CN111103544B - Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF - Google Patents

Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF Download PDF

Info

Publication number
CN111103544B
CN111103544B CN201911368020.3A CN201911368020A CN111103544B CN 111103544 B CN111103544 B CN 111103544B CN 201911368020 A CN201911368020 A CN 201911368020A CN 111103544 B CN111103544 B CN 111103544B
Authority
CN
China
Prior art keywords
ion battery
lithium ion
capacity
prediction
value
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.)
Active
Application number
CN201911368020.3A
Other languages
Chinese (zh)
Other versions
CN111103544A (en
Inventor
薛安荣
于彬鹏
杨婉琳
陈伟鹤
蔡涛
盘朝奉
何志刚
李骁淳
王丽梅
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.)
Jiangsu University
Original Assignee
Jiangsu 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 Jiangsu University filed Critical Jiangsu University
Priority to CN201911368020.3A priority Critical patent/CN111103544B/en
Publication of CN111103544A publication Critical patent/CN111103544A/en
Application granted granted Critical
Publication of CN111103544B publication Critical patent/CN111103544B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • 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
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Algebra (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)
  • Secondary Cells (AREA)
  • Tests Of Electric Status Of Batteries (AREA)

Abstract

The invention discloses a lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF, belonging to the field of lithium ion battery remaining service life prediction of new energy electric vehicles, and comprising the following specific steps of: analyzing and extracting performance degradation characteristic parameters of the lithium ion battery from the voltage, the current and the temperature of the lithium ion battery, fusing the characteristic parameters by using an improved principal component analysis method as a health index of the lithium ion battery, fully representing the performance degradation characteristics of the lithium ion battery and not containing redundant information; training a lithium ion battery capacity prediction model based on a long-time memory neural network to predict the capacity of the lithium ion battery, taking the capacity prediction value of the LSTM prediction model as an observation value of the particle filter prediction model, adjusting and updating the capacity prediction value in each step of iteration process of a particle filter algorithm, and comparing the capacity prediction value with a capacity failure threshold value so as to predict the remaining service life of the lithium ion battery. The invention can effectively monitor and predict the performance degradation process of the lithium ion battery.

Description

Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF
Technical Field
The invention belongs to the technical field of lithium ion batteries of electric vehicles, and particularly relates to a lithium ion battery remaining service life prediction method based on LSTM and PF.
Background
With the increasing of environmental awareness of people, new energy electric vehicles are favored by users. The lithium ion battery is widely applied to new energy electric vehicles due to the advantages of good safety, large specific energy, high charging and discharging efficiency and the like, but the performance degradation of the lithium ion battery is inevitable, and the performance degradation is easy to cause system failure and even catastrophic accidents of the new energy electric vehicles. Therefore, an accurate and reliable method for predicting the remaining service life of the lithium ion battery is found, the remaining service life of the lithium ion battery is accurately predicted, and the battery with the fault or the service life ending is replaced in time, so that the method is significant for avoiding and preventing accidents.
The remaining useful life of a lithium ion battery is the number of cycles that the lithium ion battery undergoes from the current charge and discharge cycle to the time the lithium ion battery fails (generally when the lithium ion battery capacity is considered to be 70% of the initial rated capacity). At present, a plurality of methods for predicting the remaining service life of the lithium ion battery mainly comprise a model-based method and a data-driven method. The model-based method is difficult to accurately predict the residual service life of the lithium ion battery because the accurate model is difficult to establish. Data-driven methods such as a correlation vector machine and a support vector machine cannot effectively process the long dependence of sequence data, and are not suitable for predicting the residual service life of the lithium ion battery.
