CN114545274A - Lithium battery residual life prediction method - Google Patents

Lithium battery residual life prediction method Download PDF

Info

Publication number
CN114545274A
CN114545274A CN202210094514.2A CN202210094514A CN114545274A CN 114545274 A CN114545274 A CN 114545274A CN 202210094514 A CN202210094514 A CN 202210094514A CN 114545274 A CN114545274 A CN 114545274A
Authority
CN
China
Prior art keywords
model
lithium battery
training
health factor
health
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210094514.2A
Other languages
Chinese (zh)
Inventor
李祖欣
张宗杰
蔡志端
钱懿
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.)
Huzhou University
Original Assignee
Huzhou 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 Huzhou University filed Critical Huzhou University
Priority to CN202210094514.2A priority Critical patent/CN114545274A/en
Publication of CN114545274A publication Critical patent/CN114545274A/en
Pending legal-status Critical Current

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
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/60—Other road transportation technologies with climate change mitigation effect
    • Y02T10/70—Energy storage systems for electromobility, e.g. batteries

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Abstract

The invention discloses a method for predicting the residual life of a lithium battery, which comprises the following steps: firstly, performing offline modeling, namely collecting offline data of a lithium battery, extracting a health factor sample set, performing weight analysis on the health factor sample set by using a random forest algorithm, determining selected health factor samples, performing BiLSTM network model training, selecting optimal hyper-parameters of a model through Bayesian optimization, and constructing a prediction model; II, secondly: and (3) performing online prediction, namely obtaining a health factor sample set corresponding to the feature selection of an offline stage through online data of the lithium battery. And then, using the prediction model in the step one to predict the service life of the lithium battery. The method can reduce the number of parameters, reduce the complexity of parameter training, reduce the loss caused by the failure of the lithium battery, improve the safety of the lithium battery and solve the problems of redundancy and deficiency in the selection of the health factors of the lithium battery and the complexity in the selection of different hyper-parameters of the neural network while maintaining the prediction accuracy of the neural network.

