CN116680983A - Lithium ion residual life prediction method based on improved particle filter model - Google Patents
Lithium ion residual life prediction method based on improved particle filter model Download PDFInfo
- Publication number
- CN116680983A CN116680983A CN202310662200.2A CN202310662200A CN116680983A CN 116680983 A CN116680983 A CN 116680983A CN 202310662200 A CN202310662200 A CN 202310662200A CN 116680983 A CN116680983 A CN 116680983A
- Authority
- CN
- China
- Prior art keywords
- model
- battery
- particle
- sequence
- lithium ion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000002245 particle Substances 0.000 title claims abstract description 133
- 238000000034 method Methods 0.000 title claims abstract description 48
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 title claims abstract description 41
- 229910001416 lithium ion Inorganic materials 0.000 title claims abstract description 41
- 230000036541 health Effects 0.000 claims abstract description 87
- 230000015556 catabolic process Effects 0.000 claims abstract description 31
- 238000006731 degradation reaction Methods 0.000 claims abstract description 31
- 238000005457 optimization Methods 0.000 claims abstract description 31
- 230000008569 process Effects 0.000 claims abstract description 17
- 238000013507 mapping Methods 0.000 claims abstract description 9
- 230000006870 function Effects 0.000 claims description 55
- 238000012549 training Methods 0.000 claims description 28
- 238000007600 charging Methods 0.000 claims description 24
- 230000032683 aging Effects 0.000 claims description 22
- 238000001914 filtration Methods 0.000 claims description 21
- 238000010277 constant-current charging Methods 0.000 claims description 18
- 238000005070 sampling Methods 0.000 claims description 15
- 238000012952 Resampling Methods 0.000 claims description 14
- 238000007599 discharging Methods 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 230000005611 electricity Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 238000010280 constant potential charging Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000001351 cycling effect Effects 0.000 claims description 3
- 230000006872 improvement Effects 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims 1
- 230000008859 change Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 abstract description 3
- 238000003062 neural network model Methods 0.000 abstract 1
- 238000002474 experimental method Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 230000003247 decreasing effect Effects 0.000 description 2
- 230000009191 jumping Effects 0.000 description 2
- 238000000342 Monte Carlo simulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010281 constant-current constant-voltage charging Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 235000003642 hunger Nutrition 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000037351 starvation Effects 0.000 description 1
Classifications
-
- 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
- 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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Computational Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Pure & Applied Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Algebra (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Operations Research (AREA)
- Databases & Information Systems (AREA)
- Tests Of Electric Status Of Batteries (AREA)
- Secondary Cells (AREA)
Abstract
The invention relates to a lithium ion residual life prediction method based on an improved particle filter model, and belongs to the technical field of batteries. Aiming at the problems that the degradation rate difference of the battery is large and the degradation data of the battery is insufficient due to the random current change and incomplete charge and discharge caused by random load in the use process of the lithium ion battery, a method for estimating the health state of the lithium ion battery and constructing a neural network model for predicting the residual life is provided, which specifically comprises the following steps: extracting health factors from the historical data; obtaining a complete health factor sequence through multi-core RVM mapping based on Bayesian optimization; establishing a sequence-to-sequence LSTM model, and estimating the real capacity of the lithium ion battery; establishing a lithium ion battery degradation double-index model based on IPSO-PF-LSTM; and predicting the residual life. The advantages of the double-index model can be better exerted through the joint optimization of various algorithms, so that the effect of life prediction is improved.
Description
Technical Field
The invention belongs to the technical field of batteries, and relates to a lithium ion residual life prediction method based on an improved particle filter model.
Background
The performance of the battery can be irreversibly affected after continuous charge and discharge cycles. The problems of battery leakage, insulation damage, local short circuit and the like can cause disastrous accidents, so accurate estimation of battery state information is necessary to avoid potential safety hazards caused by aging and performance degradation of the lithium ion battery.
The health state and the residual service life of the battery are closely related to the historical service condition of the battery, the battery is often not under the standard charge and discharge working condition in practical application, and the problems of large degradation rate difference, insufficient degradation data of the battery and the like caused by the random discharge in practice are solved, and for the situation, the complete health factor sequence after the optimization is taken into consideration, and the estimation of the health state and the prediction of the residual service life are carried out through LSTM from sequence to sequence and an improved particle filter model.
Disclosure of Invention
In view of the above, an object of the present invention is to provide a lithium ion remaining lifetime prediction method based on an improved particle filter model. Based on the problems of large battery degradation rate difference and insufficient battery degradation data, the multi-core RVM is utilized to perform health factor mapping, and meanwhile, the LSTM model from sequence to sequence is combined to obtain capacity estimation values of a plurality of future cycles.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a method for predicting lithium ion remaining life based on an improved particle filter model, the method comprising the steps of:
s1: extracting and optimizing health factors: extracting health factors aiming at a lithium ion battery charge and discharge data curve, and optimizing parameters of part of the health factors to obtain the health factor with highest correlation with capacity attenuation;
s2: health factor mapping for multi-core RVMs: taking a Relevant Vector Machine (RVM) after Bayesian optimization as a tool, and learning a mapping relation between the charging health factor and the discharging health factor to obtain a complete health factor sequence;
s3: sequence-to-sequence LSTM model: according to the construction sequence input of the complete battery health factor extracted in the step S2, using battery capacity data as a label to train the LSTM neural network;
s4: establishing a lithium ion battery degradation double-index model based on IPSO-PF: adopting an improved particle swarm optimization algorithm and a particle filtering algorithm to continuously optimize parameters of a double-index model so as to fit a battery capacity degradation trend;
s5: life prediction: training the model constructed in S4 and performing life prediction.