At present, most researches take parameters such as battery capacity, resistance and the like as the health index of the lithium ion battery to predict the residual service life of the lithium ion battery, but the parameters are difficult to directly measure in the battery operation process. In a few researches, parameters which are easy to measure such as equal pressure drop time intervals and the like are independently or jointly used as health indexes of lithium ion batteries, so that the health indexes are insufficient or redundant, and the prediction accuracy is influenced. The particle filter algorithm can effectively track and predict a nonlinear system and can effectively predict the residual service life of the lithium ion battery, but the method has no observation value in the prediction stage, so that the state cannot be updated, the prediction accuracy is low, and the prediction accuracy excessively depends on the accuracy of the model. The long-time and short-time memory neural network can effectively process the sequence data long dependence, is widely applied to the prediction of the remaining service life of the lithium ion battery at present, but the prediction accuracy is reduced because a capacity measurement value is lost in the prediction stage and a prediction value is used as model input.
Disclosure of Invention
The invention aims to provide a lithium ion battery remaining service life prediction method based on LSTM and PF so as to improve the lithium ion battery remaining service life prediction accuracy. Aiming at the problem that parameters such as battery capacity, resistance and the like which are commonly used as health indexes of the lithium ion battery are difficult to measure in the operation process of the lithium ion battery, the invention utilizes Spearman correlation analysis to extract performance degradation characteristic parameters of the lithium ion battery from the voltage, the current and the temperature of the lithium ion battery. The improved principal component analysis method is utilized to fuse characteristic parameters to replace capacity to serve as the health index of the lithium ion battery, fully represent performance degradation characteristics of the lithium ion battery, do not contain redundant information, and meanwhile remove noise data in the principal component analysis process. According to the invention, a lithium ion battery capacity model based on a long-time and short-time memory LSTM neural network is constructed, and the performance degradation trend of the lithium ion battery is represented. Aiming at the problems that the state of the particle filter prediction model cannot be updated and the prediction accuracy is low due to the fact that no measured value exists in the prediction stage, the capacity prediction value obtained by the LSTM prediction model is used as an observation value, the capacity prediction value is updated iteratively through a particle filter algorithm, the capacity prediction value is compared with a failure threshold value, the remaining service life of the lithium ion battery is predicted, and the prediction accuracy is improved. The specific technical scheme is as follows:
the lithium ion battery remaining service life prediction method based on LSTM and PF comprises the following steps:
step 1, extracting performance degradation characteristic parameters of the lithium ion battery by utilizing Spearman correlation analysis: extracting characteristic parameters from the voltage, the current and the temperature of the lithium ion battery, and determining the characteristic parameters capable of representing the performance degradation of the lithium ion battery by utilizing Spearman correlation analysis;
step 2, constructing a health index HI by using an improved principal component analysis method: improving a principal component analysis method, retaining the difference of characteristic parameter information when the dimension and quantity level difference is eliminated, and fusing characteristic parameters by using the improved principal component analysis method to replace capacity as a lithium ion battery health index;
and 3, predicting the remaining service life of the lithium ion battery based on the long-time and short-time memory LSTM neural network and the particle filter PF: training an LSTM prediction model to predict the lithium ion battery capacity, taking a double-exponential capacity degradation model as a PF prediction model state transition equation, taking a capacity prediction value obtained by the LSTM prediction model as an observation value, updating and adjusting the lithium ion battery capacity prediction value in each iteration of a particle filter algorithm, and comparing the lithium ion battery capacity prediction value with a lithium ion battery capacity failure threshold CapEOLAnd predicting the residual service life of the lithium ion battery.
Further, the characteristic parameters of the performance degradation of the lithium ion battery, which are analyzed and extracted by utilizing Spearman correlation, comprise discharge platform period duration DST, discharge platform period voltage change rate DVT, maximum discharge temperature occurrence time DTMT and constant current charging time CST.
