vector machine model is trained and tested.
In the 1 st experiment, the battery voltage value, the battery current value and the battery surface temperature value at the current moment of the retired lithium battery are taken as input values, input into a support vector machine model, and the state of charge (SOC) value at the next moment is output. The parameters of the support vector machine model are manually set, and parameters gamma=30 and C=30 are set. As shown in fig. 3, the Training origin is the Training set original SOC value, and the Training prediction is the Training set predicted SOC value. At this time, train_score=0.9864 and test_score= 0.9685, i.e., the training set prediction accuracy is 98.64% and the test set prediction accuracy is 96.85%. The score of the test set is smaller than that of the training set, which indicates that the model is too complex, the overfitting phenomenon occurs, and the generalization capability is poor. In practice, this further illustrates the importance of the appropriate parameters to the experimental results.
In the 2 nd experiment, the grid search method provided by the embodiment is combined with the cross validation method to automatically find the optimal parameters of the model, the training set and the test set data are input into the SVR model after training, and the SOC value prediction of the next moment of the retired lithium battery is carried out. By automatic optimizing, gamma value is 1.0, C value is 16, and the prediction result is shown in FIG. 4. At this time, train_score=0.9877 and test_score=0.9884, i.e., the training set prediction accuracy is 98.77% and the test set prediction accuracy is 98.84%.
In order to further improve the fitting accuracy, in this embodiment, the SOC value of the last time of the retired lithium battery is used as an influencing factor for predicting the SOC value of the next time, that is, considering that the current SOC value of the lithium battery may be related to the SOC value of the last time, the SOC value of the last time is used as an input variable to be added to the model, and the 3 rd experiment is performed.
In the 3 rd experiment, the battery voltage value, the battery current value, the battery surface temperature value and the SOC value at the previous moment of the retired lithium battery are taken as input values and are input into a support vector machine model, and the SOC value at the next moment is output. At this time, the data set is still nonlinear, the input space is mapped into the high-dimensional feature space by using the radial basis function, and the optimal parameters are found by using a grid search and cross-validation method. By automatic optimizing, gamma value is determined to be 0.0312, C value is determined to be 0.5, and the prediction result is shown in FIG. 5. At this time, train_score=0.9971 and test_score= 0.9971, i.e., the training set prediction accuracy was 99.71% and the test set prediction accuracy was 99.71%.
The experimental results by different optimization methods are shown in table 1 below.
Table 1 comparison of experimental results of different optimization methods
According to the experimental results, the model can autonomously search parameters and return optimal parameters through grid search and a cross-validation method; meanwhile, on the basis that the battery voltage value, the battery current value and the battery surface temperature value are taken as input variables, the SOC value at the moment on the retired lithium battery is also taken as the input variables, and compared with the fitting effect which only takes the battery voltage value, the battery current value and the battery surface temperature value as the input variables, the fitting effect is better and the precision is higher.
Through the experiment, the scheme of the embodiment can realize the prediction of the SOC value of the retired lithium battery at the next moment with higher precision.
Example two
The embodiment provides a charge state prediction system of a retired lithium battery, which comprises:
the data acquisition module is used for acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value at the current moment of the retired lithium battery;
the charge state prediction module is used for inputting the acquired data into the support vector machine model, and predicting to obtain a charge state SOC value of the retired lithium battery at the current moment;
the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method.
Example III
The present embodiment provides an electronic device comprising a memory and a processor, and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the steps in a state of charge prediction method for a retired lithium battery as described above.
Example IV
The present embodiment also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, perform the steps in a state-of-charge prediction method for retired lithium battery as described above.
The steps involved in the second to fourth embodiments correspond to the first embodiment of the method, and the detailed description of the second embodiment refers to the relevant description of the first embodiment. The term "computer-readable storage medium" should be taken to include a single medium or multiple media including one or more sets of instructions; it should also be understood to include any medium capable of storing, encoding or carrying a set of instructions for execution by a processor and that cause the processor to perform any one of the methods of the present application.