Optionally, the S1 specifically is: selecting the time HF of an equal voltage interval during constant current charging based on a charging phase curve 1 Time difference HF of equal current interval during constant voltage charging 2 Temperature difference HF of equal voltage interval during constant current charging 3 As health factors, specific calculation formulas are as follows:
wherein The constant current charging voltage reaches V upper Absolute time of->The constant current charging voltage reaches V lower Absolute time of (2); />The constant voltage charge time voltage reaches I upper Absolute time of->The constant voltage charge time voltage reaches I lower Absolute time of (2); />The constant current charging voltage reaches V upper Temperature of>The constant current charging voltage reaches V lower Is set at a temperature of (2);
extracting random discharge current integral value HF from random discharge curve of battery 4 Random discharge voltage sample entropy HF 5 Standard deviation of random discharge temperature HF 6 As health factors, specific calculation formulas are as follows:
where SampEn is an estimate of the conditional probability that a window of length m, which is a subsequence of the time sequence of length Nc, remains similar within the tolerance r and also matches the next electricity; t (T) K Is the Kth temperature value of the random discharge cycle of the battery; u is the average degree of random discharge cycle, n is the number of measured temperatures;
After six health factors are respectively extracted for a lithium ion battery charging and discharging curve under a random discharging working condition, optimizing the health factors extracted in a charging process through a particle swarm optimization algorithm; specifically, HF is respectively treated 1 V of (2) upper 、V lower ,HF 2 I of (2) upper 、I lower and HF3 V of (2) upper 、V lower As a parameter to be optimized, the pearson correlation coefficient of each health factor and the capacity is used as an adaptability function of the PSO, namely:
F(x)=ρ(HF,capacity) (3)
wherein F (x) is a particle swarm fitness function, HF is a corresponding health factor sequence, capability is a capacity sequence, and ρ is a pearson correlation coefficient function.
Optionally, the S2 specifically is: charging health factor HF for cycling incomplete charge and discharge 1 、HF 2 、HF 3 Together as input, the discharge health factor HF 4 、HF 5 、HF 6 Respectively as outputs; for data setsx i ∈R,t i∈R wherein xi For input of sample, t i For a target output value, R represents a real set, and D is the total number of samples; the relationship between model inputs and outputs is as follows:
t i =y(x i ,w i )+ε i (4)
wherein K (x, x) i ) Expressed as a kernel function, w= (w 1 ,w 2 ,K,w D ) T Weight vector corresponding to kernel function, w 0 Is a bias parameter;
aiming at different degradation trends of health factors, linear kernel functions, polynomial kernel functions, gaussian kernel functions and Sigmoid kernel functions are selected for linear combination, and weight coefficients of the four combined kernel functions need to be manually determined;
in the process of optimizing the RVM model through a Bayesian optimization algorithm, the method consists of two steps of Gaussian process regression and acquisition function sampling, and posterior distribution of the objective function is updated through continuously adding sampling points until the posterior distribution is consistent with the real distribution; firstly, selecting an initial super-parameter set x in a domain space χ; then calculating the probability distribution of f (x) according to the properties of the sampling set and the Gaussian distribution, and obtaining the position of the next sampling point by using an acquisition function; finally, f (x) is calculated and an objective function f is updated to reach the iteration times or meet the convergence condition;
the specific implementation process of the multi-core RVM algorithm based on Bayesian optimization is as follows: preparing a training set and a prediction set; determining super parameters and ranges which need to be optimized; and (5) building and training a Bayesian optimization RVM model.
Optionally, the S3 specifically is:
s32: constructing a data set: acquiring a complete health factor sequence according to the multi-core RVM framework in the S2, normalizing the complete health factor sequence, constructing input data, extracting capacity data and calculating to obtain a battery health state SOH, namely a label corresponding to the input; the structured dataset is:
Y={Y 1 ,Y 2 ,Y 3 ,…,Y m } (6)
for the LSTM model, the step length is set to be 10, and the current capacity value { x ] of 10 continuous loops is used 41 ,x 42 ,…,x 49 ,x 50 One prediction outputs capacity estimation values of 10 future cyclesThe 10 estimated values are used for carrying out the next prediction, the LSTM method from sequence to sequence has less iteration times and small accumulated error; y is Y i SOH value for the ith cycle; m is the number of cycles;
s32: model offline training: firstly, determining parameters to be optimized of an LSTM algorithm as time step, hidden layer node number and training times; inputting optimized parameters of the particle swarm optimization into an LSTM model, initializing other parameters, dividing the data set obtained in the step (1) into a training set and a testing set, and training the LSTM model;
s33: on-line prediction: and extracting charging data of the current cycle, combining the charging data with past data to construct and inputting the charging data into a trained model to obtain SOH predicted values of a plurality of continuous cycles.