Further, the method for constructing the health index HI by using the improved principal component analysis method comprises the following steps:
step 2.1, constructing a lithium ion battery performance degradation characteristic parameter matrix P*As shown in formula 2.1:
Figure BDA0002338949520000031
wherein n is the number of samples,
Figure BDA0002338949520000032
respectively setting the i-th cycle discharge platform period duration, the discharge platform period voltage change rate, the discharge highest temperature occurrence time and the constant current charging time;
step 2.2, the standardization process of the principal component analysis method is improved, the difference of characteristic parameter information is kept when the dimension and quantity level difference is eliminated, and the improved standardization method is shown as a formula 2.2:
Figure BDA0002338949520000033
wherein
Figure BDA0002338949520000034
Is the mean value of the jth characteristic parameter, xijFor the ith cycle, the jth characteristic parameter
Figure BDA0002338949520000035
Normalized performance degradation characteristic parameter matrix P is as follows (i ═ 1,2, …, n, j ═ 1,2,3,4), as in equation 2.3:
Figure BDA0002338949520000036
wherein DSTi,DVTi,DTMTi,CSTi(i-1, 2, …, n) are respectively the normalized values of the ith cycle discharge plateau period duration, the discharge plateau period voltage change rate, the highest discharge temperature occurrence time and the constant current charging time;
step 2.3, calculating covariance matrix COV of standardized characteristic parameter matrix Pp：
Figure BDA0002338949520000037
Wherein n is the number of samples, PTIs a transposed matrix of the matrix P;
step 2.4, calculating covariance matrix COVpThe characteristic values are sorted in descending order, and the sorted characteristic value is lambda1,λ2,λ3,λ4Each eigenvalue corresponds to a standard eigenvector of V1,V2,V3,V4；
Step 2.5, calculating the corresponding contribution rate and the accumulated contribution rate of each sorted characteristic value, as shown in formulas 2.5 and 2.6:
Figure BDA0002338949520000038
Figure BDA0002338949520000039
wherein λi(i ═ 1,2,3,4) is the ith characteristic value, CiAs a characteristic value λiContribution ratio of (C), CSiAs a characteristic value λi(ii) cumulative contribution rate of;
step 2.6, with cumulative contribution rate CSiDetermining the main component according to the condition that the main component is greater than or equal to 90%, and calculating a component matrix as the health index of the lithium ion battery, wherein the component matrix is shown as a formula 2.7:
Figure BDA00023389495200000310
wherein m is the number of main components, VjCorresponding to the standard feature vector, HI, for principal component*Is the lithium ion battery health index.
Further, the method for predicting the residual service life of the lithium ion battery based on the long-time and short-time memory LSTM neural network and the particle filter PF comprises the following steps:
step 3.1, normalizing the health index and acquiring a mapping relation between the normalized health index HI and the capacity Cap by using a polynomial fitting method;
step 3.2, taking the previous T circulation health indexes as training data, taking T as a prediction starting point and taking the rest data as test data to
Figure BDA0002338949520000041
In order to be an input, the user can select,
Figure BDA0002338949520000042
constructing a training data set for the output, wherein
Figure BDA0002338949520000043
The number of input parameters of the LSTM prediction model is the true value of the ith circulation health index, i is t-k, …, and t and k are the number of input parameters of the LSTM prediction model;
step 3.3, training an LSTM prediction model by using the training data, as shown in formula 3.1:
Figure BDA0002338949520000044
wherein XtIn order to input the prediction model, the model is input,
Figure BDA00023389495200000410
for the health index prediction value, further using an RMSprop optimization algorithm to accelerate the model training speed, and adding an L2 regularization term to avoid the problem of model overfitting;
and 3.4, predicting the health index of the lithium ion battery by using an LSTM prediction model from a prediction starting point T, and acquiring a capacity prediction value according to the normalized health index and the capacity mapping relation, wherein the model input X in the prediction stage is input into the prediction stagetAs shown in equation 3.2:
Figure BDA0002338949520000045
where w is the number of true values of the input parameter,
Figure BDA0002338949520000046
for the ith cycle health index predictor,
Figure BDA0002338949520000047
the real value of the health index;
step 3.5, constructing a state space model by using the double-index capacity degradation model as shown in the formula 3.3:
Figure BDA0002338949520000048
where Cap (k) is the capacity of the kth cycle, k is the number of cycles, p1,p2,p3,p4For a dual-exponential capacity degradation model parameter, v1,v2,v3,v4,v5Is noise;
step 3.6, using particle filteringAlgorithm tracking and determining parameter p of dual-exponential capacity degradation model shown in formula 3.31,p2,p3,p4The method comprises the following specific steps:
a. initializing algorithm related parameters: the number of particles N, process noise, measurement noise and a state initial value;
b. particle initialization: initializing particles according to the initial state values, wherein the weights of the particles are all
Figure BDA0002338949520000049
c. Importance sampling: calculating the particle value at the current moment according to a state transition equation shown as a formula 3.3;
d. weight of particles: calculating the weight of each particle and normalizing by taking the capacity value of the training data set as an observed value;
e. resampling: resampling according to the particle weight;
f. repeating the steps c-e until the cycle number is the prediction starting point T;
g. outputting a dual-exponential capacity degradation model parameter p1,p2,p3,p4；
And 3.7, taking the formula 3.3 as a state transition equation, taking the capacity predicted value obtained by the LSTM prediction model as an observed value, and iteratively updating the capacity predicted value by utilizing a particle filter algorithm.
The invention has the advantages of
The method is applied to the technical field of lithium ion batteries of new energy electric vehicles, can be used for effectively predicting the residual service life of the lithium ion battery by directly measuring parameters in the running process of the lithium ion battery, and has important significance for the health management of the lithium ion battery. Aiming at the problem that parameters such as capacity, resistance and the like which are commonly used as lithium ion battery health indexes are difficult to measure in the actual operation process of the lithium ion battery, the method utilizes Spearman correlation analysis to extract performance degradation characteristic parameters of the lithium ion battery from the voltage, the current and the temperature of the lithium ion battery, utilizes an improved principal component analysis method to fuse the characteristic parameters as the lithium ion battery health indexes, fully represents the performance degradation characteristics of the lithium ion battery, does not contain redundant information, and simultaneously removes noise data in the principal component analysis process. According to the invention, a long-time and short-time memory LSTM neural network capable of effectively expressing the dependency of the sequence data length is used as a lithium ion battery capacity prediction model to represent the performance degradation trend of the lithium ion battery, a capacity prediction value obtained by the LSTM prediction model is used as an observation value, the capacity prediction value is iteratively updated by using a particle filter algorithm, and the capacity prediction value is compared with a failure threshold value so as to predict the remaining service life of the lithium ion battery, thereby improving the prediction accuracy.
Drawings
FIG. 1 is a schematic general flow chart of a lithium ion battery remaining service life prediction method based on LSTM and PF according to the present invention;
FIG. 2 is a diagram of normalized lithium ion battery health index versus lithium ion battery capacity;
fig. 3 is a diagram of a prediction result of the remaining service life of the lithium ion battery.
Detailed Description
The following clearly and completely describes the technical solution in the embodiment of the present invention by taking the operational data of lithium ion battery No. B0005 in the NASA public data set shown in table 1 as an example and combining with the drawings in the embodiment of the present invention.
TABLE 1B 0005 lithium ion batteries
Constant current charging current/A Charge cut-off voltage/V Discharge current/A Discharge cut-off voltage/V Rated capacity/Ah
1.5 4.2 2.0 2.7 2.0
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
As shown in fig. 1, according to the embodiment of the present invention, the method for predicting the remaining service life of a lithium ion battery based on LSTM and PF comprises three basic steps: extracting performance degradation characteristic parameters of the lithium ion battery; constructing a health index by using an improved principal component analysis method; and predicting the remaining service life of the lithium ion battery based on the long-time memory LSTM neural network and the particle filter PF.
Firstly, extracting characteristic parameters of performance degradation of the lithium ion battery
Because parameters such as battery capacity, resistance and the like which are commonly used as lithium ion battery health indexes are difficult to measure in the operation process of the lithium ion battery, the invention extracts performance degradation characteristic parameters of the lithium ion battery from voltage, current and temperature which can be directly measured in the operation process of the lithium ion battery by utilizing Spearman correlation analysis, wherein the performance degradation characteristic parameters comprise discharge platform period duration DST, discharge platform period voltage change rate DVT, discharge highest temperature occurrence time DTMT and constant current charging time CST. The Spearman correlation coefficients between characteristic parameters and discharge capacity Cap and between characteristic parameters are shown in table 2.
TABLE 2 characteristic parameters and discharge capacity Spearman correlation coefficients
Cap DST DVT DTMT CST
Cap
1 0.995671 0.996170 0.999522 0.993173
DST 0.995671 1 0.996389 0.995804 0.990124
DVT 0.996170 0.996389 1 0.997220 0.992654
DTMT 0.999522 0.995804 0.997220 1 0.993722
CST 0.993173 0.990124 0.992654 0.993722 1
Second, the health index is constructed by utilizing an improved principal component analysis method
According to the Spearman correlation coefficient between the characteristic parameters of the lithium ion battery and the battery capacity and between the characteristic parameters, the characteristic parameters can be seen to effectively represent the performance degradation trend of the lithium ion battery, but a large amount of redundant information exists between the characteristic parameters, so that the characteristic parameters are fused by using an improved principal component analysis method to serve as the health index of the lithium ion battery, the performance degradation state of the lithium ion battery is fully represented, and the redundant information is not contained.
In this embodiment, constructing the health index by using the improved principal component analysis method includes the following steps:
step 1, constructing a lithium ion battery performance degradation characteristic parameter matrix P*As shown in formula 2.1:
Figure BDA0002338949520000071
wherein n is the number of samples,
Figure BDA0002338949520000072
respectively setting the i-th cycle discharge platform period duration, the discharge platform period voltage change rate, the discharge highest temperature occurrence time and the constant current charging time;
step 2, improving the standardization process of the principal component analysis method, and keeping the difference of characteristic parameter information when the dimension and quantity level difference is eliminated, wherein the improved standardization method is shown as a formula 2.2:
Figure BDA0002338949520000073
wherein
Figure BDA0002338949520000074
Is the mean value of the jth characteristic parameter, xijFor the ith cycle, the jth characteristic parameter
Figure BDA0002338949520000075
Normalized performance degradation characteristic parameter matrix P is as follows (i ═ 1,2, …, n, j ═ 1,2,3,4), as in equation 2.3:
Figure BDA0002338949520000076
wherein DSTi,DVTi,DTMTi,CSTi(i-1, 2, …, n) are respectively the normalized values of the ith cycle discharge plateau period duration, the discharge plateau period voltage change rate, the highest discharge temperature occurrence time and the constant current charging time;
step 3, calculating a covariance matrix COV of the standardized characteristic parameter matrix PpAs shown in formula 2.4:
Figure BDA0002338949520000077
wherein n represents the number of samples, PTA transposed matrix representing the matrix P;
step 4, calculating a covariance matrix COVpThe characteristic values are sorted in descending order, and the sorted characteristic value is lambda1,λ2,λ3,λ4Each eigenvalue corresponds to a standard eigenvector of V1,V2,V3,V4；
Step 5, calculating the corresponding contribution rate and the accumulated contribution rate of each sorted characteristic value, as shown in formulas 2.5 and 2.6:
Figure BDA0002338949520000078
Figure BDA0002338949520000079
wherein λi(i ═ 1,2,3,4) is the ith characteristic value, CiAs a characteristic value λiContribution ratio of (C), CSiAs a characteristic value λi(ii) cumulative contribution rate of;
step 6, accumulating contribution rate CSiDetermining the main component according to the condition that the main component is greater than or equal to 90%, and calculating a component matrix as the health index of the lithium ion battery, wherein the component matrix is shown as a formula 2.7:
Figure BDA00023389495200000710
wherein m is the number of main components, VjCorresponding to the standard feature vector, HI, for principal component*Is the lithium ion battery health index.
Predicting the remaining service life of the lithium ion battery based on long-time memory LSTM neural network and particle filter PF
In this example, predicting the remaining service life of the lithium ion battery based on the long-time and short-time memory LSTM neural network and the particle filter PF comprises the following steps:
step 1, normalizing the health index and obtaining a mapping relation between the normalized health index HI and the capacity Cap by utilizing a polynomial fitting method, wherein the relation between the normalized health index and the lithium ion battery discharge capacity is shown in figure 2;
step 2, taking the former 90 circulation health indexes as training data and the rest data as test data to
Figure BDA0002338949520000081
In order to be an input, the user can select,
Figure BDA0002338949520000082
constructing a training data set for the output, wherein
Figure BDA0002338949520000083
The real value of the ith cycle health index is t-10, … and t;
step 3, training an LSTM prediction model by using the training data, as shown in formula 3.1:
Figure BDA0002338949520000084
wherein XtIn order to input the prediction model, the model is input,
Figure BDA0002338949520000085
for the health index prediction value, further using an RMSprop optimization algorithm to accelerate the model training speed, and adding an L2 regularization term to avoid the problem of model overfitting;
step 4, predicting the health index of the lithium ion battery by using an LSTM prediction model from the 90 th lithium ion battery charging and discharging cycle, obtaining a capacity prediction value according to the normalized health index and the capacity mapping relation, and inputting X by the model in the prediction stagetAs shown in equation 3.2:
Figure BDA0002338949520000086
where w is the number of true values of the input parameter,
Figure BDA0002338949520000087
for the ith cycle health index predictor,
Figure BDA0002338949520000088
the real value of the health index;
step 5, constructing a state space model by using the double-index capacity degradation model as shown in the formula 3.3:
Figure BDA0002338949520000089
where Cap (k) is the capacity of the kth cycle, k is the number of cycles, p1,p2,p3,p4For a dual-exponential capacity degradation model parameter, v1,v2,v3,v4,v5Is noise;
step 6, tracking and determining a parameter p of the dual-exponential capacity degradation model shown in the formula 3.3 by using a particle filter algorithm1,p2,p3,p4The method comprises the following specific steps:
a. initializing algorithm related parameters: the number of particles N, process noise, measurement noise and a state initial value;
b. particle initialization: initializing particles according to the initial state values, wherein the weights of the particles are all
Figure BDA00023389495200000810
c. Importance sampling: calculating the particle value at the current moment according to a state transition equation shown as a formula 3.3;
d. weight of particles: calculating the weight of each particle and normalizing by taking the capacity value of the training data set as an observed value;
e. resampling: resampling according to the particle weight;
f. repeating steps c-e until the cycle number is 90;
g. outputting a dual-exponential capacity degradation model parameter p1,p2,p3,p4；
Step 7, taking equation 3.3 as a state transition equation, taking the capacity predicted value obtained by the LSTM prediction model as an observed value, and iteratively updating the capacity predicted value by using a particle filter algorithm, wherein the specific steps of the particle filter algorithm are substantially consistent with those of step 6, and what needs to be emphasized is that: in the step, the particle filter algorithm iteratively updates the volume predicted value, the volume predicted value obtained by the LSTM prediction model is used as an observed value to calculate the weight of each particle, the updated volume predicted value is calculated after resampling, and the cycle end condition is that the volume predicted value is less than or equal to a volume failure threshold CapEOLAnd outputting the cycle number k at the time, so that the lithium ion battery has the residual service lifeHit prediction value RUL ═ k, Cap in this exampleEOLSet to 1.4 Ah;
step 8, comparing the prediction result of the remaining service life of the lithium ion battery based on the LSTM and the PF with the prediction results of the LSTM prediction model and the PF prediction model, as shown in fig. 3 and table 3, the Root Mean Square Error (RMSE) is:
Figure BDA0002338949520000091
wherein m is the number of prediction cycles,
Figure BDA0002338949520000092
for the capacity prediction value of the ith prediction cycle,
Figure BDA0002338949520000093
the capacity true value of the ith prediction cycle.
TABLE 3 comparison of predicted results
Starting point of prediction True RUL Predicting RUL Absolute error RMSE
LSTM 90 33 54 21 0.0530456
PF 90 33 15 18 0.0422868
LSTM-PF 90 33 25 8 0.0175367
In summary, the invention discloses a lithium ion battery remaining service life prediction method based on LSTM and PF, belonging to the field of new energy electric vehicle lithium ion battery remaining service life prediction, and comprising the following specific steps: (1) extracting performance degradation characteristic parameters of the lithium ion battery from the voltage, the current and the temperature of the lithium ion battery by utilizing Spearman correlation analysis, fusing the characteristic parameters by utilizing an improved principal component analysis method to serve as a Health Index (HI) of the lithium ion battery, fully representing the performance degradation characteristics of the lithium ion battery and not containing redundant information; (2) training a lithium ion battery capacity prediction model based on a Long Short-Term Memory (LSTM) neural network to predict the capacity of a lithium ion battery, taking the capacity prediction value of the LSTM prediction model as an observed value of a Particle Filter (PF) prediction model, adjusting and updating the capacity prediction value in each iteration process of a Particle Filter algorithm, and comparing the capacity prediction value with a capacity failure threshold value so as to predict the Remaining service Life (RUL) of the lithium ion battery. The method can be applied to prediction of the remaining service life of the lithium ion battery of the new energy electric vehicle, and can effectively monitor and predict the performance degradation process of the lithium ion battery.

Claims (4)

1. The lithium ion battery remaining service life prediction method based on LSTM and PF is characterized by comprising the following steps:
step 1, extracting performance degradation characteristic parameters of the lithium ion battery by utilizing Spearman correlation analysis: extracting characteristic parameters from the voltage, the current and the temperature of the lithium ion battery, and determining the characteristic parameters capable of representing the performance degradation of the lithium ion battery by utilizing Spearman correlation analysis;
step 2, constructing a health index HI by using an improved principal component analysis method: improving a principal component analysis method, retaining the difference of characteristic parameter information when the dimension and quantity level difference is eliminated, and fusing characteristic parameters by using the improved principal component analysis method to replace capacity as a lithium ion battery health index;
and 3, predicting the remaining service life of the lithium ion battery based on the long-time and short-time memory LSTM neural network and the particle filter PF: training an LSTM prediction model to predict the lithium ion battery capacity, taking a double-exponential capacity degradation model as a PF prediction model state transition equation, taking a capacity prediction value obtained by the LSTM prediction model as an observation value, updating and adjusting the lithium ion battery capacity prediction value in each iteration of a particle filter algorithm, and comparing the lithium ion battery capacity prediction value with a lithium ion battery capacity failure threshold CapEOLAnd predicting the residual service life of the lithium ion battery.
2. The lithium ion battery residual service life prediction method based on LSTM and PF of claim 1, wherein the characteristic parameters of performance degradation of lithium ion battery extracted by Spearman correlation analysis comprise discharge plateau period duration DST, discharge plateau period voltage change rate DVT, maximum discharge temperature occurrence time DTMT and constant current charging time CST.
