CN114755586A - Lithium ion battery residual life prediction method - Google Patents

Lithium ion battery residual life prediction method Download PDF

Info

Publication number
CN114755586A
CN114755586A CN202210452229.3A CN202210452229A CN114755586A CN 114755586 A CN114755586 A CN 114755586A CN 202210452229 A CN202210452229 A CN 202210452229A CN 114755586 A CN114755586 A CN 114755586A
Authority
CN
China
Prior art keywords
algorithm
lithium ion
ion battery
whale
equivalent circuit
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.)
Granted
Application number
CN202210452229.3A
Other languages
Chinese (zh)
Other versions
CN114755586B (en
Inventor
徐自强
朱洪涛
吴孟强
周海平
冯婷婷
张庶
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Yangtze River Delta Research Institute of UESTC Huzhou
Original Assignee
University of Electronic Science and Technology of China
Yangtze River Delta Research Institute of UESTC Huzhou
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China, Yangtze River Delta Research Institute of UESTC Huzhou filed Critical University of Electronic Science and Technology of China
Priority to CN202210452229.3A priority Critical patent/CN114755586B/en
Publication of CN114755586A publication Critical patent/CN114755586A/en
Application granted granted Critical
Publication of CN114755586B publication Critical patent/CN114755586B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • 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
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
    • Y02E60/10—Energy storage using batteries

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Artificial Intelligence (AREA)
  • Computational Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Operations Research (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Medical Informatics (AREA)
  • Algebra (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Secondary Cells (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

The invention provides a lithium ion battery residual life prediction method based on a random forest RF fusion equivalent circuit model optimized by an improved whale algorithm IWOA. The method establishes an equivalent circuit model of the lithium ion battery, and uses Electrochemical Impedance Spectroscopy (EIS) test data to fit model parameters; the whale algorithm WOA is improved, the RF algorithm is used as a main body of a prediction algorithm, the IWOA algorithm is used for optimizing the hyperparameter in the RF algorithm, and therefore the prediction of the residual life of the lithium ion battery by means of the fitted equivalent circuit model parameters is achieved. The prediction algorithm can effectively and accurately predict the current residual cycle number of the lithium ion battery, and simultaneously ensures the real-time property and the nondestructive property of battery detection, thereby better planning the actual application of the lithium ion battery.

Description

Lithium ion battery residual life prediction method
Technical Field
The invention provides an algorithm for predicting the residual life of a lithium ion battery. In particular to a method for predicting the residual life of a lithium ion battery based on a random forest optimized by an improved whale algorithm and by fusing an equivalent circuit model.
Background
The remaining life of the battery is an important index of the lithium ion battery, reflects the remaining complete charge and discharge cycle times of the lithium ion battery before the end of the life, and provides reference and basis for scrapping, recycling and replacing a battery system. In order to avoid safety accidents in the use process of the lithium ion battery and check whether the service life of the lithium ion battery is ended in real time, the residual service life of the lithium ion battery needs to be predicted. The method has the advantages that the residual life is accurately estimated, the premise of full utilization and safe use of the lithium ion battery is provided, the maximum utilization of the performance of the lithium ion battery and the prolonging of the service life of the lithium ion battery can be realized, and the method is scientifically applied to actual life.
Disclosure of Invention
The invention provides an improved whale algorithm random forest IWOA-RF fused with an equivalent circuit model, and aims to accurately predict the remaining service life through EIS excitation, namely a sinusoidal signal of one-time frequency conversion.
In order to achieve the purpose, the design scheme of the invention is as follows:
a lithium ion battery residual life prediction method is based on random forest RF optimized by improved whale algorithm IWOA fused with an equivalent circuit model, and comprises the steps of establishing and fitting the equivalent circuit model, improving whale algorithm WOA, optimizing random forest RF, and predicting lithium ion battery residual life by improved whale algorithm-random forest IWOA-RF algorithm, wherein the prediction process comprises the following steps:
(1) establishing an equivalent circuit model of the lithium ion battery, and fitting parameter values of equivalent circuit elements by performing EIS test on the equivalent circuit model;
(2) improving a whale algorithm, taking a random forest RF algorithm as a main body of a prediction algorithm, and using the improved whale IWOA algorithm for optimization of hyper-parameters in the random forest RF algorithm;
(3) dividing a data set into a training set and a testing set, training an algorithm model by using training set data based on an improved whale random forest IWOA-RF algorithm; substituting the test set model into the algorithm model, and quantitatively evaluating the prediction capability of the model by establishing an evaluation system.
Preferably, the establishment of the equivalent circuit model comprises the following steps:
firstly, corresponding Nyquist diagrams of the lithium ion battery under different cycle times are measured, and corresponding equivalent circuit models of the lithium ion battery are designed step by step according to the characteristics of the Nyquist diagrams of different equivalent circuit elements;
fitting parameter values of each element in the equivalent circuit model through EIS data of the lithium ion battery so as to represent the current service life state of the lithium ion battery;
and thirdly, selecting EIS data and residual life data of the lithium ion battery with the available capacity of more than 70% of the rated capacity.
Preferably, the whale algorithm WOA simulates three stages of whale predation: an envelope phase, a foaming phase and a search phase.
Preferably, the improved whale algorithm IWOA comprises the following improvement process:
introducing a nonlinear weight factor: in order to improve the globality of the whale algorithm, a nonlinear adaptive weight factor is provided:
Figure BDA0003619157270000021
where ω is a weight factor, k is an adjustment coefficient, and k is>1, t is the current iteration number, ImaxIs the maximum number of iterations; introducing non-linear weightsAfter the factor, the WOA algorithm formula is updated as:
Figure BDA0003619157270000022
Figure BDA0003619157270000023
in the formula (I), the compound is shown in the specification,
Figure BDA0003619157270000024
is the position vector of the current whale individual,
Figure BDA0003619157270000025
is the whale position vector with the highest current fitness,
Figure BDA0003619157270000026
is a vector of coefficients that is a function of,
Figure BDA0003619157270000027
and
Figure BDA0003619157270000028
is the value of the intermediate process, and,
Figure BDA0003619157270000029
a distance vector representing the current whale position and prey, b being a constant related to the spiral shape, l being [ -1,1 [ -1]A random number within the interval of time,
Figure BDA00036191572700000210
a position vector representing a current random whale;
difference variation disturbance term: designing a differential variation disturbance term, wherein the coefficient of the differential variation disturbance term changes with the iteration number in a linear decreasing mode, and the term is defined as:
Figure BDA00036191572700000211
Figure BDA00036191572700000212
in the formula, delta is a differential variation disturbance term, and h is a dynamic adjustment coefficient; after a differential variation disturbance term is introduced, the WOA algorithm formula is updated as follows:
Figure BDA00036191572700000213
adaptive adjustment of search strategy: a probability threshold is designed such that the algorithm produces a set of solutions as randomly as possible within the initial global scope:
Figure BDA00036191572700000214
Figure BDA0003619157270000031
wherein Q is a probability threshold value,
Figure BDA0003619157270000032
as the mean fitness of the population, fminIs the worst fitness in the population, fmaxFor optimal fitness in the population, g is [0,1 ]]The random number within the interval is a random number,
Figure BDA0003619157270000033
and
Figure BDA0003619157270000034
are respectively as
Figure BDA0003619157270000035
Maximum and minimum values of (a); for each whale, use one [0,1 ]]The random number Q in between is compared with a probability threshold Q, if Q<Q, updating the positions of the whales according to the formula (8), wherein the positions of other whales are unchanged;otherwise, updating the positions of other whales according to the formula (3).
Preferably, the improvement of whale algorithm WOA and the optimization of random forest RF comprise the following steps:
setting an adjusting coefficient k in an improved whale algorithm IWOA to be 2, and setting a population scale and an evolution frequency;
secondly, the mean square error MSE is used as an objective function of IWOA and a characteristic division function of a CART tree in random forest RF;
thirdly, the number of CART trees in the RF and the leaf node number of each CART tree are used as optimization objects for improving the whale algorithm IWOA;
as a preferable mode, the establishment of a whale algorithm random forest IWOA-RF algorithm model is improved, 60% of data is divided into a training set, a network is trained, the other 40% of data is divided into a testing set, the accuracy of the network is tested, and evaluation indexes are established.
Preferably, the evaluation index includes an average absolute error MAE and a root mean square error RMSE, and the prediction effect of the algorithm is comprehensively evaluated from multiple dimensions.
Preferably, the step (1) is specifically: the method comprises the following steps of firstly, researching the relation between the real part and the imaginary part of the impedance of a common equivalent circuit element on a frequency domain complex plane, and deducing a function image of the impedance of each element. And measuring corresponding Nyquist diagrams of the lithium ion battery under different cycle times, and designing corresponding equivalent circuit models step by step according to the characteristics of the Nyquist diagrams of different equivalent circuit elements. Aiming at a Nyquist diagram of a tested lithium ion battery, an equivalent circuit model is designed to be in a form of LR (RQ) and (RQ) Q, wherein L is an inductor, R is a resistor, and Q is a constant phase element. The impedance of Q is expressed as:
Figure BDA0003619157270000036
Figure BDA0003619157270000037
Figure BDA0003619157270000038
wherein Z is the impedance of Q, ZReIs the real part of impedance, ZImIs the imaginary part of the impedance, j is the unit of imaginary number, ω is the frequency, Y0And N is two parameters of the constant phase element Q, wherein N is a dispersion index, is dimensionless and takes a value in an interval (0, 1); y is0Has the unit of Ω-1·cm-2·s-NIt is always positive.
And secondly, fitting each element parameter value in the equivalent circuit model by using software through EIS data of the lithium ion battery, thereby representing the current service life state of the lithium ion battery. The method obtains the corresponding parameter value of the complete equivalent circuit model under each sampling cycle.
Preferably, the step (2) is specifically: establishing a random forest algorithm model, and combining a plurality of CART decision tree models in parallel through a Bagging algorithm to form an integrated regressor, wherein the final result is as follows:
Figure BDA0003619157270000041
in the formula, H is the final integrated regressor, n is the number of single regressors selected from the original data set, HiThe decision result of each single regressor.
The CART decision tree is used for solving a regression problem, and MSE (mean square error) is specified to be used as a division basis of data when an RF algorithm trains a data set. Randomly selecting partial data sets from the original data sets repeatedly through a Bagging algorithm, then respectively using the data set subsets selected each time as training sets of a plurality of weak regressors, training a plurality of weak learning models, and combining the weak learning models to form an integrated regressor.
The invention has the advantages that:
the WOA algorithm is improved, the optimization capability of the WOA algorithm is improved, and the WOA algorithm has stronger global search capability and is not easy to fall into a local optimal solution; the equivalent circuit model is fused, the residual life of the lithium ion battery can be obtained in real time, the battery cannot be damaged, and the method has high universality and high prediction accuracy.
Drawings
FIG. 1 is an example of the equivalent circuit model and the fitting results of the parameters of the components thereof.
FIG. 2 is a comparison of the original sampling points of the lithium ion battery EIS of the present invention with the fitted curve.
FIG. 3 shows the prediction result of the remaining life of the IWOA-RF lithium ion battery fused with the equivalent circuit model.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention.
Example 1
The embodiment provides a lithium ion battery residual life prediction method, which is based on a random forest RF optimized by a whale algorithm IWOA improved based on a fusion equivalent circuit model, and comprises the steps of establishing and fitting the equivalent circuit model, improving a whale algorithm WOA and optimizing the random forest RF, and predicting the lithium ion battery residual life by the whale algorithm-random forest IWOA-RF algorithm, wherein the prediction process comprises the following steps:
(1) establishing an equivalent circuit model of the lithium ion battery, and fitting parameter values of equivalent circuit elements by performing EIS test on the equivalent circuit model;
(2) improving a whale algorithm, taking a random forest RF algorithm as a main body of a prediction algorithm, and using the improved whale IWOA algorithm for optimization of hyper-parameters in the random forest RF algorithm;
(3) dividing a data set into a training set and a testing set, training an algorithm model by using training set data based on an improved whale random forest IWOA-RF algorithm; substituting the test set model into the algorithm model, and quantitatively evaluating the prediction capability of the model by establishing an evaluation system.
Example 2
The embodiment provides a lithium ion battery residual life prediction method, which is based on a random forest RF optimized by a whale algorithm IWOA improved based on a fusion equivalent circuit model, and comprises the steps of establishing and fitting the equivalent circuit model, improving a whale algorithm WOA and optimizing the random forest RF, and predicting the lithium ion battery residual life by the whale algorithm-random forest IWOA-RF algorithm, wherein the prediction process comprises the following steps:
(1) establishing an equivalent circuit model of the lithium ion battery, and fitting parameter values of equivalent circuit elements by performing EIS test on the equivalent circuit model;
(2) improving a whale algorithm, taking a random forest RF algorithm as a main body of a prediction algorithm, and applying the improved whale IWOA algorithm to optimization of an over-parameter in the random forest RF algorithm;
(3) dividing a data set into a training set and a testing set, training an algorithm model by using training set data based on an improved whale random forest IWOA-RF algorithm; substituting the test set model into the algorithm model, and quantitatively evaluating the prediction capability of the model by establishing an evaluation system.
The establishment of the equivalent circuit model comprises the following steps:
firstly, measuring corresponding Nyquist diagrams of the lithium ion battery under different cycle times, and designing corresponding equivalent circuit models step by step according to the characteristics of the Nyquist diagrams of different equivalent circuit elements;
fitting parameter values of each element in the equivalent circuit model through EIS data of the lithium ion battery so as to represent the current service life state of the lithium ion battery;
and thirdly, selecting EIS data and residual life data of the lithium ion battery with the available capacity of more than 70% of the rated capacity.
The whale algorithm WOA simulates three stages of whale predation: an envelope phase, a foaming phase and a search phase.
The improved whale algorithm IWOA comprises the following improvement processes:
introducing a nonlinear weight factor: in order to improve the globality of whale algorithm, a nonlinear adaptive weight factor is proposed:
Figure BDA0003619157270000061
where ω is a weight factor, k is an adjustment coefficient, k>1, t is the current iteration number, ImaxIs the maximum iteration number; after the nonlinear weight factor is introduced, the WOA algorithm formula is updated as follows:
Figure BDA0003619157270000062
Figure BDA0003619157270000063
in the formula (I), the compound is shown in the specification,
Figure BDA0003619157270000064
is the position vector of the current whale individual,
Figure BDA0003619157270000065
is the whale position vector with the highest current fitness,
Figure BDA0003619157270000066
is a vector of coefficients that is a function of,
Figure BDA0003619157270000067
and
Figure BDA0003619157270000068
is the value of the intermediate process, and,
Figure BDA0003619157270000069
a distance vector representing the current whale position and prey, b being a constant related to the spiral shape, l being [ -1,1 [ -1]The random number within the interval is a random number,
Figure BDA00036191572700000610
a position vector representing a current random whale;
difference variation disturbance term: designing a differential variation disturbance term, wherein the coefficient of the differential variation disturbance term changes with the iteration number in a linear decreasing mode, and the term is defined as:
Figure BDA00036191572700000611
Figure BDA00036191572700000612
in the formula, delta is a differential variation disturbance term, and h is a dynamic adjustment coefficient; after a differential variation disturbance term is introduced, the WOA algorithm formula is updated as follows:
Figure BDA00036191572700000613
adaptive adjustment of search strategy: a probability threshold is designed such that the algorithm produces a set of solutions as randomly as possible within the initial global scope:
Figure BDA00036191572700000614
Figure BDA00036191572700000615
wherein Q is a probability threshold value,
Figure BDA00036191572700000616
as the mean fitness of the population, fminIs the worst fitness in the population, fmaxFor optimal fitness in the population, g is [0,1 ]]The random number within the interval is a random number,
Figure BDA00036191572700000617
and
Figure BDA00036191572700000618
are respectively as
Figure BDA00036191572700000619
Maximum and minimum values of; for each whale, use one [0,1 ]]The random number Q in between is compared with a probability threshold Q, if Q<Q, updating the positions of the whales according to the formula (8), wherein the positions of other whales are unchanged; otherwise, updating the positions of other whales according to the formula (3).
Improvement of whale algorithm WOA and optimization of random forest RF, comprising the following steps:
setting an adjusting coefficient k in an improved whale algorithm IWOA to be 2, and setting a population scale and an evolution frequency;
secondly, the mean square error MSE is used as an objective function of IWOA and a characteristic division function of a CART tree in random forest RF;
thirdly, the number of CART trees in the RF and the leaf node number of each CART tree are used as optimization objects for improving the whale algorithm IWOA;
the improved whale algorithm random forest IWOA-RF algorithm model is established by dividing 60% of data into a training set and training a network, dividing the other 40% of data into a test set, testing the accuracy of the network and establishing an evaluation index.
The evaluation indexes comprise an average absolute error MAE and a root mean square error RMSE, and the prediction effect of the multi-dimensional comprehensive evaluation algorithm is obtained.
Example 3
The embodiment provides a lithium ion battery residual life prediction method, which is a random forest RF optimized by an improved whale algorithm IWOA based on a fusion equivalent circuit model, and the specific implementation process comprises the following steps:
(1) and establishing an equivalent circuit model of the lithium ion battery. And (3) researching the relation between the real part and the imaginary part of the impedance of the common equivalent circuit element on the complex plane of the frequency domain, and deriving a function image of the impedance of each element. 4 sections of the Song NCR18650B lithium ion battery are taken as experimental objects to obtain EIS data and residual life data of the whole life cycle. And measuring corresponding Nyquist diagrams of the lithium ion battery under different cycle times, and designing corresponding equivalent circuit models step by step according to the characteristics of the Nyquist diagrams of different equivalent circuit elements.
The Nyquist plot decreases in frequency from left to right. In the high frequency region, the portion of the imaginary part below the horizontal axis is represented by one L; the part of the imaginary part above the horizontal axis presents a semicircle, denoted by one (RQ); the intersection of the curve with the horizontal axis is translated horizontally by changing the magnitude of R. In the intermediate frequency region, a semicircle is presented, also denoted with one (RQ). In the low frequency region, a straight line is present, denoted by a Q. All the parts are connected in series to obtain a complete equivalent circuit which is an LR (RQ) or (RQ) Q structure. Wherein L is an inductor, R is a resistor, and Q is a constant phase element. The impedance of Q is expressed as:
Figure BDA0003619157270000071
Figure BDA0003619157270000072
Figure BDA0003619157270000073
wherein Z is the impedance of Q, ZReIs the real part of impedance, ZImIs the imaginary part of the impedance, j is the unit of imaginary number, ω is the frequency, Y0And N is two parameters of the constant phase element Q, wherein N is a dispersion index, is dimensionless and takes a value in an interval (0, 1); y is0Has the unit of Ω-1·cm-2·s-NIt is always positive.
And fitting each element parameter value in the equivalent circuit model by using software through EIS data of the lithium ion battery so as to represent the current life state of the lithium ion battery. First, two semi-circle parts are separated, and R (RQ) parameter values of the two parts are respectively obtained and connected in series, wherein R is the average value of R of the two semi-circles. The remaining L and Q parameter values are then fitted, keeping the fitted parameter values of r (rq) and (rq) unchanged. Then LR (RQ) is taken out independently, the parameter values are updated gradually for all parts except the low-frequency area straight line, when the parameter with the front serial number is updated, the parameter value with the rear serial number is ensured to be unchanged, otherwise, the parameter with the front serial number is ensured to be changeable. And substituting the obtained result into the low-frequency region Q which is fit at the beginning, updating the parameters of the Q, and finally updating the whole. Thus, a set of complete equivalent circuit model parameter values can be obtained, the equivalent circuit model and the fitting result of the element parameters thereof are shown in fig. 1, and the comparison between the original sampling point of the lithium ion battery EIS and the fitting curve is shown in fig. 2.
The method obtains the corresponding parameter value of the complete equivalent circuit model under each sampling cycle.
(2) The whale algorithm is improved, the random forest RF algorithm is taken as a main body of the prediction algorithm, and the improved whale IWOA algorithm is used for optimization of the hyperparameters in the random forest RF algorithm. Establishing a random forest algorithm model, and combining a plurality of CART decision tree models in parallel through a Bagging algorithm to form an integrated regressor, wherein the final result is as follows:
Figure BDA0003619157270000081
in the formula, H is the final integrated regressor, n is the number of single regressors selected from the original data set, HiThe decision result of each single regressor.
The CART decision tree is used for solving a regression problem, and MSE (mean square error) is defined as a data division basis when an RF algorithm trains a data set. Randomly selecting a part of data sets from the original data sets repeatedly through a Bagging algorithm, then respectively using the data set subsets selected each time as training sets of a plurality of weak regressors, training a plurality of weak learning models, and combining the weak learning models to form an integrated regressor.
An IWOA algorithm is provided, and compared with a WOA algorithm, the main improvement contents are as follows:
introducing nonlinear weight factors. In order to improve the globality of the whale algorithm, a nonlinear adaptive weight factor is provided:
Figure BDA0003619157270000082
where ω is a weight factor and k is an adjustment coefficient (k)>1) T is the current iteration number, ImaxIs the maximum number of iterations. After the nonlinear weight factor is introduced, the WOA algorithm formula is updated as follows:
Figure BDA0003619157270000091
Figure BDA0003619157270000092
in the formula (I), the compound is shown in the specification,
Figure BDA0003619157270000093
is the position vector of the current whale individual,
Figure BDA0003619157270000094
is the whale position vector with the highest current fitness,
Figure BDA0003619157270000095
is a vector of coefficients that is a function of,
Figure BDA0003619157270000096
and
Figure BDA0003619157270000097
is the value of the middle of the process,
Figure BDA0003619157270000098
a distance vector representing the current whale position and prey, b being a constant related to the spiral shape, l being [ -1,1 [ -1]The random number within the interval is a random number,
Figure BDA0003619157270000099
representing the current position vector of some random whale.
And (2) differentiating the variation disturbance term. Designing a differential variation disturbance term, wherein the coefficient of the differential variation disturbance term changes with the iteration number in a linear decreasing mode, and the term is defined as:
Figure BDA00036191572700000910
Figure BDA00036191572700000911
in the formula, Δ is a differential variation disturbance term, and h is a dynamic adjustment coefficient. After a differential variation disturbance term is introduced, the WOA algorithm formula is updated as follows:
Figure BDA00036191572700000912
and adjusting the search strategy in a self-adaptive manner. A probability threshold is designed such that the algorithm produces a set of solutions as randomly as possible within the initial global scope:
Figure BDA00036191572700000913
Figure BDA00036191572700000914
wherein Q is a probability threshold value,
Figure BDA00036191572700000915
as the mean fitness of the population, fminIs the worst fitness in the population, fmaxFor optimal fitness in the population, g is [0,1 ]]The random number within the interval is a random number,
Figure BDA00036191572700000916
and
Figure BDA00036191572700000917
are respectively as
Figure BDA00036191572700000918
Maximum and minimum values of. For each whale, use one [0,1 ]]The random number Q in between is compared with a probability threshold Q, if Q<Q, updating the positions of the whales according to the formula (12), wherein the positions of other whales are unchanged; otherwise, the positions of other whales are updated according to the formula (7).
(3) Dividing a data set into a training set and a testing set, training an algorithm model by using training set data based on an improved whale random forest IWOA-RF algorithm; substituting the test set model into the algorithm model, and quantitatively evaluating the prediction capability of the model by establishing an evaluation system.
Improvement of whale algorithm WOA and optimization of random forest RF, comprising the following steps:
setting an adjusting coefficient k in an IWOA algorithm as 2, and setting the population scale and the evolution times; MSE (mean square error) is used as an objective function of IWOA and a characteristic partition function of a CART tree in RF; and taking the number of CART trees in RF and the number of leaf nodes of each CART tree as the optimization objects of the IWOA algorithm.
The method comprises the steps of improving the establishment of a whale algorithm random forest IWOA-RF algorithm model, dividing 60% of data into a training set, training a network, dividing the other 40% of data into a test set, testing the accuracy of the network, and fusing an IWOA-RF lithium ion battery residual life prediction result of an equivalent circuit model as shown in figure 3. Evaluation indexes MAE (mean absolute error) and RMSE (root mean square error) were calculated as 9.8269 and 11.937, respectively.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.

Claims (9)

1. A lithium ion battery residual life prediction method is a random forest RF optimized by an improved whale algorithm IWOA based on a fusion equivalent circuit model, and is characterized in that: the method comprises the steps of establishing and fitting an equivalent circuit model, improving a whale algorithm WOA and optimizing random forest RF, improving a whale algorithm-random forest IWOA-RF algorithm to predict the residual life of the lithium ion battery, wherein the prediction process comprises the following steps:
(1) establishing an equivalent circuit model of the lithium ion battery, and fitting parameter values of equivalent circuit elements by performing EIS test on the equivalent circuit model;
(2) improving a whale algorithm, taking a random forest RF algorithm as a main body of a prediction algorithm, and using the improved whale IWOA algorithm for optimization of hyper-parameters in the random forest RF algorithm;
(3) dividing a data set into a training set and a testing set, training an algorithm model by using training set data based on an improved whale random forest IWOA-RF algorithm; substituting the test set model into the algorithm model, and quantitatively evaluating the prediction capability of the model by establishing an evaluation system.
2. The method for predicting the remaining life of the lithium ion battery according to claim 1, wherein the establishment of the equivalent circuit model comprises the following steps:
firstly, corresponding Nyquist diagrams of the lithium ion battery under different cycle times are measured, and corresponding equivalent circuit models of the lithium ion battery are designed step by step according to the characteristics of the Nyquist diagrams of different equivalent circuit elements;
fitting parameter values of each element in the equivalent circuit model through EIS data of the lithium ion battery so as to represent the current service life state of the lithium ion battery;
and thirdly, selecting EIS data and residual life data of the lithium ion battery with the available capacity of more than 70% of the rated capacity.
3. The method for predicting the remaining life of the lithium ion battery according to claim 1, wherein: the whale algorithm WOA simulates three stages of whale predation: an envelope phase, a foaming phase and a search phase.
4. The lithium ion battery remaining life prediction method of claim 1, wherein improving whale algorithm IWOA comprises the following improvement process:
introducing a nonlinear weight factor: in order to improve the globality of whale algorithm, a nonlinear adaptive weight factor is proposed:
Figure FDA0003619157260000011
where ω is a weight factor, k is an adjustment coefficient, and k is>1, t is the current iteration number, ImaxIs the maximum iteration number; after the nonlinear weight factor is introduced, the WOA algorithm formula is updated as follows:
Figure FDA0003619157260000012
Figure FDA0003619157260000021
in the formula (I), the compound is shown in the specification,
Figure FDA0003619157260000022
is the position vector of the current whale individual,
Figure FDA0003619157260000023
is the whale position vector with the highest current fitness,
Figure FDA0003619157260000024
is a vector of coefficients that is a function of,
Figure FDA0003619157260000025
and
Figure FDA0003619157260000026
is the value of the intermediate process, and,
Figure FDA0003619157260000027
a distance vector representing the current whale position and prey, b is a constant related to the spiral shape, 1 is [ -1,1 [ -1 [ ]]The random number within the interval is a random number,
Figure FDA0003619157260000028
a position vector representing a current random whale;
difference variation disturbance term: designing a differential variation disturbance term, wherein the coefficient of the differential variation disturbance term changes with the iteration number in a linear decreasing mode, and the term is defined as:
Figure FDA0003619157260000029
Figure FDA00036191572600000210
in the formula, delta is a differential variation disturbance term, and h is a dynamic adjustment coefficient; after a differential variation disturbance term is introduced, the WOA algorithm formula is updated as follows:
Figure FDA00036191572600000211
adaptive adjustment of search strategy: a probability threshold is designed such that the algorithm produces a set of solutions as randomly as possible within the initial global scope:
Figure FDA00036191572600000212
Figure FDA00036191572600000213
wherein Q is a probability threshold value,
Figure FDA00036191572600000218
as the mean fitness of the population, fminIs the worst fitness in the population, fmaxFor optimal fitness in the population, g is [0,1 ]]The random number within the interval is a random number,
Figure FDA00036191572600000215
and
Figure FDA00036191572600000216
are respectively as
Figure FDA00036191572600000217
Maximum and minimum values of; for each whale, use one [0,1 ]]Comparing the random number Q with a probability threshold value Q, if Q is less than Q, updating the position of the whale according to a formula (8), and keeping the positions of other whales unchanged; otherwise, updating the positions of other whales according to the formula (3).
5. The lithium ion battery residual life prediction method according to claim 1, wherein the improvement of whale algorithm WOA and the optimization of random forest RF comprise the following steps:
setting an adjusting coefficient k in an improved whale algorithm IWOA to be 2, and setting the population scale and the evolution times;
secondly, the mean square error MSE is used as an objective function of IWOA and a characteristic division function of a CART tree in random forest RF;
and thirdly, taking the number of CART trees in the RF and the leaf node number of each CART tree as optimization objects for improving the whale algorithm IWOA.
6. The method for predicting the remaining life of a lithium ion battery according to claim 1, wherein: the establishment of a whale algorithm random forest IWOA-RF algorithm model is improved, 60% of data is divided into a training set, a network is trained, the other 40% of data is divided into a testing set, the accuracy of the network is tested, and evaluation indexes are established.
7. The method for predicting the remaining life of a lithium ion battery according to claim 6, wherein: the evaluation indexes comprise an average absolute error MAE and a root mean square error RMSE, and the prediction effect of the multi-dimensional comprehensive evaluation algorithm is obtained.
8. The method for predicting the remaining life of a lithium ion battery according to claim 1, wherein: the step (1) is specifically as follows: firstly, researching the relation between the real part and the imaginary part of the impedance of a common equivalent circuit element on a frequency domain complex plane, deducing a function image of the impedance of each element, measuring corresponding Nyquist diagrams of the lithium ion battery under different cycle times, and designing corresponding equivalent circuit models step by step according to the characteristics of the Nyquist diagrams of different equivalent circuit elements; aiming at a Nyquist diagram of a tested lithium ion battery, an equivalent circuit model is designed to be in a form of LR (RQ) Q, wherein L is an inductor, R is a resistor, Q is a constant phase element, and the impedance of Q is represented as:
Figure FDA0003619157260000031
Figure FDA0003619157260000032
Figure FDA0003619157260000033
wherein Z is the impedance of Q, ZReIs the real part of impedance, ZImIs the imaginary part of the impedance, j is the unit of imaginary number, ω is the frequency, Y0And N is two parameters of the constant phase element Q, wherein N is a dispersion index, is dimensionless and takes a value in an interval (0, 1); y is0Has the unit of Ω-1·cm-2·s-NConstantly positive;
fitting each element parameter value in the equivalent circuit model by using software through EIS data of the lithium ion battery so as to represent the current service life state of the lithium ion battery; the method obtains the corresponding parameter value of the complete equivalent circuit model under each sampling cycle.
9. The method for predicting the remaining life of a lithium ion battery according to claim 1, wherein: the step (2) is specifically as follows: establishing a random forest algorithm model, and combining a plurality of CART decision tree models in parallel through a Bagging algorithm to form an integrated regressor, wherein the final result is as follows:
Figure FDA0003619157260000041
in the formula, H is a final integrated regressor, n is the number of single regressors selected from the original data set, and hi is a decision result of each single regressor;
the CART decision tree is used for solving a regression problem, and a Mean Square Error (MSE) is defined as a data division basis when an RF algorithm training data set is used; randomly selecting a part of data sets from the original data sets repeatedly through a Bagging algorithm, then respectively using the data set subsets selected each time as training sets of a plurality of weak regressors, training a plurality of weak learning models, and combining the weak learning models to form an integrated regressor.
CN202210452229.3A 2022-04-27 2022-04-27 Lithium ion battery residual life prediction method Active CN114755586B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210452229.3A CN114755586B (en) 2022-04-27 2022-04-27 Lithium ion battery residual life prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210452229.3A CN114755586B (en) 2022-04-27 2022-04-27 Lithium ion battery residual life prediction method

Publications (2)

Publication Number Publication Date
CN114755586A true CN114755586A (en) 2022-07-15
CN114755586B CN114755586B (en) 2023-08-01

Family

ID=82333880

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210452229.3A Active CN114755586B (en) 2022-04-27 2022-04-27 Lithium ion battery residual life prediction method

Country Status (1)

Country Link
CN (1) CN114755586B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115267558A (en) * 2022-09-30 2022-11-01 深圳先进技术研究院 Service life prediction and state estimation method of bidirectional lithium ion battery

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110295532A1 (en) * 2010-05-27 2011-12-01 Stukenberg Todd J Electronic storage battery diagnostic system
CN108389211A (en) * 2018-03-16 2018-08-10 西安电子科技大学 Based on the image partition method for improving whale Optimization of Fuzzy cluster
CN110568374A (en) * 2019-09-02 2019-12-13 东北电力大学 Lithium ion battery residual service life prediction method based on consideration of internal and external characteristics
CN110888059A (en) * 2019-12-03 2020-03-17 西安科技大学 Algorithm based on improved random forest combined cubature Kalman power battery state of charge estimation
CN111965562A (en) * 2020-10-20 2020-11-20 江苏慧智能源工程技术创新研究院有限公司 Method for predicting residual cycle life of lithium battery based on random forest model

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110295532A1 (en) * 2010-05-27 2011-12-01 Stukenberg Todd J Electronic storage battery diagnostic system
CN108389211A (en) * 2018-03-16 2018-08-10 西安电子科技大学 Based on the image partition method for improving whale Optimization of Fuzzy cluster
CN110568374A (en) * 2019-09-02 2019-12-13 东北电力大学 Lithium ion battery residual service life prediction method based on consideration of internal and external characteristics
CN110888059A (en) * 2019-12-03 2020-03-17 西安科技大学 Algorithm based on improved random forest combined cubature Kalman power battery state of charge estimation
CN111965562A (en) * 2020-10-20 2020-11-20 江苏慧智能源工程技术创新研究院有限公司 Method for predicting residual cycle life of lithium battery based on random forest model

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115267558A (en) * 2022-09-30 2022-11-01 深圳先进技术研究院 Service life prediction and state estimation method of bidirectional lithium ion battery

Also Published As

Publication number Publication date
CN114755586B (en) 2023-08-01

Similar Documents

Publication Publication Date Title
CN110135630B (en) Short-term load demand prediction method based on random forest regression and multi-step optimization
Mao et al. Optimal BP neural network algorithm for state of charge estimation of lithium-ion battery using PSO with Levy flight
CN110457789B (en) Lithium ion battery residual life prediction method
CN109001640B (en) Data processing method and device for power battery
Qin et al. State of health prediction for lithium-ion battery using a gradient boosting-based data-driven method
CN113702843B (en) Lithium battery parameter identification and SOC estimation method based on suburb optimization algorithm
CN110941929A (en) Battery health state assessment method based on ARMA and Elman neural network combined modeling
CN114970332B (en) Lithium battery model parameter identification method based on chaotic quantum sparrow search algorithm
CN113125960A (en) Vehicle-mounted lithium ion battery charge state prediction method based on random forest model
CN112269133A (en) SOC estimation method based on pre-charging circuit model parameter identification
CN115219906A (en) Multi-model fusion battery state of charge prediction method and system based on GA-PSO optimization
CN115271167B (en) BP neural network-based tire vulcanization quality RFV index prediction method and system
CN114755586A (en) Lithium ion battery residual life prediction method
CN116125278A (en) Lithium battery SOC estimation method and system based on LSTM-EKF algorithm
CN115963407A (en) ICGWO (intensive care unit) optimization ELM (element-based robust model) based lithium battery SOC estimation method
CN114280490B (en) Lithium ion battery state of charge estimation method and system
CN116774088A (en) Lithium ion battery health state estimation method based on multi-objective optimization
CN113884915A (en) Method and system for predicting state of charge and state of health of lithium ion battery
CN117007974A (en) Solid-state battery SOC estimation method based on model fusion
CN116859259A (en) Battery health state estimation method based on integrated learning and real vehicle big data
CN116662893A (en) Water quality prediction method for optimizing SVM (support vector machine) based on improved goblet sea squirt algorithm
Song et al. A hierarchical state of charge estimation method for lithium-ion batteries via xgboost and kalman filter
CN115792680A (en) Retired battery health state assessment method based on Bayes deep learning
CN115938496A (en) Quality Estimation Method Based on XGBoost Algorithm
CN110276476A (en) A kind of automobile engine smoke intensity prediction technique based on PSO-RBF neural network

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
GR01 Patent grant
GR01 Patent grant