Optionally, in the step S4, a double-index model is used as an empirical model of the lithium ion battery to fit the battery capacity degradation trend; the model built therein is as follows:
wherein k is the cycle number of charge and discharge cycles, Q k A is the capacity of the kth cycle of the battery, a 1 、a 2 、a 3 、a 4 The aging parameters to be identified for the double-index model are related to the internal characteristics of the battery, wherein a 1 、a 3 Related to the internal resistance of the current cycle of the battery, a 2 、a 4 Related to the degradation rate of the battery;
selecting x r,k =[x 1,k x 2,k x 3,k x 4,k ] T Discretizing the state variable of the battery capacity attenuation model in the kth cycle according to a state equation and an observation equation of particle filtering, wherein the state variable is as follows:
Q k =a 1,k ·exp(a 2,k ·k)+a 3,k ·exp(a 4,k ·k) (8)
in the formula ,x1,k 、x 2,k 、x 3,k 、x 4,k The initial value of (a) is a double-index model parameter a 1 、a 2 、a 3 、a 4 ;y k Is the output of the model, Q k Gamma, the capacity of lithium ion battery k =[γ 1,k γ 2,k γ 3,k γ 4,k ] T In order for the process to be noisy,is measurement noise;
in order to alleviate the particle degradation phenomenon, the particle can better adapt to the battery aging trend under the random discharging working condition, and on the basis of the established battery aging model, the capacity is predicted by using an IPSO-PF algorithm;
the specific steps of the IPSO-PF algorithm are as follows:
(1) initialization of
Setting the current charge and discharge cycle number as k; when the particle filtering is performed, k=0 is taken, and the filter is determined according to the initial prior probability density p (x r,0 ) Extracting n initial state particles
(2) Importance sampling
When k=1, 2,..n, selecting the prior PDF as the state PDF, i.e
Extracting new samples based on importance probability densityThen adopting improved particle swarm algorithm to collect particlesIterative optimization is carried out, and the specific steps are as follows:
step 11: setting the maximum iteration number T max The method comprises the steps of carrying out a first treatment on the surface of the The population scale n is consistent with the particle quantity selection in the particle filtering, and a particle range R is setMaximum speed V max Inertial weight omega, acceleration factor c 1 、c 2 Waiting for initial parameters; let the fitness function of particle swarm optimization be:
wherein K is the number of training set samples, L is the step size of LSTM based on sequence to sequence,the estimated value of the LSTM for L capacities in the future is the last subsection;
step 12: let the iteration number t=1, set the initial position of the particle as the initial parameter of the particle in the particle filterSetting the current position as an individual history optimal pBest, setting the optimal individuals in the group as a group history optimal gBest, and calculating the fitness function value of each particle;
step 13: in the current iteration loop, if the current fitness function value is better than the historical optimal value, updating the individual historical optimal, and if the current fitness function is better than the global historical optimal value, updating the population historical optimal; at the end of one cycle, the d-th dimension of the velocity and position of each particle i are updated separately as follows:
in the formula c1 、c 1 Taking a fixed value of 2.0 as a learning factor; ω is inertial weight, initialized to 0.9, and a linear differential decremental strategy is selected for improvement, and the algorithm is as follows:
step 14: judging whether the maximum iteration number is reached or the set optimization extremum is reached, if so, optimizing, and setting particles in the PFOtherwise, jumping to the step 2;
(3) particle weight update
Reconstructing a particle weight formula, and carrying out weight judgment in a plurality of time ranges by combining an aging model by means of a first iteration result from an LSTM sequence to a sequence, wherein the specific formula is as follows:
wherein ,is the weight of the i-th particle, +.>For the LSTM first iteration result based on sequence-to-sequence, y k In order to consider the output value of the aging model of the noise, R is the variance of the measured noise, and N is the total charge-discharge cycle number of the battery;
(4) resampling
Firstly, judging whether the cycle needs resampling or not; number N of effective particles eff The method comprises the following steps:
if N eff <N threshold For particlesRandom resampling is carried out to obtain a new particle setOtherwise, not resampling; />
(5) State estimation
The particles in the particle set are weighted to obtain an estimated value:
(6) battery RUL prediction
Assuming that the battery is currently in the kth cycle, according to the state equation and the observation equation of particle filtering, the capacity of each particle in the current cycle for predicting j future cycles is calculated as follows:
the predicted capacity at the kth cycle for the future j cycles is:
posterior PDF approximation of battery capacity is written:
the battery RUL predicted value is calculated as follows:
optionally, the step S5 is to input the test set data into the model established in the step S4, and perform model training based on particle filtering and perform life prediction after obtaining the parameters of the double-index model.
The invention has the beneficial effects that: aiming at the problems of different rates in the later stage of battery life degradation caused by random discharge current and the RUL prediction method based on data driving, the invention provides an improved RUL prediction method of a particle filter fused LSTM neural network, which comprises the steps of firstly establishing a lithium ion battery degradation double-index model, optimizing parameters of the double-index model through particle filter, optimizing an improved particle swarm algorithm aiming at the particle degradation phenomenon in PF, improving the influence of the random discharge current on the later stage degradation degree of the battery, using an importance sampling step and a particle weight updating step in PF, using the health state estimation result of LSTM based on sequence-to-sequence for the importance sampling and the particle weight updating of the particle filter, and establishing an adaptability function and a weight judgment formula considering the future capacity change trend of the battery, thereby improving the effect of residual life prediction.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objects and other advantages of the invention may be realized and obtained by means of the instrumentalities and combinations particularly pointed out in the specification.
Drawings
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in the following preferred detail with reference to the accompanying drawings, in which:
FIG. 1 is a structural frame diagram of the present invention;
FIG. 2 is a graph of a health factor sequence and a capacity sequence correlation analysis; (a) is a charge health factor versus capacity map; (b) is a discharge health factor versus capacity graph;
FIG. 3 is a flow chart of optimizing health factor parameters by a particle swarm algorithm;
fig. 4 is a graph showing estimation results of the health factor of RW4 at 60% of the duty cycle of the incomplete charge-discharge cycle; (a) estimating HF4 for HF1, HF2, HF 3; (b) estimating an absolute error for HF4; (c) estimating HF5 for HF1, HF2, HF 3; (d) estimating an absolute error for HF5; (e) estimating HF6 for HF1, HF2, HF 3; (f) estimating an absolute error for HF6;
FIG. 5 is an overall frame diagram of a battery state of health estimation method;
FIG. 6 is a graph of predicted remaining life of a lithium ion battery under standard operating conditions;
fig. 7 is a graph of a predicted remaining life of a lithium ion battery under random discharge conditions.
Detailed Description
Other advantages and effects of the present invention will become apparent to those skilled in the art from the following disclosure, which describes the embodiments of the present invention with reference to specific examples. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention. It should be noted that the illustrations provided in the following embodiments merely illustrate the basic idea of the present invention by way of illustration, and the following embodiments and features in the embodiments may be combined with each other without conflict.
Wherein the drawings are for illustrative purposes only and are shown in schematic, non-physical, and not intended to limit the invention; for the purpose of better illustrating embodiments of the invention, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the size of the actual product; it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The same or similar reference numbers in the drawings of embodiments of the invention correspond to the same or similar components; in the description of the present invention, it should be understood that, if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc., that indicate an azimuth or a positional relationship based on the azimuth or the positional relationship shown in the drawings, it is only for convenience of describing the present invention and simplifying the description, but not for indicating or suggesting that the referred device or element must have a specific azimuth, be constructed and operated in a specific azimuth, so that the terms describing the positional relationship in the drawings are merely for exemplary illustration and should not be construed as limiting the present invention, and that the specific meaning of the above terms may be understood by those of ordinary skill in the art according to the specific circumstances.
As shown in fig. 1, the implementation process of the present invention includes:
s1: extracting and optimizing health factors: and extracting health factors aiming at the charge and discharge data curve of the lithium ion battery, and optimizing parameters of part of the health factors to obtain the health factor with the highest correlation with capacity attenuation.
S2: health factor mapping for multi-core RVMs: and learning the mapping relation between the charging health factor and the discharging health factor by taking a Relevant Vector Machine (RVM) subjected to Bayesian optimization as a tool so as to obtain a complete health factor sequence.
S3: sequence-to-sequence LSTM model: and (3) according to the input of the complete battery health factor construction sequence extracted in the step S2, training the LSTM neural network by using the battery capacity data as a label.
S4: establishing a lithium ion battery degradation double-index model based on IPSO-PF: and continuously optimizing parameters of the double-index model by adopting an improved particle swarm optimization algorithm and a particle filtering algorithm so as to fit the capacity degradation trend of the battery.
S5: life prediction: training the model constructed in S4 and performing life prediction.
Experiment
The battery aging analysis, the estimation of the health state under the random discharging working condition and the study of the RUL prediction of the lithium ion battery under the random discharging working condition are carried out by combining the battery aging data set and the oxford battery aging data set provided by the American space agency-Ames prediction center. The oxford dataset contained battery aging data for 8 lithium ion pouch cells with a rated capacity of 0.74Ah, all data were performed in an incubator at 40 ℃ with a constant-constant voltage charge mode, and the discharge mode was repeatedly performed based on the Urban Artemis driving cycle, and a standard charge-discharge cycle was performed after every 100 charge-discharge cycles to determine battery characteristics. The NASA random discharge data set simulates discharge current working conditions of different styles, and the battery charge and discharge experiment is subjected to three processes of a random discharge cycle experiment, a reference charge and discharge experiment and a pulse discharge experiment, and the residual life of the lithium ion battery is predicted by taking RW4 battery data as an example.
S1: extracting and optimizing health factors, firstly extracting characteristics of a battery in a charging stage based on a characteristic curve change diagram of the battery in constant-current-constant-voltage charging,
selecting a time (HF) of an equal voltage interval during constant current charging based on a charging phase curve 1 ) Time difference of equal current interval during constant voltage charging (HF 2 ) Temperature difference (HF) between equal voltage intervals during constant current charging 3 ) As health factors, specific calculation formulas are as follows:
wherein The constant current charging voltage reaches V upper Absolute time of->The constant current charging voltage reaches V lower Absolute time of (2); />The constant voltage charge time voltage reaches I upper Absolute time of->The constant voltage charge time voltage reaches I lower Absolute time of (2); />The constant current charging voltage reaches V upper Is of the temperature of (1)Degree (f)>The constant current charging voltage reaches V lower Is set at a temperature of (2);
in addition, it is necessary to extract a random discharge current integral value (HF) from a random discharge curve of the battery 4 ) Random discharge voltage sample entropy (HF 5 ) Standard deviation of random discharge temperature (HF 6 ) As health factors, specific calculation formulas are as follows:
where SampEn is an estimate of the conditional probability that a window of length m (a subsequence of a time series of length Nc) remains similar within the tolerance r and also matches the next electricity; t (T) K Is the kth temperature value of the battery random discharge cycle. u is the average of the random discharge cycles and n is the number of temperatures measured. As shown in fig. 2, a graph of health factor versus capacity obtained for the RW4 dataset is provided. In fig. 2, (a) is a graph of charge health factor versus capacity; (b) is a graph of discharge health factor versus capacity.
After six health factors are respectively extracted for the charge-discharge curve of the lithium ion battery under the random discharge working condition, the health factors extracted in the charging process can be optimized through a particle swarm optimization algorithm. Specifically, HF is respectively treated 1 V of (2) upper 、V lower ,HF 2 I of (2) upper 、I lower and HF3 V of (2) upper 、V lower As a parameter to be optimized, the pearson correlation coefficient of each health factor and the capacity is used as an fitness function of the PSO, at this time, the population is selected to be 50, and the iteration number is selected to be 20, namely:
F(x)=ρ(HF,capacity) (3)
wherein F (x) is a particle swarm fitness function, HF is a corresponding health factor sequence, capability is a capacity sequence, and ρ is a pearson correlation coefficient function. As shown in fig. 3, a flow chart for optimizing health factor parameters for a particle swarm algorithm is provided.
S2: charging health factor HF for cycling incomplete charge and discharge 1 、HF 2 、HF 3 Together as input, the discharge health factor HF 4 、HF 5 、HF 6 Respectively as outputs. For data setsx i ∈R,t i∈R wherein xi For input of sample, t i For the target output value, R represents the real set and D is the total number of samples. The relationship between model inputs and outputs is as follows:
t i =y(x i ,w i )+ε i (4)
wherein K (x, x) i ) Expressed as a kernel function, w= (w 1 ,w 2 ,K,w D ) T Weight vector corresponding to kernel function, w 0 Is a bias parameter.
Aiming at different degradation trends of health factors, linear kernel functions, polynomial kernel functions, gaussian kernel functions and Sigmoid kernel functions are selected for linear combination, and weight coefficients of the four combined kernel functions need to be determined manually.
In addition, in the process of optimizing the RVM model through a Bayesian optimization algorithm, the RVM model mainly comprises two steps of Gaussian process regression and acquisition function sampling, and posterior distribution of the objective function is updated through continuous addition of sampling points until the posterior distribution is consistent with the real distribution.
When the health factor mapping is performed, the ratio of the number of incomplete charge and discharge cycles in the training set to the total training set has an important influence on the estimation accuracy of the health factor. To demonstrate the estimated performance of the multi-core RVM, taking the battery RW4 as an example, the experiment was illustrated with an incomplete discharge cycle ratio of 60% for random current discharge, with the incomplete discharge cycles being randomly distributed throughout the discharge cycle. As shown in fig. 4, a graph of the estimation result of the RW4 incomplete charge-discharge cycle duty ratio 60% health factor is shown. In fig. 4, (a) HF4 is estimated for HF1, HF2, HF 3; (b) estimating an absolute error for HF4; (c) estimating HF5 for HF1, HF2, HF 3; (d) estimating an absolute error for HF5; (e) estimating HF6 for HF1, HF2, HF 3; (f) estimating an absolute error for HF 6.