It will be appreciated by those skilled in the art that the modules or steps of the application described above may be implemented by general-purpose computer means, alternatively they may be implemented by program code executable by computing means, whereby they may be stored in storage means for execution by computing means, or they may be made into individual integrated circuit modules separately, or a plurality of modules or steps in them may be made into a single integrated circuit module. The present application is not limited to any specific combination of hardware and software.
The above description is only of the preferred embodiments of the present application and is not intended to limit the present application, but various modifications and variations can be made to the present application by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
While the foregoing description of the embodiments of the present application has been presented in conjunction with the drawings, it should be understood that it is not intended to limit the scope of the application, but rather, it is intended to cover all modifications or variations within the scope of the application as defined by the claims of the present application.
Claims (10)
1. The charge state prediction method of the retired lithium battery is characterized by comprising the following steps of:
acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value of a retired lithium battery at the current moment;
inputting the acquired data into a support vector machine model, and predicting to obtain a state of charge (SOC) value of the retired lithium battery at the current moment;
the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method.
2. The method for predicting the state of charge of a retired lithium battery according to claim 1, wherein the support vector machine model building and training comprises:
performing offline charge and discharge test on the retired lithium battery to obtain measured data of the retired lithium battery under different working conditions, wherein the measured data comprise a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value;
sampling the obtained actual measurement data to construct a training sample set;
and training a support vector machine model by using the constructed training sample set, and constructing a regression model.
3. The method for predicting the state of charge of a retired lithium battery according to claim 2, wherein after a training sample set is obtained, data in the training sample set is preprocessed; the pretreatment comprises the following steps: and carrying out standardization processing on the sampled battery voltage value, battery current value, battery surface temperature value and state of charge (SOC) value.
4. A method of predicting state of charge of a retired lithium battery according to claim 3, wherein the normalized sample value is a ratio of a difference between the sample value and a minimum value of the sample to a difference between a maximum value and a minimum value of the sample value.
5. The method for predicting the state of charge of a retired lithium battery according to claim 1, wherein the constructing of the regression model comprises:
constructing a linear regression function, constructing a linear loss function according to the linear regression function, introducing a relaxation variable, and converting the linear loss function into a constraint relation;
converting the constraint relation into a dual problem by utilizing a Lagrange dual theory, solving to obtain Lagrange multipliers in the dual problem, and further obtaining regression coefficients and threshold values of the linear regression function;
based on the obtained linear regression function, adding nonlinear transformation, and mapping an input feature space of the model into a high-dimensional feature space;
constructing a kernel function, combining a training sample set, solving in an input feature space to obtain an inner product in a high-dimensional feature space, and training to obtain a final support vector machine model.
6. The method of claim 1, wherein determining the optimal combination of parameters for the support vector machine model using a grid search method in combination with a cross-validation method comprises:
cross-verifying various possible radial basis function parameter combination values, and determining a combination pair with highest cross-verification accuracy;
each kernel function parameter combination is used for training a Support Vector Machine (SVM) model, and cross verification is used for evaluating performance;
after the fitting function attempts all the parameter combinations, an optimal classification surface is returned, and the optimal parameter combinations are automatically adjusted, so that the errors of all training samples from the optimal classification surface are minimized.
7. The method of claim 6, wherein the parameter combination includes gamma parameters of the radial basis function and penalty parameters C in the regression model.
8. A state of charge prediction system for retired lithium batteries, comprising:
the data acquisition module is used for acquiring a battery voltage value, a battery current value, a battery surface temperature value and a state of charge (SOC) value at the current moment of the retired lithium battery;
the charge state prediction module is used for inputting the acquired data into the support vector machine model, and predicting to obtain a charge state SOC value of the retired lithium battery at the current moment;
the support vector machine model is a regression model based on a radial basis function, the input feature space of the model is mapped into a high-dimensional feature space, and the optimal parameter combination of the support vector machine model is determined by combining a grid search method and a cross verification method.
9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and running on the processor, which when executed by the processor, perform the steps of a method of predicting state of charge of a retired lithium battery according to any one of claims 1-7.