Description

Lithium battery residual life prediction method
Technical Field
The invention belongs to the technical field of lithium batteries, and particularly relates to a method for predicting the residual life of a lithium battery.
Background
Since the advent of lithium batteries, lithium batteries have been widely used in electronic products, aerospace, electric vehicles, and other devices because of their excellent electrochemical properties. However, during the use process, the capacity of the lithium ion battery gradually decreases with the increase of the cycle number, thereby bringing about the problems of safety and reliability. And the accurate prediction of the remaining service life of the lithium ion battery is of great significance to the safe operation of a power supply system.
The current commonly used lithium battery life prediction methods include: traditional modeling, electrochemical modeling, data-driven. Because the chemical reaction inside the lithium battery is very complicated, the established model is often too complicated and the model is difficult to obtain. In addition, the influence factors of the service life of the lithium battery are more, and the prediction precision of the service life of the lithium battery can be influenced by different temperatures and charge-discharge depths, which are not considered by the traditional model method. The data driving method comprises the steps of analyzing and processing historical charging and discharging data of the lithium battery, mining battery information by utilizing the historical data, and predicting the residual service life of the lithium battery by combining a machine learning algorithm. The bidirectional long-short term memory neural network (BilSTM) has an attention focusing mechanism, can better extract time sequence information in historical data, avoids gradient disappearance and gradient explosion at the same time, and has a better application prospect in the field of lithium battery service life prediction. However, the current data-driven method has the following two problems: (1) the selection of the health factors has redundancy and insufficiency, and the precision of the prediction of the residual life of the lithium battery is influenced; (2) the accuracy of the model is affected by different super-parameter selections of the neural network, and a large number of super-parameter tests are usually needed to determine the final model parameters, which is time-consuming and labor-consuming.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a lithium battery residual life prediction method integrating random forest health factor selection and Bayesian optimization BiLSTM. The method comprises the following steps:
the method comprises the following steps: performing offline modeling, namely collecting offline data of the lithium battery, extracting a lithium battery degradation characteristic sequence to obtain a health factor sample set, and performing weight analysis on the health factor sample set by using a random forest algorithm to determine the selected health factor sample according to the occupied weight; adding the selected samples into training data in sequence, randomly generating initialization parameters, and substituting the initialization parameters into a BilSTM model for training; when the precision of the training result does not meet the requirement, the Bayesian optimization model is used for: firstly, establishing a Gaussian process regression model, solving an acquisition function expression, solving a next group of model parameters according to an acquisition function until an optimal hyper-parameter meeting the precision requirement is found, and constructing a prediction model;
step two: in the online prediction, a health factor sample set is obtained by collecting online data of the lithium battery and corresponding to the feature selection in an offline stage; and then, using a BilSTM prediction model established in an off-line stage to predict the service life of the lithium battery.
Preferably, the offline training process in step one is as follows:
step 11: collecting off-line data of the lithium battery, including current, voltage and temperature in the charging and discharging process, and extracting degradation characteristics; obtaining the constant pressure rise charging time teq-vcCharging voltage rise V at equal time intervalseq-tcConstant current charging duration thcEqual time interval charging current drop Ieq-tcVolume increase curve ic, initial temperature T0Mean temperature TavgMedian temperature TmidMaximum temperature TmaxAverage rate of temperature change TrateConstant voltage drop discharge time teq-vdEqual time interval discharge voltage drop Veq-tdMean voltage decay MVF, voltage sample entropy VSampEn(ii) a The calculation formula of each degradation characteristic is as follows:
Figure BDA0003490526770000021
Figure BDA0003490526770000022
Figure BDA0003490526770000023
Figure BDA0003490526770000024
wherein T is temperature, VnIs a nominal voltage of 4.2V, K is the sampling frequency, Bm(r) is the probability that two voltage sequences match m points with a similar tolerance r, Am(r) is the probability that two voltage sequences match m +1 points;
step 12: when all the health factor sample sets are obtained, the health factor X is extracted1，X2，X3，...,XqStacking into a matrix to obtain a health factor sample training set X belonging to Rp×qWherein p represents the number of training samples, and q represents the number of health factors;
step 13: selecting health factors by using a random forest algorithm, extracting the health factors in a training set X by using a sampling and putting back method, randomly sampling n samples, and generating t training sets; generating a corresponding decision tree by the training set, randomly and repeatedly selecting d features at each generated node, dividing the training set Y by the d features, finding the optimal division feature by using a Gini coefficient, and repeating the steps until a random forest model is obtained; the Gini index is expressed by GI, and the Gini index score VIM of each health factor is calculated(Gini)The Gini index is calculated by the formula:
Figure BDA0003490526770000031
wherein K represents K categories, pdkRepresenting the proportion of the category k in the node d; health factor XjThe importance at node m is:
Figure BDA0003490526770000032
GIland GIrRespectively representing Gini indexes of two nodes after branching;
health factor XjThe node appearing in decision tree i is set M, then XjThe importance in the ith tree is:
Figure BDA0003490526770000033
if there are n trees in the random forest model, then:
Figure BDA0003490526770000034
and (3) normalizing all the obtained importance scores to obtain the importance scores of the health factors:
Figure BDA0003490526770000035
step 14: selecting parameters to be optimized of a BiLSTM model, and setting m initial hyper-parameters Xm＝[X1,X2,...,Xm]Substituting the model into a BilSTM model for training; to obtain XmResult Y on the objective functionm＝[Y1,Y2,...,Ym]Building a matrix D { (X)1,Y1)...(Xm,Ym) }; verifying whether the initial model meets the precision requirement on the training set, and if so, and if the iteration number is less than the maximum iteration number, the hyper-parameter of the model is XmAnd if not, carrying out Bayesian optimization loop iteration. The Bayesian optimization process is as follows: and establishing a Gaussian process regression model by using the matrix D. The next group of hyper-parameters X is carried out through the collection function g (X)m+1And (4) calculating. Substituting the new hyper-parameters into the BilSTM model to perform model training and calculate the result Y of the objective functionmAnd judging whether the precision requirement is met.
Preferably, the online detection process in step two is as follows:
step 21: using the health factor with the importance greater than 0.3 calculated by the random forest model in the step 13 of the off-line training stage in the step one, and acquiring corresponding health factor data on line;
step 22: establishing an online prediction model by using Bayesian optimization BiLSTM model parameters in step 14 of the offline training stage in the step one;
step 23: inputting corresponding health factor data acquired on line, updating the cell state through a forgetting gate and an input-output gate, acquiring the time sequence relation between the previous moment and the subsequent moment, and finally acquiring a final lithium battery residual life prediction result through a full connection layer: if the predicted battery capacity is less than 80%, the end of the battery life is determined.
The invention has the beneficial effects that: the invention selects the health factors by using the random forest algorithm in the training stage, solves the problem of redundancy in the selection of the health factors and reduces the training amount. And the network parameters are determined by Bayesian optimization, so that the time for selecting the super parameters is shortened, and the prediction precision of the residual life of the lithium battery is improved.
Drawings
FIG. 1 is a block flow diagram of the present invention;
FIG. 2 is a comparison of the predicted results of the method of the present invention and the prior art;
fig. 3 is a comparison of the residual results of the inventive method and the prior art.
Detailed Description
The invention relates to a lithium battery residual life prediction method based on random forest health factor selection and Bayesian optimization BiLSTM, which comprises the following steps:
the method comprises the following steps: and (3) performing offline modeling, collecting offline data such as current, voltage, temperature and the like in the charging and discharging process of the lithium battery, and extracting a lithium battery degradation characteristic sequence to obtain a health factor sample set. And then, carrying out weight analysis on the health factor sample set by using a random forest algorithm to determine the selected health factor sample according to the occupied weight. And adding the selected samples into a training set in sequence, randomly generating initialization parameters, substituting the initialization parameters into a BilSTM model for training, solving an acquisition function expression by establishing a Gaussian process regression model when the precision of a training result does not meet the requirement, solving the next group of model parameters according to the acquisition function until the optimal hyper-parameter meeting the precision requirement is found, and constructing a prediction model.
Step two: and (4) online prediction. And acquiring corresponding health factor data on line by calculating a health factor with the importance greater than 0.3 through a random forest model, establishing an on-line prediction model by using BiLSTM model parameters optimized by Bayesian optimization in an off-line stage, inputting the corresponding health factor data acquired on line, and judging that the service life of the battery is finished if the predicted battery capacity is lower than 80%.
The method comprises the following steps: the specific off-line modeling process of off-line modeling is as follows:
1) and extracting degradation characteristics according to the collected current, voltage and temperature data of the lithium battery in the charging and discharging process. Obtaining the constant pressure rise charging time teq-vcCharging voltage rise V at equal time intervalseq-tcConstant current charging duration thcEqual time interval charging current drop Ieq-tcVolume increase curve ic, initial temperature T0Mean temperature TavgMedian temperature TmidMaximum temperature TmaxAverage rate of temperature change TrateConstant voltage drop discharge time teq-vdEqual time interval discharge voltage drop Veq-tdMean voltage decay MVF, voltage sample entropy VSampEn；
The calculation formula of each degradation characteristic is as follows:
Figure BDA0003490526770000051
Figure BDA0003490526770000052
Figure BDA0003490526770000053
Figure BDA0003490526770000054
wherein T is temperature, VnIs a nominal voltage of 4.2V, K is the sampling frequency, Bm(r) is the probability that two voltage sequences match m points with a similar tolerance r, Am(r) is two electronsProbability of matching m +1 points of the press sequence;
2) when all the health factor sample sets are obtained, the health factor X is used1，X2，X3，...,XqStacking into a matrix to obtain a health factor sample training set X belonging to Rp×qWherein p represents the number of training samples, and q represents the number of health factors;
3) selecting health factors by using a random forest algorithm, extracting the health factors in a training set X by using a sampling and returning method (bootstrap), and randomly sampling n samples to generate t training sets; and generating a corresponding decision tree by the training set, randomly and repeatedly selecting d features at each generated node, dividing the training set Y by using the d features, finding the optimal division feature by using the Gini coefficient, and repeating the steps until a random forest model is obtained. The Gini index is expressed by GI, and the Gini index score VIM of each health factor is calculated(Gini)The Gini index is calculated as:
Figure BDA0003490526770000055
wherein K represents K categories, pdkRepresenting the proportion of the class k in the node d;
health factor XjThe importance at node m is:
Figure BDA0003490526770000056
GIland GIrRespectively representing Gini indexes of two nodes after branching;
health factor XjThe node appearing in decision tree i is set M, then XjThe importance in the ith tree is:
Figure BDA0003490526770000061
if there are n trees in the random forest model, then:
Figure BDA0003490526770000062
and (3) normalizing all the obtained importance scores to obtain the importance scores of the health factors:
Figure BDA0003490526770000063
4) selecting parameters to be optimized of the BilSTM model, and setting m initial hyper-parameters Xm＝[X1,X2,...,Xm]And substituting the model into a BilSTM model for training. To obtain XmResult Y on the objective functionm＝[Y1,Y2,...,Ym]Building a matrix D { (X)1,Y1)...(Xm,Ym)}. Verifying whether the initial model meets the precision requirement on the training set, and if so, and if the iteration number is less than the maximum iteration number, the hyper-parameter of the model is XmAnd if not, carrying out Bayesian optimization loop iteration. The Bayesian optimization process is as follows: and establishing a Gaussian process regression model by using the matrix D. The next group of hyper-parameters X is carried out through the collection function g (X)m+1And (4) calculating. Substituting the new hyper-parameters into the BilSTM model to perform model training and calculate the result Y of the objective functionmAnd judging whether the precision requirement is met.
Step two: and (4) online prediction. The method comprises the following specific steps:
1) calculating the health factor with the importance greater than 0.3 by using the random forest model in the step 3) of the off-line training stage of the step one, and acquiring corresponding health factor data on line;
2) establishing an online prediction model by using Bayesian optimization BiLSTM model parameters in the step 4) of the offline training stage in the step one;
3) and inputting corresponding health factor data acquired on line, updating the cell state through a forgetting gate and an input-output gate, acquiring the time sequence relation between the previous moment and the subsequent moment, and finally obtaining the final lithium battery residual life prediction result through a full connection layer. If the predicted battery capacity is less than 80%, the end of the battery life is determined.
And (4) functional verification of the method:
the effectiveness of the method of the present invention is illustrated below with reference to the aging experimental data of lithium batteries. A commercial 18650 lithium battery is adopted, and the lithium battery is charged at a Constant Current (CC) at a room temperature of 24 ℃ until the current is 1.5A and the voltage reaches 4.2V, and then is charged at a Constant Voltage (CV) until the current is reduced to 20 mA. Discharge was discharged at a constant current of 2A until the cutoff voltage of each cell was reached.
The following detailed description of the implementation steps of the present invention is provided in conjunction with the specific process:
1. according to the lithium battery charging and discharging data set, taking B5 battery as an example, the health factor is extracted for the isobaric charging time (t) of 70, 80 and 90 cycleseq-vc) Charging voltage rise (V) at equal time intervalseq-tc) Duration of constant current charging (t)hc) Charging current drop (I) at equal time intervalseq-tc) Volume increase curve (ic), initial temperature (T)0) Mean temperature (T)avg) Median temperature (T)mid) Maximum temperature (T)max) Average rate of temperature change (T)rate) Constant pressure drop discharge time (t)eq-vd) Discharge voltage drop (V) at equal time intervalseq-td) Mean voltage decay (MVF), entropy of voltage samples (V)SampEn) And stacking the health factor row vectors into a matrix to obtain a health factor sample training set, namely obtaining a training set X of the health factors.
2, selecting health factors by using a random forest algorithm, extracting in a training set X by using a sampling and returning method (bootstrap), calculating Gini index score of each health factor, and screening out the health factors with the importance greater than 0.3.
3. Selecting parameters needing to be optimized of the BilSTM model, and setting 3 initial hyper-parameters, the number of hidden layers, the number of neurons of the hidden layers and the size of Dropout. The number range of the hidden layers is set to be 2-4, the number of the neurons of the hidden layers is set to be 200-500, the Dropout search range is 0-0.9, the search space is adjusted according to the whole hundred, Bayesian iterative optimization is carried out, and after the accuracy requirement is met, the optimal parameter model is obtained, wherein the number of the hidden layers is 2 at the moment, the number of the neurons of the hidden layers is 450, and the Dropout is 0.2.
4. And inputting the corresponding health factor data acquired on line into the optimal parameter model to obtain a final lithium battery residual life prediction result. If the predicted battery capacity is less than 80%, the end of the battery life is determined. The lithium battery life prediction accuracy of the traditional BilSTM neural network and the method of the invention is shown in Table 1, and B5 battery prediction results are shown in figures 2 and 3. Lithium battery residual life prediction method based on random forest health factor selection and Bayesian optimization BilSTM has the advantages that when 70 training samples are used, MAE is 0.00985Ah, and MSE is 1.3402 multiplied by 10-4Ah, the prediction progress is gradually improved along with the increase of the training samples, the error of the RUL prediction value is 0 in 90 training samples, the error of the prediction value is maintained within 1.5% each time, and the prediction effect is superior to that of the traditional BilSTM network.
Table 1: b5 prediction result error of different training data of battery
Figure BDA0003490526770000071
Figure BDA0003490526770000081
The above-described embodiments are intended to illustrate rather than to limit the invention, and any modifications and variations of the present invention are within the spirit of the invention and the scope of the appended claims.

Claims (3)

1. A method for predicting the residual life of a lithium battery is characterized by comprising the following steps:
the method comprises the following steps: performing offline modeling, namely collecting offline data of the lithium battery, extracting a lithium battery degradation characteristic sequence to obtain a health factor sample set, and performing weight analysis on the health factor sample set by using a random forest algorithm to determine the selected health factor sample according to the occupied weight; adding the selected samples into training data in sequence, randomly generating initialization parameters, and substituting the initialization parameters into a BilSTM model for training; when the precision of the training result does not meet the requirement, the Bayesian optimization model is used for: establishing a Gaussian process regression model, solving an acquisition function expression, solving the next group of model parameters according to the acquisition function until the optimal hyper-parameter meeting the precision requirement is found, and constructing a prediction model;
step two: in the online prediction, a health factor sample set is obtained by collecting online data of the lithium battery and corresponding to the feature selection in an offline stage; and then, using a BilSTM prediction model established in an off-line stage to predict the service life of the lithium battery.
2. The method for predicting remaining life of lithium battery as claimed in claim 1, wherein the off-line training process of step one is as follows:
step 11: collecting off-line data of the lithium battery, including current, voltage and temperature in the charging and discharging process, and extracting degradation characteristics; obtaining the constant pressure rise charging time teq-vcCharging voltage rise V at equal time intervalseq-tcConstant current charging duration thcEqual time interval charging current drop Ieq-tcVolume increase curve ic, initial temperature T0Mean temperature TavgMedian temperature TmidMaximum temperature TmaxAverage rate of temperature change TrateConstant voltage drop discharge time teq-vdEqual time interval discharge voltage drop Veq-tdMean voltage decay MVF, voltage sample entropy VSampEn(ii) a The calculation formula of each degradation characteristic is as follows:
Figure FDA0003490526760000011
Figure FDA0003490526760000012
Figure FDA0003490526760000013
Figure FDA0003490526760000014
wherein T is temperature, VnIs a nominal voltage of 4.2V, K is the sampling frequency, Bm(r) is the probability that two voltage sequences match m points with a similar tolerance r, Am(r) is the probability that two voltage sequences match m +1 points;
step 12: when all the health factor sample sets are obtained, the health factor X is extracted1，X2，X3，...,XqStacking into a matrix to obtain a health factor sample training set X belonging to Rp×qWherein p represents the number of training samples, and q represents the number of health factors;
step 13: selecting health factors by using a random forest algorithm, extracting the health factors in a training set X by using a sampling and putting back method, randomly sampling n samples, and generating t training sets; generating a corresponding decision tree by the training set, randomly and repeatedly selecting d features at each generated node, dividing the training set Y by the d features, finding the optimal division feature by using a Gini coefficient, and repeating the steps until a random forest model is obtained; the Gini index is expressed by GI, and the Gini index score VIM of each health factor is calculated(Gini)The Gini index is calculated by the formula:
Figure FDA0003490526760000021
wherein K represents K categories, pdkRepresenting the proportion of the category k in the node d;
health factor XjThe importance at node m is:
Figure FDA0003490526760000022
GIland GIrRespectively representing Gini indexes of two nodes after branching;
health factor XjThe node appearing in decision tree i is set M, then XjThe importance in the ith tree is:
Figure FDA0003490526760000023
if there are n trees in the random forest model, then:
Figure FDA0003490526760000024
and (3) normalizing all the obtained importance scores to obtain the importance scores of the health factors:
Figure FDA0003490526760000025
step 14: selecting parameters to be optimized of a BiLSTM model, and setting m initial hyper-parameters Xm＝[X1,X2,...,Xm]Substituting the model into a BilSTM model for training; to obtain XmResult Y on the objective functionm＝[Y1,Y2,...,Ym]Building a matrix D { (X)1,Y1)...(Xm,Ym) }; verifying whether the initial model meets the precision requirement on the training set, and if so, and if the iteration number is less than the maximum iteration number, the hyper-parameter of the model is XmIf not, carrying out Bayesian optimization loop iteration;
the Bayesian optimization process comprises the following steps: establishing a Gaussian process regression model by using the matrix D; the next group of hyperparameters X is carried out through the acquisition function g (X)m+1Calculating (1); substituting the new hyper-parameters into the BilSTM model to perform model training and calculate the result Y of the objective functionmAnd judging whether the precision requirement is met.
3. The method for predicting the remaining life of the lithium battery as claimed in claim 1, wherein the online detection process in the second step is as follows:
step 21: using the health factor with the importance greater than 0.3 calculated by the random forest model in the step 13 of the off-line training stage in the step one, and acquiring corresponding health factor data on line;
step 22: establishing an online prediction model by using Bayesian optimization BiLSTM model parameters in step 14 of the offline training stage in the step one;
step 23: inputting corresponding health factor data acquired on line, updating the cell state through a forgetting gate and an input-output gate, acquiring the time sequence relation between the previous moment and the subsequent moment, and finally acquiring a final lithium battery residual life prediction result through a full connection layer: if the predicted battery capacity is less than 80%, the end of the battery life is determined.
CN202210094514.2A 2022-01-26 2022-01-26 Lithium battery residual life prediction method Pending CN114545274A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210094514.2A CN114545274A (en) 2022-01-26 2022-01-26 Lithium battery residual life prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210094514.2A CN114545274A (en) 2022-01-26 2022-01-26 Lithium battery residual life prediction method

Publications (1)

Publication Number Publication Date
CN114545274A true CN114545274A (en) 2022-05-27

Family

ID=81674022

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210094514.2A Pending CN114545274A (en) 2022-01-26 2022-01-26 Lithium battery residual life prediction method

Country Status (1)

Country Link
CN (1) CN114545274A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115308603A (en) * 2022-07-13 2022-11-08 中国长江三峡集团有限公司 Battery life prediction method based on multi-dimensional features and machine learning
CN115436814A (en) * 2022-10-17 2022-12-06 吉林大学 Probability prediction method for residual life of lithium ion battery
CN116011109A (en) * 2023-01-13 2023-04-25 北京控制工程研究所 Spacecraft service life prediction method and device, electronic equipment and storage medium
CN117150334A (en) * 2023-06-16 2023-12-01 合肥工业大学 Lithium battery multi-condition prediction method and device based on optimized BiLSTM neural network
CN117540879A (en) * 2023-12-26 2024-02-09 淮阴工学院 Method for predicting state of charge of lithium battery of new energy electric car
CN117723999A (en) * 2024-02-07 2024-03-19 深圳市东田通利电业制品有限公司 Battery service life prediction method, device, equipment and storage medium

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115308603A (en) * 2022-07-13 2022-11-08 中国长江三峡集团有限公司 Battery life prediction method based on multi-dimensional features and machine learning
CN115436814A (en) * 2022-10-17 2022-12-06 吉林大学 Probability prediction method for residual life of lithium ion battery
CN116011109A (en) * 2023-01-13 2023-04-25 北京控制工程研究所 Spacecraft service life prediction method and device, electronic equipment and storage medium
CN116011109B (en) * 2023-01-13 2023-09-08 北京控制工程研究所 Spacecraft service life prediction method and device, electronic equipment and storage medium
CN117150334A (en) * 2023-06-16 2023-12-01 合肥工业大学 Lithium battery multi-condition prediction method and device based on optimized BiLSTM neural network
CN117540879A (en) * 2023-12-26 2024-02-09 淮阴工学院 Method for predicting state of charge of lithium battery of new energy electric car
CN117723999A (en) * 2024-02-07 2024-03-19 深圳市东田通利电业制品有限公司 Battery service life prediction method, device, equipment and storage medium

Similar Documents

Publication Publication Date Title
CN114545274A (en) Lithium battery residual life prediction method
CN111443294B (en) Method and device for indirectly predicting remaining life of lithium ion battery
CN110221225B (en) Spacecraft lithium ion battery cycle life prediction method
CN110687452B (en) Lithium battery capacity online prediction method based on K-means clustering and Elman neural network
Fan et al. A novel machine learning method based approach for Li-ion battery prognostic and health management
CN111443293A (en) Lithium battery state of health (SOH) estimation method based on data driving
CN111999649A (en) XGboost algorithm-based lithium battery residual life prediction method
CN110658462B (en) Lithium battery online service life prediction method based on data fusion and ARIMA model
CN113917334B (en) Battery health state estimation method based on evolution LSTM self-encoder
CN114861527A (en) Lithium battery life prediction method based on time series characteristics
CN115469227B (en) Aggregation variation self-encoder and dynamic regular lithium battery abnormality detection method
CN113406521B (en) Lithium battery health state online estimation method based on feature analysis
CN108009585B (en) Lead-acid battery health state prediction method based on local information fusion
CN114726045B (en) Lithium battery SOH estimation method based on IPEA-LSTM model
CN112816874A (en) RVM and PF algorithm fusion-based battery remaining service life prediction method
CN115656857A (en) Multi-scale fusion prediction method for remaining service life of lithium ion battery
CN115409263A (en) Method for predicting remaining life of lithium battery based on gating and attention mechanism
CN108303649A (en) A kind of cell health state recognition methods
CN112098873B (en) Lithium battery health state estimation method based on charging voltage curve geometric characteristics
CN114384435A (en) WSA-LSTM algorithm-based self-adaptive prediction method for residual service life of new energy automobile power battery
CN113901707A (en) SOH time sequence prediction method for vehicle lithium ion power battery
CN116736133A (en) Early prediction method for capacity degradation track of lithium ion battery in full life cycle
Savargaonkar et al. A cycle-based recurrent neural network for state-of-charge estimation of li-ion battery cells
CN115902642A (en) Battery state of charge estimation method and device, electronic equipment and storage medium
CN114460481A (en) Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism

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