S3: and (2) acquiring a complete health factor sequence according to the multi-core RVM framework in the S2, normalizing the complete health factor sequence, constructing input data, extracting capacity data and calculating to obtain a battery health State (SOH), namely a label corresponding to the input. The structured dataset is:
Y={Y 1 ,Y 2 ,Y 3 ,…,Y m } (6)
for the LSTM model, a step size of 10 is assumed, i.e. the capacity value { x over 10 cycles currently in succession 41 ,x 42 ,…,x 49 ,x 50 One prediction can output capacity estimation values of 10 future cyclesAnd the 10 estimated values are used for carrying out the next prediction, so that the LSTM method from sequence to sequence has less iteration times and small accumulated error. Y is Y i SOH value for the ith cycle. m is the number of cycles. (2) Model offline training: firstly, determining parameters to be optimized of an LSTM algorithm as time step, hidden layer node number and training times. Inputting optimized parameters of the particle swarm optimization into an LSTM model, initializing other parameters, dividing the data set obtained in the step (1) into a training set and a testing set, and training the LSTM model. (3) On-line prediction: and extracting charging data of the current cycle, combining the charging data with past data to construct and inputting the charging data into a trained model to obtain SOH predicted values of a plurality of continuous cycles. As shown in fig. 5, an overall frame diagram of the battery state of health estimation method is shown.
S4: and adopting a double-index model as an empirical model of the lithium ion battery to fit the capacity degradation trend of the battery. The model built therein is as follows:
wherein k is the cycle number of charge and discharge cycles, Q k A is the capacity of the kth cycle of the battery, a 1 、a 2 、a 3 、a 4 The aging parameters to be identified for the double-index model are related to the internal characteristics of the battery, wherein a 1 、a 3 Related to the internal resistance of the current cycle of the battery, a 2 、a 4 Related to the degradation rate of the battery.
Selecting x r,k =[x 1,k x 2,k x 3,k x 4,k ] T Discretizing the state variable of the battery capacity attenuation model in the kth cycle according to a state equation and an observation equation of particle filtering, wherein the state variable is as follows:
Q k =a 1,k ·exp(a 2,k ·k)+a 3,k ·exp(a 4,k ·k) (8)
in the formula ,x1,k 、x 2,k 、x 3,k 、x 4,k The initial value of (a) is a double-index model parameter a 1 、a 2 、a 3 、a 4 。y k Is the output of the model, Q k Gamma, the capacity of lithium ion battery k =[γ 1,k γ 2,k γ 3,k γ 4,k ] T In order for the process to be noisy,is the measurement noise.
In order to alleviate the particle degradation phenomenon, the particle can better adapt to the battery aging trend under the random discharging working condition, and the capacity is predicted by using an IPSO-PF algorithm on the basis of the established battery aging model. The specific steps of the IPSO-PF algorithm are as follows:
(1) initializing: let the current charge-discharge cycle number be k. When the particle filtering is performed, k=0 is firstly taken, and according to the initial prior probability density p (x r,0 ) Extracting n initial state particlesIncreasing the number of particles in particle filtering can effectively delay particle degradation, but can increase the amount of computation.
(2) Importance sampling: when k=1, 2,..n, selecting the prior PDF as the state PDF, i.e
Extracting new samples based on importance probability densityThen adopting improved particle swarm algorithm to collect particlesIterative optimization is carried out, and the specific steps are as follows:
step 11: setting the maximum iteration number T max The method comprises the steps of carrying out a first treatment on the surface of the The population scale n is consistent with the particle quantity selection in particle filtering, and a particle range R and a maximum speed V are set max Inertial weight omega, acceleration factor c 1 、c 2 Etc. initial parameters. Let the fitness function of particle swarm optimization be:
wherein K is the number of training set samples, L is the step size of LSTM based on sequence to sequence,is an estimate of the last section LSTM for the L capacities in the future.
Step 12: let the iteration number t=1, set the initial position of the particle as the initial parameter of the particle in the particle filterSetting the current position as the individual history optimal pBest, setting the optimal individuals in the group as the group history optimal gBest, and calculating the fitness function value of each particle.
Step 13: and in the current iteration loop, updating the historical optimum of the individual if the current fitness function value is better than the historical optimum, and updating the historical optimum of the population if the current fitness function value is better than the global historical optimum. At the end of one cycle, the d-th dimension of the velocity and position of each particle i are updated separately as follows:
in the formula c1 、c 1 For learning factor, a fixed value of 2.0 is taken. ω is inertial weight, the initialization is set to 0.9, considering that the particle swarm algorithm may trap into local extremum during iteration if linear decreasing strategy is adopted for ω, resulting in particle loss diversity, the linear differential decreasing strategy is selected for improvement, and the algorithm is as follows:
step 14: judging whether the maximum iteration number is reached or the set optimization extremum is reached, if so, optimizing, and setting particles in the PFOtherwise, jumping to step 2.
(3) Particle weight updating: the particle weight formula is reconstructed to emphasize the future trend of the particles, and is improved on the basis of the particle weight formula, the response degree of the future trend can be deeper by means of the first iteration result from the LSTM sequence to the LSTM sequence, and the weight judgment of multiple time ranges can be carried out by combining an aging model, wherein the specific formula is as follows:
wherein ,is the weight of the i-th particle, +.>For the LSTM first iteration result based on sequence-to-sequence, y k To consider the aging model output value of the noise, R is the variance of the measured noise and N is the total charge-discharge cycle number of the battery.
(4) Resampling: resampling is mainly to solve the particle starvation phenomenon that occurs in classical monte carlo methods. If the number of effective particles generated in the current cycle is small, resampling is needed to avoid the situation that the number of active particles is too small to be trapped into local optimum, and meanwhile, the calculated amount is prevented from being consumed on particles with small weights. It is first necessary to determine whether the cycle requires resampling. Number N of effective particles eff The method comprises the following steps:
if N eff <N threshold For particlesRandom resampling is carried out to obtain a new particle setOtherwise no resampling is performed. General choice +.>
(5) State estimation: the particles in the particle set are weighted to obtain an estimated value:
(6) battery RUL prediction: assuming that the battery is currently in the kth cycle, according to the state equation and the observation equation of particle filtering, the capacity of each particle in the current cycle for predicting j future cycles is calculated as follows:
thus, the predicted capacity at the kth cycle for the future j cycles is:
the posterior PDF of battery capacity can be written approximately:
the battery RUL predicted value is calculated as follows:
model training and prediction based on particle filtering can be performed after obtaining the double-index model parameters. Based on the algorithm flow of 4), the invention sets the particle number in particle filtering to 200, and the LSTM estimation sequence length to 15. In order to verify that the proposed IPSO-PF-LSTM algorithm can realize the RUL prediction of the lithium ion battery under the standard working condition and the random discharging working condition and the applicability of the lithium ion battery in different battery packs, experiments are compared with a plurality of different algorithms, and the results show that the error of the algorithm in a random current aging test data set and a standard working condition aging test data set is respectively within 6.2% and 7.0%, and the performance indexes are superior to those of other RUL prediction methods. As shown in fig. 6 and fig. 7, the predicted result diagrams of the remaining service life of the lithium ion battery under the standard working condition and the random discharging working condition are shown respectively.
Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the present invention, which is intended to be covered by the claims of the present invention.
Claims (6)
1. The lithium ion residual life prediction method based on the improved particle filter model is characterized by comprising the following steps of: the method comprises the following steps:
s1: extracting and optimizing health factors: extracting health factors aiming at a lithium ion battery charge and discharge data curve, and optimizing parameters of part of the health factors to obtain the health factor with highest correlation with capacity attenuation;
s2: health factor mapping for multi-core RVMs: taking a Relevant Vector Machine (RVM) after Bayesian optimization as a tool, and learning a mapping relation between the charging health factor and the discharging health factor to obtain a complete health factor sequence;
s3: sequence-to-sequence LSTM model: according to the construction sequence input of the complete battery health factor extracted in the step S2, using battery capacity data as a label to train the LSTM neural network;
s4: establishing a lithium ion battery degradation double-index model based on IPSO-PF: adopting an improved particle swarm optimization algorithm and a particle filtering algorithm to continuously optimize parameters of a double-index model so as to fit a battery capacity degradation trend;
s5: life prediction: training the model constructed in S4 and performing life prediction.
2. The method for predicting lithium ion remaining life based on an improved particle filter model of claim 1, wherein: the S1 specifically comprises the following steps: selecting the time HF of an equal voltage interval during constant current charging based on a charging phase curve 1 Time difference HF of equal current interval during constant voltage charging 2 Temperature difference HF of equal voltage interval during constant current charging 3 As health factors, specific calculation formulas are as follows:
wherein The constant current charging voltage reaches V upper Absolute time of->The constant current charging voltage reaches V lower Absolute time of (2); />The constant voltage charge time voltage reaches I upper Absolute time of->The constant voltage charge time voltage reaches I lower Absolute time of (2); />The constant current charging voltage reaches V upper Temperature of>The constant current charging voltage reaches V lower Is set at a temperature of (2);
extraction of random discharge electricity according to random discharge curve of batteryFlow integral value HF 4 Random discharge voltage sample entropy HF 5 Standard deviation of random discharge temperature HF 6 As health factors, specific calculation formulas are as follows:
where SampEn is an estimate of the conditional probability that a window of length m, which is a subsequence of the time sequence of length Nc, remains similar within the tolerance r and also matches the next electricity; t (T) K Is the Kth temperature value of the random discharge cycle of the battery; u is the average degree of random discharge cycle, n is the number of measured temperatures;
after six health factors are respectively extracted for a lithium ion battery charging and discharging curve under a random discharging working condition, optimizing the health factors extracted in a charging process through a particle swarm optimization algorithm; specifically, HF is respectively treated 1 V of (2) upper 、V lower ,HF 2 I of (2) upper 、I lower and HF3 V of (2) upper 、V lower As a parameter to be optimized, the pearson correlation coefficient of each health factor and the capacity is used as an adaptability function of the PSO, namely:
F(x)=ρ(HF,capacity) (3)
wherein F (x) is a particle swarm fitness function, HF is a corresponding health factor sequence, capability is a capacity sequence, and ρ is a pearson correlation coefficient function.
3. The method for predicting lithium ion remaining life based on an improved particle filter model of claim 2, wherein: the step S2 is specifically as follows: charging health factor HF for cycling incomplete charge and discharge 1 、HF 2 、HF 3 Together as input, the discharge health factor HF 4 、HF 5 、HF 6 Respectively as outputs; for data setsx i ∈R,t i∈R wherein xi For input of sample, t i For a target output value, R represents a real set, and D is the total number of samples; the relationship between model inputs and outputs is as follows:
t i =y(x i ,w i )+ε i (4)
wherein K (x, x) i ) Expressed as a kernel function, w= (w 1 ,w 2 ,K,w D ) T Weight vector corresponding to kernel function, w 0 Is a bias parameter;
aiming at different degradation trends of health factors, linear kernel functions, polynomial kernel functions, gaussian kernel functions and Sigmoid kernel functions are selected for linear combination, and weight coefficients of the four combined kernel functions need to be manually determined;
in the process of optimizing the RVM model through a Bayesian optimization algorithm, the method consists of two steps of Gaussian process regression and acquisition function sampling, and posterior distribution of the objective function is updated through continuously adding sampling points until the posterior distribution is consistent with the real distribution; firstly, selecting an initial super-parameter set x in a domain space χ; then calculating the probability distribution of f (x) according to the properties of the sampling set and the Gaussian distribution, and obtaining the position of the next sampling point by using an acquisition function; finally, f (x) is calculated and an objective function f is updated to reach the iteration times or meet the convergence condition;
the specific implementation process of the multi-core RVM algorithm based on Bayesian optimization is as follows: preparing a training set and a prediction set; determining super parameters and ranges which need to be optimized; and (5) building and training a Bayesian optimization RVM model.
4. The method for predicting lithium ion remaining life based on an improved particle filter model of claim 3, wherein: the step S3 is specifically as follows:
s32: constructing a data set: acquiring a complete health factor sequence according to the multi-core RVM framework in the S2, normalizing the complete health factor sequence, constructing input data, extracting capacity data and calculating to obtain a battery health state SOH, namely a label corresponding to the input; the structured dataset is:
Y={Y 1 ,Y 2 ,Y 3 ,…,Y m } (6)
for the LSTM model, the step length is set to be 10, and the current capacity value { x ] of 10 continuous loops is used 41 ,x 42 ,…,x 49 ,x 50 One prediction outputs capacity estimation values of 10 future cyclesThe 10 estimated values are used for carrying out the next prediction, the LSTM method from sequence to sequence has less iteration times and small accumulated error; y is Y i SOH value for the ith cycle; m is the number of cycles;
s32: model offline training: firstly, determining parameters to be optimized of an LSTM algorithm as time step, hidden layer node number and training times; inputting optimized parameters of the particle swarm optimization into an LSTM model, initializing other parameters, dividing the data set obtained in the step (1) into a training set and a testing set, and training the LSTM model;
s33: on-line prediction: and extracting charging data of the current cycle, combining the charging data with past data to construct and inputting the charging data into a trained model to obtain SOH predicted values of a plurality of continuous cycles.
5. The method for predicting lithium ion remaining life based on an improved particle filter model of claim 4, wherein: in the step S4, a double-index model is adopted as an empirical model of the lithium ion battery to fit the capacity degradation trend of the battery; the model built therein is as follows:
wherein k is the cycle number of charge and discharge cycles, Q k A is the capacity of the kth cycle of the battery, a 1 、a 2 、a 3 、a 4 The aging parameters to be identified for the double-index model are related to the internal characteristics of the battery, wherein a 1 、a 3 Related to the internal resistance of the current cycle of the battery, a 2 、a 4 Related to the degradation rate of the battery;
selecting x r,k =[x 1,k x 2,k x 3,k x 4,k ] T Discretizing the state variable of the battery capacity attenuation model in the kth cycle according to a state equation and an observation equation of particle filtering, wherein the state variable is as follows:
Q k =a 1,k ·exp(a 2,k ·k)+a 3,k ·exp(a 4,k ·k) (8)
in the formula ,x1,k 、x 2,k 、x 3,k 、x 4,k The initial value of (a) is a double-index model parameter a 1 、a 2 、a 3 、a 4 ;y k Is the output of the model, Q k Gamma, the capacity of lithium ion battery k =[γ 1,k γ 2,k γ 3,k γ 4,k ] T In order for the process to be noisy,is measurement noise;
in order to alleviate the particle degradation phenomenon, the particle can better adapt to the battery aging trend under the random discharging working condition, and on the basis of the established battery aging model, the capacity is predicted by using an IPSO-PF algorithm;
the specific steps of the IPSO-PF algorithm are as follows:
(1) initialization of
Setting the current charge and discharge cycle number as k; when the particle filtering is performed, k=0 is taken, and the filter is determined according to the initial prior probability density p (x r,0 ) Extracting n initial state particles
(2) Importance sampling
When k=1, 2,..n, selecting the prior probability density function as the state probability density function, i.e.
Extracting new samples based on importance probability densityThen adopting improved particle swarm algorithm to make particle set +.>Iterative optimization is carried out, and the specific steps are as follows:
step 11: setting the maximum iteration number T max The method comprises the steps of carrying out a first treatment on the surface of the The population scale n is consistent with the particle quantity selection in particle filtering, and a particle range R and a maximum speed V are set max Inertial weight omega, acceleration factor c 1 、c 2 Waiting for initial parameters; let the fitness function of particle swarm optimization be:
wherein K is the number of training set samples, L is the step size of LSTM based on sequence to sequence,the estimated value of the LSTM for L capacities in the future is the last subsection;
step 12: let iteration number t=1, set grainThe sub-initial position is the initialization parameter of the particles in the particle filteringSetting the current position as an individual history optimal pBest, setting the optimal individuals in the group as a group history optimal gBest, and calculating the fitness function value of each particle;
step 13: in the current iteration loop, if the current fitness function value is better than the historical optimal value, updating the individual historical optimal, and if the current fitness function is better than the global historical optimal value, updating the population historical optimal; at the end of one cycle, the d-th dimension of the velocity and position of each particle i are updated separately as follows:
in the formula c1 、c 1 Taking a fixed value of 2.0 as a learning factor; ω is inertial weight, initialized to 0.9, and a linear differential decremental strategy is selected for improvement, and the algorithm is as follows:
step 14: judging whether the maximum iteration number is reached or the set optimization extremum is reached, if so, optimizing, and setting particles in the PFOtherwise jump to step2;
(3) Particle weight update
Reconstructing a particle weight formula, and carrying out weight judgment in a plurality of time ranges by combining an aging model by means of a first iteration result from an LSTM sequence to a sequence, wherein the specific formula is as follows:
wherein ,is the weight of the i-th particle, +.>For the LSTM first iteration result based on sequence-to-sequence, y k In order to consider the output value of the aging model of the noise, R is the variance of the measured noise, and N is the total charge-discharge cycle number of the battery;
(4) resampling
Firstly, judging whether the cycle needs resampling or not; number N of effective particles eff The method comprises the following steps:
if N eff <N thresh old, particle pairRandom resampling is carried out to obtain a new particle set +.>Otherwise, not resampling; />
(5) State estimation
The particles in the particle set are weighted to obtain an estimated value:
(6) battery RUL prediction
Assuming that the battery is currently in the kth cycle, according to the state equation and the observation equation of particle filtering, the capacity of each particle in the current cycle for predicting j future cycles is calculated as follows:
the predicted capacity at the kth cycle for the future j cycles is:
posterior PDF approximation of battery capacity is written:
the battery RUL predicted value is calculated as follows:
6. the method for predicting lithium ion remaining life based on an improved particle filter model of claim 5, wherein: and S5, inputting the test set data into the model established in the S4, and performing model training based on particle filtering and life prediction after obtaining the double-index model parameters.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310662200.2A CN116680983A (en) | 2023-06-06 | 2023-06-06 | Lithium ion residual life prediction method based on improved particle filter model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310662200.2A CN116680983A (en) | 2023-06-06 | 2023-06-06 | Lithium ion residual life prediction method based on improved particle filter model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116680983A true CN116680983A (en) | 2023-09-01 |
Family
ID=87780529
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310662200.2A Pending CN116680983A (en) | 2023-06-06 | 2023-06-06 | Lithium ion residual life prediction method based on improved particle filter model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116680983A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117563184A (en) * | 2024-01-15 | 2024-02-20 | 东营昆宇电源科技有限公司 | Energy storage fire control system based on thing networking |
CN117630682A (en) * | 2024-01-25 | 2024-03-01 | 江西理工大学 | Random degradation process-based RUL prediction method for lithium ion battery |
CN117665630A (en) * | 2024-02-01 | 2024-03-08 | 云储新能源科技有限公司 | Battery life prediction method and system based on charge-discharge cycle data |
CN117706406A (en) * | 2024-02-05 | 2024-03-15 | 安徽布拉特智能科技有限公司 | Lithium battery health state monitoring model, method, system and storage medium |
CN117786444A (en) * | 2024-02-23 | 2024-03-29 | 深圳市拓湃新能源科技有限公司 | Battery remaining service life prediction method, device, equipment and storage medium |
-
2023
- 2023-06-06 CN CN202310662200.2A patent/CN116680983A/en active Pending
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117563184A (en) * | 2024-01-15 | 2024-02-20 | 东营昆宇电源科技有限公司 | Energy storage fire control system based on thing networking |
CN117563184B (en) * | 2024-01-15 | 2024-03-22 | 东营昆宇电源科技有限公司 | Energy storage fire control system based on thing networking |
CN117630682A (en) * | 2024-01-25 | 2024-03-01 | 江西理工大学 | Random degradation process-based RUL prediction method for lithium ion battery |
CN117630682B (en) * | 2024-01-25 | 2024-04-05 | 江西理工大学 | Random degradation process-based RUL prediction method for lithium ion battery |
CN117665630A (en) * | 2024-02-01 | 2024-03-08 | 云储新能源科技有限公司 | Battery life prediction method and system based on charge-discharge cycle data |
CN117665630B (en) * | 2024-02-01 | 2024-04-12 | 云储新能源科技有限公司 | Battery life prediction method and system based on charge-discharge cycle data |
CN117706406A (en) * | 2024-02-05 | 2024-03-15 | 安徽布拉特智能科技有限公司 | Lithium battery health state monitoring model, method, system and storage medium |
CN117706406B (en) * | 2024-02-05 | 2024-04-16 | 安徽布拉特智能科技有限公司 | Lithium battery health state monitoring model, method, system and storage medium |
CN117786444A (en) * | 2024-02-23 | 2024-03-29 | 深圳市拓湃新能源科技有限公司 | Battery remaining service life prediction method, device, equipment and storage medium |
CN117786444B (en) * | 2024-02-23 | 2024-05-14 | 深圳市拓湃新能源科技有限公司 | Battery remaining service life prediction method, device, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116680983A (en) | Lithium ion residual life prediction method based on improved particle filter model | |
Wang et al. | An improved feedforward-long short-term memory modeling method for the whole-life-cycle state of charge prediction of lithium-ion batteries considering current-voltage-temperature variation | |
CN111443294B (en) | Method and device for indirectly predicting remaining life of lithium ion battery | |
CN110888059B (en) | Charge state estimation algorithm based on improved random forest combined volume Kalman | |
Ma et al. | State of health estimation and remaining useful life prediction for lithium-ion batteries by improved particle swarm optimization-back propagation neural network | |
CN113917337A (en) | Battery health state estimation method based on charging data and LSTM neural network | |
Lipu et al. | Real-time state of charge estimation of lithium-ion batteries using optimized random forest regression algorithm | |
CN113702843B (en) | Lithium battery parameter identification and SOC estimation method based on suburb optimization algorithm | |
CN113109717B (en) | Lithium battery state of charge estimation method based on characteristic curve optimization | |
CN112834927A (en) | Lithium battery residual life prediction method, system, device and medium | |
CN113504483B (en) | Integrated prediction method for residual life of lithium ion battery considering uncertainty | |
Anton et al. | A new predictive model for the state-of-charge of a high-power lithium-ion cell based on a PSO-optimized multivariate adaptive regression spline approach | |
CN115201686B (en) | Lithium ion battery health state assessment method under incomplete charge and discharge data | |
Duan et al. | Battery SOH estimation and RUL prediction framework based on variable forgetting factor online sequential extreme learning machine and particle filter | |
CN114660497A (en) | Lithium ion battery service life prediction method aiming at capacity regeneration phenomenon | |
CN114091282A (en) | Lithium ion battery state estimation method and system based on fractional order model | |
CN114384435A (en) | WSA-LSTM algorithm-based self-adaptive prediction method for residual service life of new energy automobile power battery | |
CN115856633A (en) | Lithium ion battery capacity estimation method based on graph neural network | |
CN116774088A (en) | Lithium ion battery health state estimation method based on multi-objective optimization | |
CN115902642A (en) | Battery state of charge estimation method and device, electronic equipment and storage medium | |
Cao et al. | A flexible battery capacity estimation method based on partial voltage curves and polynomial fitting | |
Zhang et al. | Lithium-ion battery calendar aging mechanism analysis and impedance-based State-of-Health estimation method | |
Zhu et al. | State of health prediction for li-ion batteries with end-to-end deep learning | |
Tang et al. | Parameter identification for lithium batteries: Model variable-coupling analysis and a novel cooperatively coevolving identification algorithm | |
CN111337833B (en) | Lithium battery capacity integrated prediction method based on dynamic time-varying weight |
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 |