10. A computer readable storage medium storing computer instructions which, when executed by a processor, perform the steps of a method of predicting state of charge of a retired lithium battery according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310752959.XA CN116794518A (en) | 2023-06-21 | 2023-06-21 | Method and system for predicting charge state of retired lithium battery |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310752959.XA CN116794518A (en) | 2023-06-21 | 2023-06-21 | Method and system for predicting charge state of retired lithium battery |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116794518A true CN116794518A (en) | 2023-09-22 |
Family
ID=88034235
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310752959.XA Pending CN116794518A (en) | 2023-06-21 | 2023-06-21 | Method and system for predicting charge state of retired lithium battery |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116794518A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117954716A (en) * | 2024-03-27 | 2024-04-30 | 深圳市杰成镍钴新能源科技有限公司 | Discharge voltage rebound control method and device for retired lithium battery |
-
2023
- 2023-06-21 CN CN202310752959.XA patent/CN116794518A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117954716A (en) * | 2024-03-27 | 2024-04-30 | 深圳市杰成镍钴新能源科技有限公司 | Discharge voltage rebound control method and device for retired lithium battery |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chang et al. | Lithium-ion battery state of health estimation using the incremental capacity and wavelet neural networks with genetic algorithm | |
| Deng et al. | Battery health estimation with degradation pattern recognition and transfer learning | |
| Shu et al. | Stage of charge estimation of lithium-ion battery packs based on improved cubature Kalman filter with long short-term memory model | |
| Ma et al. | A novel data-model fusion state-of-health estimation approach for lithium-ion batteries | |
| Jiang et al. | State of health estimation for lithium-ion battery using empirical degradation and error compensation models | |
| Zhang et al. | Online State of Charge Estimation of Lithium‐Ion Cells Using Particle Filter‐Based Hybrid Filtering Approach | |
| Mamo et al. | Long short-term memory with attention mechanism for state of charge estimation of lithium-ion batteries | |
| CN103616647A (en) | Battery remaining capacity estimation method for electric car battery management system | |
| KR102246633B1 (en) | Battery Output Voltage Response and State-of-Charge Forecasting Optimization Method using Hybrid VARMA and LSTM | |
| Wei et al. | Lyapunov-based state of charge diagnosis and health prognosis for lithium-ion batteries | |
| Samadani et al. | A review study of methods for lithium-ion battery health monitoring and remaining life estimation in hybrid electric vehicles | |
| Jiang et al. | An aging-aware soc estimation method for lithium-ion batteries using xgboost algorithm | |
| Duan et al. | Online parameter identification and state of charge estimation of battery based on multitimescale adaptive double Kalman filter algorithm | |
| Tang et al. | Comprehensive study and improvement of experimental methods for obtaining referenced battery state-of-power | |
| CN117805649A (en) | Method for identifying abnormal battery cells based on SOH quantized battery capacity attenuation | |
| CN116794518A (en) | Method and system for predicting charge state of retired lithium battery | |
| Wang et al. | Charging diagnosis of power battery based on adaptive STCKF and BLS for electric vehicles | |
| CN112946480B (en) | Lithium battery circuit model simplification method for improving SOC estimation real-time performance | |
| Singh et al. | State of charge estimation techniques of Li-ion battery of electric vehicles | |
| Wu et al. | State of health estimation for lithium-ion batteries in real-world electric vehicles | |
| Poh et al. | A review of machine learning applications for li-ion battery state estimation in electric vehicles | |
| Song et al. | A hierarchical state of charge estimation method for lithium-ion batteries via xgboost and kalman filter | |
| Sugashini et al. | Hybrid Deep Learning Algorithm for the State of Charge Prediction of the Lithium-Ion Battery for Electric Vehicles | |
| Aliberti et al. | Comparative Analysis of Neural Networks Techniques for Lithium-ion Battery SOH Estimation | |
| Wang et al. | Online capacity estimation of lithium-ion batteries by partial incremental capacity curve |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |