CN114236414A - Satellite lithium battery life prediction method based on KCC-PF algorithm - Google Patents

Satellite lithium battery life prediction method based on KCC-PF algorithm Download PDF

Info

Publication number
CN114236414A
CN114236414A CN202111541174.5A CN202111541174A CN114236414A CN 114236414 A CN114236414 A CN 114236414A CN 202111541174 A CN202111541174 A CN 202111541174A CN 114236414 A CN114236414 A CN 114236414A
Authority
CN
China
Prior art keywords
lithium battery
capacity
particle
kcc
algorithm
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
CN202111541174.5A
Other languages
Chinese (zh)
Other versions
CN114236414B (en
Inventor
陈自强
崔显
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.)
Shanghai Jiaotong University
Original Assignee
Shanghai Jiaotong University
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 Shanghai Jiaotong University filed Critical Shanghai Jiaotong University
Priority to CN202111541174.5A priority Critical patent/CN114236414B/en
Publication of CN114236414A publication Critical patent/CN114236414A/en
Application granted granted Critical
Publication of CN114236414B publication Critical patent/CN114236414B/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]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • 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

Abstract

A satellite lithium battery remaining life prediction method based on a KCC-PF algorithm is characterized in that after a lithium battery capacity attenuation model is built and model parameters to be iterated are determined, state tracking of the lithium battery capacity attenuation model is carried out through a particle filter algorithm, and an initial particle set is built; updating model parameters of a lithium battery capacity attenuation model through a KCC-PF algorithm, and performing multi-step iterative prediction of capacity based on the updated lithium battery capacity attenuation model to obtain the residual life of the satellite lithium battery; and introducing a Kendel rank order correlation coefficient into resampling of particle filtering, and introducing historical data of the capacity into calculation of a current cycle while solving the problem of particle shortage, so as to guide weight distribution of particles and obtain a more accurate capacity predicted value, thereby improving the prediction accuracy of the RUL of the lithium battery. The residual service life of the satellite battery can be accurately predicted.

Description

Satellite lithium battery life prediction method based on KCC-PF algorithm
Technical Field
The invention relates to the technical field of lithium battery running state measurement, in particular to a satellite lithium battery residual life online prediction method based on a Kendel rank order correlation coefficient-particle filter (KCC-PF) algorithm.
Background
Lithium ion batteries have the advantages of long cycle life, high energy density, and the like, and have begun to become energy storage devices in the field of aviation. The lithium battery system is the only energy source of the satellite during the shadow period, and when the satellite cannot work normally due to battery aging, serious safety accidents and huge economic losses can be caused. The Remaining service Life (RUL) of the battery refers to the number of charge and discharge cycles that the battery undergoes from the current cycle to the time when the capacity is reduced to 80% of the initial capacity, and the State of health (SOH) of the battery is comprehensively reflected. Accurate RUL prediction can help satellite managers to master the working condition of the satellite lithium battery more clearly, and the reliability of a satellite energy storage system is greatly improved, so that the research on the RUL prediction technology of the satellite lithium battery has important significance.
The method is a common RUL prediction method at present, and the method utilizes a filtering algorithm to obtain model output in real time, deduces the performance attenuation trend of the model output and further completes RUL prediction. The current common battery degradation models include single-exponential, double-exponential, polynomial models, and the like, and the filtering algorithm is most widely applied as Particle Filter (PF).
The problems of weight degradation and particle shortage inevitably exist during the operation of the traditional PF algorithm, and when the PF algorithm is applied to the prediction of the RUL of the lithium battery, the accuracy and the reliability of a prediction result are seriously influenced. In addition, the current PF-based lithium battery RUL prediction method only estimates the capacity according to the particle distribution of the current cycle, and does not consider the guiding effect of historical real data on the estimation of the current state variable.
Disclosure of Invention
The invention aims to solve the problem of particle shortage after multi-step iteration of traditional particle filtering, and provides a satellite lithium battery service life prediction method based on an improved particle filtering algorithm.
The invention is realized by the following technical scheme:
the invention relates to a satellite lithium battery residual life prediction method based on a Kendel rank order correlation coefficient-particle filter (KCC-PF) algorithm, which comprises the following steps:
step 1, collecting capacity attenuation data of a lithium battery, and selecting a training battery for generating training data and a testing battery for testing.
The present embodiment selects 75% of the rated capacity of the battery as the failure threshold of the battery.
Step 2, constructing a lithium battery capacity attenuation model
Figure BDA0003414349050000021
Wherein: k is the current charge-discharge cycle number of the battery, CkFor battery capacity, the parameters to be iteratively updated include coulombic efficiency μC,kCoefficient of exponential term beta1,kAnd an exponential correction factor beta2,k。
Step 3, determining an initial range of the parameter to be iteratively updated: mu.sC，kHas an initial value of 0.997, beta1，k∈[0.3，1]，β2，k∈[1，10]。
And 4, setting a prediction Starting Point (SP) of the test battery.
Step 5, performing state tracking on L lithium battery capacity attenuation data before the predicted starting point through a Particle Filter (PF) algorithm to construct an initial particle set, which specifically comprises the following steps:
step 5.1, using the parameter mu to be iteratively updated of the lithium battery capacity attenuation modelC,k、β1,kAnd beta2,kA state space equation is constructed by taking the battery capacity as an observation variable as a state variable (z)
Figure BDA0003414349050000022
Wherein: state variable zk＝[μC，k，β1，k，β2，k]T。
And 5.2, sequentially sampling importance, and generating N samples from an importance density function q (z):
Figure BDA0003414349050000023
Figure BDA0003414349050000024
i.e. the particles are empty from stateAnd calculating the corresponding capacity value of each particle by using an intermediate equation. Calculating the weight of each particle according to the corresponding capacity value of each particle
Figure BDA0003414349050000025
Wherein: ckAnd Ck-1Representing the true capacity values of the k and k-1 cycles, τ being the observation noise δkThe variance of (a); normalization weight:
Figure BDA0003414349050000026
step 5.3, performing polynomial resampling to obtain a new particle set
Figure BDA0003414349050000027
Step 5.4, New particle Collection
Figure BDA0003414349050000028
Recorded as a column vector.
And 5.5, enabling k to be k +1, repeating the steps 5.2-5.4, and constructing the initial particle set of the L pieces of real capacity data before SP
Figure BDA0003414349050000029
Figure BDA00034143490500000210
Step 6, updating parameters to be iteratively updated of the lithium battery capacity attenuation model through a KCC-PF algorithm, and performing multi-step iterative prediction of the capacity based on the updated lithium battery capacity attenuation model, wherein the method specifically comprises the following steps:
step 6.1, generating new particles from the initial particle set and the state space equation obtained in step 5:
Figure BDA0003414349050000031
calculating importance weights for each particle
Figure BDA0003414349050000032
And normalizing the weights
Figure BDA0003414349050000033
Figure BDA0003414349050000034
Step 6.2, when the effective sample size
Figure BDA0003414349050000035
The resampling of step 6.3 is performed, otherwise no resampling is performed.
Step 6.3, resampling based on Kendall rank order correlation coefficient (KCC): calculating the current particle and the initial particle set constructed in step 5
Figure BDA0003414349050000036
And performing resampling on the KCCs, redistributing the weight values of the particles, and performing weighted summation to obtain a new state variable, namely the posterior estimated value of three parameters of the capacity fading model:
Figure BDA0003414349050000037
thereby obtaining an updated lithium battery capacity attenuation model.
Step 6.4, obtaining a one-step predicted value of the capacity based on the capacity attenuation model updated in the step 6.3
Figure BDA0003414349050000038
An approximate distribution of the predicted capacity is obtained from the particle distribution.
Step 6.5, update the initial particle set
Figure BDA0003414349050000039
Will be provided with
Figure BDA00034143490500000310
Regarded as historical data, and the newly acquired particle set is placed in the initial particle set
Figure BDA00034143490500000311
Last column of (3), delete
Figure BDA00034143490500000312
First line of (1), hold
Figure BDA00034143490500000313
Is constant.
Step 6.6, when the capacity predicted value
Figure BDA00034143490500000314
When the value is lower than the failure threshold value, the iteration step number k-SP is output as a prediction result of the residual service life, and meanwhile, the prediction distribution of the real residual service life (RUL) of the battery, namely RUL (Cycle), can be output according to the particle distributionEOL-CyclecurrentWherein: cycleEOLRepresents the number of charge-discharge cycles, at which the battery capacity reaches a failure thresholdcurrentRepresents the current charge-discharge cycle number of the battery; otherwise, let k be k +1, return to step 6.1, and enter the next iteration calculation.
Technical effects
Compared with the prior conventional technical means, the invention has the technical effects that:
1) through a resampling algorithm based on KCC, a brand-new particle can be generated in each step, and a proper weight is distributed to the particle, so that the diversity of the particle can be ensured even after multi-step iteration, and the problem of particle shortage of the traditional particle filtering algorithm is solved.
2) And introducing the historical reliable capacity data into the calculation of the current cycle for guiding the weight distribution of the particles to obtain a more accurate capacity predicted value, thereby improving the prediction precision of the residual service life of the lithium battery.
3) The lithium ion battery RUL prediction comparison test based on standard PF and KCC-PF algorithms is implemented based on the lithium ion battery aging test data of NASA's Ames prediction center (PCoE), and the results show that: when the prediction starting point is the 60 th cycle, the relative prediction error of the lithium battery RUL based on KCC-PF is within 10%, the relative error can be reduced to be within 3% as the prediction starting point moves backwards, and the error is reduced by about 50% compared with the prediction method based on PF.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic diagram of a capacity-decay curve and an initial prediction model of B0005;
FIG. 3 is a schematic diagram of a capacity fade curve and an initial prediction model of B0006;
FIG. 4 is a schematic diagram of a capacity fade curve and an initial prediction model of B0007;
FIG. 5 is a schematic diagram of a capacity fade curve and an initial prediction model of B00018;
FIG. 6 is a schematic diagram of the remaining service life prediction of B0005 at a prediction starting point of 60;
FIG. 7 is a schematic diagram of the remaining useful life prediction of B0006 at a starting point of prediction of 60;
FIG. 8 is a comparison diagram of the predicted remaining useful life of B0005 at the prediction start point of 60/80/100;
fig. 9 is a schematic diagram comparing the predicted remaining useful life results of B0006 at the prediction starting point of 60/80/100.
Detailed Description
The lithium battery aging test data used in the present embodiment is from the dums prediction center (PCoE) by NASA, and the test object is a 18650 type rechargeable battery with a rated capacity of 2 Ah. The aging test method comprises the steps of circularly charging, discharging and measuring impedance of each group of batteries at constant environmental temperature, and recording data such as battery voltage, current, impedance and the like in the experimental process.
When B0005 is taken as a test battery, B0006, B7 and B18 are taken as training batteries; with B0006 as the test cell, B0005, 7 and 18 will be used as training cells.
As shown in fig. 1, a satellite lithium battery life prediction method based on an improved particle filter algorithm according to this embodiment includes:
step 1, battery aging data is collated, a training battery and a testing battery are selected, and a capacity failure threshold value is set to be 1.38 Ah.
Step 2, constructing a lithium battery capacity attenuation model:
Figure BDA0003414349050000041
wherein: k is charge and dischargeNumber of cycles, CkCoulombic efficiency mu is used for the charging capacity of the kth charging and discharging period and the comprehensive influence of charging and discharging circulation on capacity attenuation in the aging process of the batteryC,kTo describe; the capacity regeneration phenomenon can occur when the battery is in standby for a long time, so that the capacity decay curve is locally raised, the characteristic is described by an exponential term of a capacity degradation model, delta tkThe time interval from the kth period to the (k + 1) th period is a parameter related to the actual charging and discharging frequency of the lithium battery, and can be regarded as beta in the model2,kCoefficient of (2) without taking Δ tkTo a constant value of 1, Δ t is calculatedkIs included in beta2,kThe calculation result is not affected in the adjustment of (2). This example is for muC,k、β1,kAnd beta2,kAnd performing iterative updating.
Step 3, determining initial parameters of the model: fitting models for B0005, 6, 7 and 18 were calculated, respectively. Initial parameter mu of modelC,k、β1,kAnd beta2,kIt cannot be fitted directly from historical data, but three parameters have their empirical ranges: mu.sC,kMay take a value of about 0.997, beta1,k∈[0.3,1]，β2,k∈[1,10]And determining three parameters by adopting a combined optimization mode.
And 4, determining a prediction Starting Point (SP).
In the present embodiment, for predicting the remaining service life of the battery, the capacity fading condition after the predicted starting point SP is more concerned, and when performing optimization tests of different combinations on three parameters, the Root Mean Square Error (RMSE) of the sample point fitting after the predicted starting point is used as an index of model accuracy:
Figure BDA0003414349050000051
when B0005 is taken as a test battery, that is, B0005 is taken as a battery to be predicted, and the capacity attenuation data after the prediction starting point is an unknown quantity, the initial prediction model is determined by the parameter average value of the fitting models of B0006, 7 and 18. Similarly, the initial prediction model of B0006 was determined from the parameter averages of the fitted models of B0005, 7 and 18.
The finally determined model parameters are shown in table 1:
battery numbering μC β1 β2 RMSE/Ah
B5 0.9968 0.484 6.700 0.0223
B6 0.9965 0.468 6.660 0.0346
B7 0.9969 0.484 5.860 0.0252
B18 0.9966 0.493 5.997 0.0391
Average value of B6, 7, 18 0.9967 0.4817 6.1723 0.0271
Average value of B5, 7, 18 0.9968 0.4870 6.1857 0.0523
Average value of B5, 6, 18 0.9967 0.4817 6.4523 0.0673
Average value of B5, 6, 7 0.9968 0.4787 6.4067 0.0416
The graphs of the respective models are shown in fig. 2-5.
Step 5, PF state tracking, and constructing an initial particle set, which specifically comprises:
step 5.1, constructing an initial particle set by utilizing a tracking function of particle filtering, wherein a state space equation is as follows:
Figure BDA0003414349050000052
wherein: state variable zk＝[μC,k,β1,k,β2,k]T，εk＝[εμ,k,εβ1,k,εβ2,k]TAs process noise, its covariance matrix is
Figure BDA0003414349050000053
δkFor observing noise, the variance is τ2Let all noise be white gaussian noise with variance 0.0001.
Step 5.2, sequential importance sampling: let the importance density function q (z)k)＝p(zk|zk-1) I.e. gaussian distribution with the initial parameters as desired and the system noise variance as variance. N samples are generated from the distribution:
Figure BDA0003414349050000054
Figure BDA0003414349050000055
i.e. a collection of particles. Calculating the capacity value corresponding to each particle according to the state space equation, and calculating the weight of each particle according to the capacity value corresponding to each particle
Figure BDA0003414349050000056
Wherein: ckAnd Ck-1The real capacity values of the k and k-1 circulation are shown, i represents the particle number; normalization weight:
Figure BDA0003414349050000061
step 5.3, standard polynomial resampling, from interval (0, 1)]Uniformly sampling to obtain N independent sample sets with the same distribution
Figure BDA0003414349050000062
When in use
Figure BDA0003414349050000063
Then, the sequence number function l (U (i)) m is substituted for each U (i) in U to obtain a sequence number set
Figure BDA0003414349050000064
When the weight of a particle is larger, the corresponding interval
Figure BDA0003414349050000065
The larger the probability that u (i) falls in this section, the larger the number of sequence numbers m in L. The total number of the equal serial numbers in the L is counted, so as to obtain a set
Figure BDA0003414349050000066
Mixing the particles
Figure BDA0003414349050000067
Replication of niNext, a new set of particles is obtained:
Figure BDA0003414349050000068
step 5.4, New particle Collection
Figure BDA0003414349050000069
Recorded as a column vector.
And 5.5, enabling k to be k +1, repeating the steps 5.2-5.4, and constructing an initial particle set based on L real capacity data before SP
Figure BDA00034143490500000610
Figure BDA00034143490500000611
Wherein:
Figure BDA00034143490500000612
the set of particles for the L cycles before SP,
Figure BDA00034143490500000613
is the state vector of the state space equation, i.e. the parameters of the battery capacity fade model:
Figure BDA00034143490500000614
i is the particle number and j is the number of charge-discharge cycles.
Step 6, operating a KCC-PF algorithm, updating model parameters, and performing multi-step iterative prediction of capacity, wherein the method specifically comprises the following steps:
step 6.1, updating the state: new particles are generated from the initial set of particles and the state-transfer equation:
Figure BDA00034143490500000615
calculating importance weights for each particle
Figure BDA00034143490500000616
And normalizing the weights
Figure BDA00034143490500000617
Step 6.2, when the effective sample size
Figure BDA00034143490500000618
The resampling of step 6.3 is performed, otherwise no resampling is performed.
Step 6.3, resampling based on Kendel rank order correlation coefficient (KCC) is carried out to obtain new weight, and the new weight specifically comprises the following steps:
i) preparing an initial set of particles
Figure BDA00034143490500000619
ii) calculating the current particle
Figure BDA00034143490500000620
And initial particle set
Figure BDA00034143490500000621
The positive rank correlation coefficient between the two is specifically as follows: the ratio of the difference between the number P of data pairs with the same consistency and the number Q of data pairs with different consistency to the total combination number L (L-1)/2 in all the combinations of the data pairs, namely
Figure BDA00034143490500000622
Wherein: n, 1, 2. Wherein the same identity means: the actual capacity value and the estimated capacity value are combined pairwise in time sequence:
Figure BDA00034143490500000623
wherein: i represents the serial number of the particles, and N is the total number of the particles; for two sets of data pairs
Figure BDA0003414349050000071
And
Figure BDA0003414349050000072
when s is<t and satisfy xs＜xtAnd is
Figure BDA0003414349050000073
Or xs＞xtAnd is
Figure BDA0003414349050000074
The two data pairs are said to be identical in consistency; when x is satisfieds＜xtAnd is
Figure BDA0003414349050000075
Or xs＞xtAnd is
Figure BDA0003414349050000076
The two data pairs are said to be of different consistency; when x is satisfieds＝xtOr
Figure BDA0003414349050000077
The two data pairs are said to not have consistency. When there is no data pair without consistency, then there is
Figure BDA0003414349050000078
It can be inferred that the value range of KCC is [ -1,1 [)]。
Preferably, to transform the KCC into the positive range, an exponential function with parameters is taken
Figure BDA0003414349050000079
The KCC is treated: when 0 is present<α<1, alpha. KCCkCentered toward the origin, parameter γkThe value range of (A) tends to be reduced, and the corresponding particle weight is concentrated; when alpha is>1, alpha. KCCkDiverging from the origin to the two ends, parameter gammakThe value range of (a) tends to be enlarged, and the corresponding particle weights are more dispersed.
In this embodiment, the parameter α is 10.
iii) reassigning the particle weights to:
Figure BDA00034143490500000710
step 6.4, the posterior estimation of the state variables is realized through weighted summation
Figure BDA00034143490500000711
One-step prediction of ultimate realized capacity
Figure BDA00034143490500000712
Step 6.5, update the initial particle
Figure BDA00034143490500000713
Collecting: will be provided with
Figure BDA00034143490500000714
Regarding as historical data and placing its particle set in the initial particle
Figure BDA00034143490500000715
Last column of (3), delete
Figure BDA00034143490500000716
The first list of (a).
Step 6.6, when the capacity predicted value
Figure BDA00034143490500000717
When the output iteration number k-SP is lower than the failure threshold value, the output iteration number k-SP is the residual service life of the battery, and meanwhile, the approximate distribution of the predicted capacity can be obtained(ii) a Otherwise, let k be k +1, repeat steps 6.1-6.5, and enter the next iteration.
The test cells were B0005 and B0006, respectively, with a failure threshold of 1.38 Ah.
As shown in fig. 6 and 7, in order to predict the starting point SP of 60, the lithium battery RUL is predicted based on the conventional PF algorithm and the KCC-PF algorithm.
TABLE 2 prediction error
Figure BDA00034143490500000718
RUL is the actual remaining life of the battery, and PRUL is the predicted value of RUL.
The prediction error of the RUL prediction algorithm based on the KCC-PF is lower than that of the PF, and the probability density distribution function (PDF) width is narrower, so that the KCC-PF improves the prediction accuracy of the PF, improves the problem of particle shortage, and reduces the uncertainty of a prediction result.
The prediction accuracy can be improved by moving the prediction starting point backwards.
As shown in fig. 8 and 9, the prediction results of the lithium battery RUL based on the KCC-PF algorithm are obtained when the starting point SP is predicted to be 80 or 100.
Table 3 shows the prediction error
Figure BDA0003414349050000081
The prediction accuracy is obviously improved by moving the prediction starting point backwards, which shows that the prediction accuracy of the RUL obtained by the invention is improved along with the increase of the aging degree of the battery.
The foregoing embodiments may be modified in many different ways by those skilled in the art without departing from the spirit and scope of the invention, which is defined by the appended claims and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

Claims (5)

1. A satellite lithium battery remaining life prediction method based on a KCC-PF algorithm is characterized in that after a lithium battery capacity attenuation model is constructed and model parameters to be iterated are determined, state tracking of the lithium battery capacity attenuation model is carried out through a particle filter algorithm, and an initial particle set is constructed; updating model parameters of a lithium battery capacity attenuation model through a KCC-PF algorithm, and performing multi-step iterative prediction of capacity based on the updated lithium battery capacity attenuation model to obtain the residual life of the satellite lithium battery;
the lithium battery capacity attenuation model
Figure FDA0003414349040000011
Wherein: k is the current charge-discharge cycle number of the battery, delta tkIs the time interval from the kth cycle to the (k + 1) th cycle, CkFor battery capacity, the parameters to be iteratively updated include coulombic efficiency μC,kCoefficient of exponential term beta1,kAnd an exponential correction factor beta2,k；
The residual service life of the satellite lithium battery RUL is equal to CycleEOL-CyclecurrentWherein: cycleEoLRepresents the number of charge-discharge cycles, at which the battery capacity reaches a failure thresholdcurrentIndicating the current number of charge and discharge cycles of the battery.
2. The method for predicting the residual life of a satellite lithium battery based on the KCC-PF algorithm as claimed in claim 1, wherein the state tracking specifically comprises:
step 5.1, using the parameter mu to be iteratively updated of the lithium battery capacity attenuation modelC,k、β1,kAnd beta2,kA state space equation is constructed by taking the battery capacity as an observation variable as a state variable (z)
Figure FDA0003414349040000012
Wherein: state variable zk＝[μC，k，β1，k，β2，k]T；
And 5.2, sequentially sampling importance, and generating N samples from an importance density function q (z):
Figure FDA0003414349040000013
Figure FDA0003414349040000014
namely particles, calculating the capacity value corresponding to each particle by a state space equation; calculating the weight of each particle according to the corresponding capacity value of each particle
Figure FDA0003414349040000015
Wherein: ckAnd Ck-1Representing the true capacity values of the k and k-1 cycles, τ being the observation noise δkThe variance of (a);
step 5.3, performing polynomial resampling to obtain a new particle set
Figure FDA0003414349040000016
Step 5.4, New particle Collection
Figure FDA0003414349040000017
Recording as a column vector;
and 5.5, enabling k to be k +1, repeating the steps 5.2-5.4, and constructing an initial particle set based on L real capacity data before SP
Figure FDA0003414349040000018
Figure FDA0003414349040000021
3. The satellite lithium battery residual life prediction method based on the KCC-PF algorithm as claimed in claim 1, wherein the multi-step iterative prediction specifically comprises:
step 6.1, generating new particles from the initial particle set and the state transfer equation:
Figure FDA0003414349040000022
calculating importance weights for each particle
Figure FDA0003414349040000023
And normalizing the weights
Figure FDA0003414349040000024
Step 6.2, when the effective sample size
Figure FDA0003414349040000025
Performing the resampling of step 6.3, otherwise not performing the resampling;
step 6.3, resampling based on Kendall rank order correlation coefficient KCC: calculating the current particle and the initial particle set
Figure FDA0003414349040000026
KCC in between, resampling, and redistributing weight of particles
Figure FDA0003414349040000027
And weighting and summing to obtain a new state variable, namely the posterior estimated value of three parameters of the capacity fading model:
Figure FDA0003414349040000028
thereby completing the update of the capacity attenuation model of the lithium battery;
step 6.4, obtaining a one-step predicted value of the capacity by using the observation equation and the expectation formula
Figure FDA0003414349040000029
Obtaining an approximate distribution of the predicted capacity from the particle distribution;
step 6.5, update the initial particle set
Figure FDA00034143490400000210
Will be provided with
Figure FDA00034143490400000211
Regarding the data as historical data, and placing the particle set in the initial particle set
Figure FDA00034143490400000212
Deleting the first column while deleting the last column;
step 6.6, when the capacity predicted value
Figure FDA00034143490400000213
When the output iteration number k-SP is lower than the failure threshold value, the output iteration number k-SP is the residual service life of the battery, and meanwhile, the approximate distribution of the predicted capacity can be obtained; otherwise, let k be k +1, repeat steps 6.1-6.5, and enter the next iteration.
4. The method for predicting the residual life of the satellite lithium battery based on the KCC-PF algorithm according to claim 4, wherein the indexes of the accuracy of the lithium battery capacity attenuation model are as follows:
Figure FDA00034143490400000214
5. the method for predicting the remaining life of a satellite lithium battery based on the KCC-PF algorithm as claimed in claim 4, wherein in order to transform the KCC into a positive number range, an exponential function with parameters is taken
Figure FDA00034143490400000215
The KCC is treated: when 0 is present<α<1, alpha. KCCkCentered toward the origin, parameter γkThe value range of (A) tends to be reduced, and the corresponding particle weight is concentrated; when alpha is>1, alpha. KCCkDiverging from the origin to the two ends, parameter gammakThe value range of (a) tends to be enlarged, and the corresponding particle weights are more dispersed.
CN202111541174.5A 2021-12-16 2021-12-16 Satellite lithium battery life prediction method based on KCC-PF algorithm Active CN114236414B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111541174.5A CN114236414B (en) 2021-12-16 2021-12-16 Satellite lithium battery life prediction method based on KCC-PF algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111541174.5A CN114236414B (en) 2021-12-16 2021-12-16 Satellite lithium battery life prediction method based on KCC-PF algorithm

Publications (2)

Publication Number Publication Date
CN114236414A true CN114236414A (en) 2022-03-25
CN114236414B CN114236414B (en) 2023-02-10

Family

ID=80756944

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111541174.5A Active CN114236414B (en) 2021-12-16 2021-12-16 Satellite lithium battery life prediction method based on KCC-PF algorithm

Country Status (1)

Country Link
CN (1) CN114236414B (en)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150349385A1 (en) * 2014-04-01 2015-12-03 Medtronic, Inc. Method and System for Predicting Useful Life of a Rechargeable Battery
CN106845866A (en) * 2017-02-27 2017-06-13 四川大学 Equipment method for predicting residual useful life based on improved particle filter algorithm
CN107918103A (en) * 2018-01-05 2018-04-17 广西大学 A kind of lithium ion battery residual life Forecasting Methodology based on grey particle filter
CN108205114A (en) * 2017-12-29 2018-06-26 上海电气集团股份有限公司 The Forecasting Methodology and system of battery life
CN108535656A (en) * 2018-03-22 2018-09-14 中北大学 Lithium ion battery remaining life prediction technique and system based on PCA-NARX neural networks
CN110457789A (en) * 2019-07-25 2019-11-15 桂林电子科技大学 A kind of lithium ion battery residual life prediction technique merged based on improvement particle filter with double exponential decay experience physical models
CN110703120A (en) * 2019-09-29 2020-01-17 上海海事大学 Lithium ion battery service life prediction method based on particle filtering and long-and-short time memory network
CN112986831A (en) * 2021-04-30 2021-06-18 上海海事大学 Lithium ion battery life prediction method based on correlation coefficient particle filtering
CN113255199A (en) * 2021-04-09 2021-08-13 南京工程学院 Battery remaining life prediction method based on particle filtering

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150349385A1 (en) * 2014-04-01 2015-12-03 Medtronic, Inc. Method and System for Predicting Useful Life of a Rechargeable Battery
CN106845866A (en) * 2017-02-27 2017-06-13 四川大学 Equipment method for predicting residual useful life based on improved particle filter algorithm
CN108205114A (en) * 2017-12-29 2018-06-26 上海电气集团股份有限公司 The Forecasting Methodology and system of battery life
CN107918103A (en) * 2018-01-05 2018-04-17 广西大学 A kind of lithium ion battery residual life Forecasting Methodology based on grey particle filter
CN108535656A (en) * 2018-03-22 2018-09-14 中北大学 Lithium ion battery remaining life prediction technique and system based on PCA-NARX neural networks
CN110457789A (en) * 2019-07-25 2019-11-15 桂林电子科技大学 A kind of lithium ion battery residual life prediction technique merged based on improvement particle filter with double exponential decay experience physical models
CN110703120A (en) * 2019-09-29 2020-01-17 上海海事大学 Lithium ion battery service life prediction method based on particle filtering and long-and-short time memory network
CN113255199A (en) * 2021-04-09 2021-08-13 南京工程学院 Battery remaining life prediction method based on particle filtering
CN112986831A (en) * 2021-04-30 2021-06-18 上海海事大学 Lithium ion battery life prediction method based on correlation coefficient particle filtering

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
GAO D J等: "A Method for Predicting the Remaining Useful Life of Lithium-Ion Batteries Based on Particle Filter Using Kendall Rank Correlation Coefficient", 《ENERGIES》 *
崔显 等: "基于 KCC-PF 的锂离子电池剩余使用寿命预测", 《装备环境工程》 *
张吉宣等: "电动汽车供电系统锂电池剩余寿命预测", 《电子测量与仪器学报》 *
王帅等: "基于粒子滤波的锂离子电池剩余寿命预测", 《电源技术》 *
韦海燕等: "新陈代谢灰色粒子滤波实现电池剩余寿命预测", 《电工技术学报》 *

Also Published As

Publication number Publication date
CN114236414B (en) 2023-02-10

Similar Documents

Publication Publication Date Title
CN113009349B (en) Lithium ion battery health state diagnosis method based on deep learning model
CN111443293A (en) Lithium battery state of health (SOH) estimation method based on data driving
CN107741568B (en) Lithium battery SOC estimation method based on state transition optimization RBF neural network
CN113740736B (en) Electric vehicle lithium battery SOH estimation method based on deep network self-adaption
CN112986831A (en) Lithium ion battery life prediction method based on correlation coefficient particle filtering
CN113917334B (en) Battery health state estimation method based on evolution LSTM self-encoder
CN111753416A (en) Lithium ion battery RUL prediction method based on two-stage Wiener process
CN112684363A (en) Lithium ion battery health state estimation method based on discharge process
CN113189490B (en) Lithium battery health state estimation method based on feature screening and Gaussian process regression
CN113109715A (en) Battery health condition prediction method based on feature selection and support vector regression
CN111832221A (en) Lithium battery life prediction method based on feature screening
CN110703120A (en) Lithium ion battery service life prediction method based on particle filtering and long-and-short time memory network
CN111175666A (en) SOH detection method and device
CN112989690A (en) Multi-time scale state of charge estimation method for lithium battery of hybrid electric vehicle
CN115267555A (en) Battery SOH (State of health) evaluation system of energy storage system based on battery multipoint temperature measurement
CN112834927A (en) Lithium battery residual life prediction method, system, device and medium
CN115201686B (en) Lithium ion battery health state assessment method under incomplete charge and discharge data
CN113030761A (en) Method and system for evaluating health state of battery of super-large-scale energy storage power station
CN111950205A (en) Lithium battery SOH prediction method based on FWA optimization extreme learning machine
CN114384435A (en) WSA-LSTM algorithm-based self-adaptive prediction method for residual service life of new energy automobile power battery
Xu et al. State of health diagnosis and remaining useful life prediction of lithium-ion batteries based on multi-feature data and mechanism fusion
CN114839538A (en) Method for extracting degradation characteristics of lithium ion battery for estimating residual life
CN116298914A (en) Lithium battery state of charge and health state joint estimation method based on deep learning
CN116754959A (en) SOC estimation method based on improved GWO optimized forgetting factor on-line parameter identification
CN115219918A (en) Lithium ion battery life prediction method based on capacity decline combined model

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