3. The LSTM and PF based prediction method of remaining useful life of li-ion battery of claim 1 wherein the constructing of health index HI using improved principal component analysis comprises the steps of:
step 2.1, constructing a lithium ion battery performance degradation characteristic parameter matrix P*As shown in formula 2.1:
Figure FDA0003281639990000011
where n is the number of samples, i is 1,2, …, n,
Figure FDA0003281639990000012
respectively setting the i-th cycle discharge platform period duration, the discharge platform period voltage change rate, the discharge highest temperature occurrence time and the constant current charging time;
step 2.2, the standardization process of the principal component analysis method is improved, the difference of characteristic parameter information is kept when the dimension and quantity level difference is eliminated, and the improved standardization method is shown as a formula 2.2:
Figure FDA0003281639990000021
wherein
Figure FDA0003281639990000022
Is the mean value of the jth characteristic parameter, xijFor the ith cycle, the jth characteristic parameter
Figure FDA0003281639990000023
J ═ 1,2,3,4, and the normalized performance degradation characteristic parameter matrix P is as in equation 2.3:
Figure FDA0003281639990000024
wherein DSTi,DVTi,DTMTi,CSTi(i-1, 2, …, n) is the ith cycle discharge plateauThe standardized values of the duration, the voltage change rate of the discharge plateau, the highest discharge temperature occurrence time and the constant current charging time;
step 2.3, calculating covariance matrix COV of standardized characteristic parameter matrix Pp：
Figure FDA0003281639990000025
Wherein, PTIs a transposed matrix of the matrix P;
step 2.4, calculating covariance matrix COVpThe characteristic values are sorted in descending order, and the sorted characteristic value is lambda1,λ2,λ3,λ4Each eigenvalue corresponds to a standard eigenvector of V1,V2,V3,V4；
Step 2.5, calculating the corresponding contribution rate and the accumulated contribution rate of each sorted characteristic value, as shown in formulas 2.5 and 2.6:
Figure FDA0003281639990000026
Figure FDA0003281639990000027
wherein, when i is 1,2,3,4, λiIs the ith characteristic value, CiAs a characteristic value λiContribution ratio of (C), CSiAs a characteristic value λi(ii) cumulative contribution rate of;
step 2.6, with cumulative contribution rate CSiDetermining the main component according to the condition that the main component is greater than or equal to 90%, and calculating a component matrix as the health index of the lithium ion battery, wherein the component matrix is shown as a formula 2.7:
Figure FDA0003281639990000028
wherein m is the number of main components, VjCorresponding to the standard feature vector, HI, for principal component*Is the lithium ion battery health index.
4. The LSTM and PF based lithium ion battery remaining service life prediction method of claim 1, wherein predicting the lithium ion battery remaining service life based on long-and-short-term memory LSTM neural network and particle filter PF comprises the steps of:
step 3.1, normalizing the health index and acquiring a mapping relation between the normalized health index HI and the capacity Cap by using a polynomial fitting method;
step 3.2, taking the previous T circulation health indexes as training data, taking T as a prediction starting point and taking the rest data as test data to
Figure FDA0003281639990000031
In order to be an input, the user can select,
Figure FDA0003281639990000032
constructing a training data set for the output, wherein
Figure FDA0003281639990000033
The number of input parameters of the LSTM prediction model is the true value of the ith circulation health index, i is t-k, …, and t and k are the number of input parameters of the LSTM prediction model;
step 3.3, training an LSTM prediction model by using the training data, as shown in formula 3.1:
Figure FDA0003281639990000034
wherein XtIn order to input the prediction model, the model is input,
Figure FDA0003281639990000035
for the health index prediction value, further using an RMSprop optimization algorithm to accelerate the model training speed, and adding an L2 regularization term to avoid the problem of model overfitting;
step 3.4, starting from the prediction starting point T, using the LSTM prediction modelMeasuring the health index of the lithium ion battery, and obtaining a capacity predicted value according to the normalized health index and the capacity mapping relation, wherein the model input X in the prediction stagetAs shown in equation 3.2:
Figure FDA0003281639990000036
where w is the number of true values of the input parameter,
Figure FDA0003281639990000037
for the ith cycle health index predictor,
Figure FDA0003281639990000038
the real value of the health index;
step 3.5, constructing a state space model by using the double-index capacity degradation model as shown in the formula 3.3:
Figure FDA0003281639990000039
where Cap (k) is the capacity of the kth cycle, k is the number of cycles, p1,p2,p3,p4For a dual-exponential capacity degradation model parameter, v1,v2,v3,v4,v5Is noise;
step 3.6, tracking and determining a parameter p of the dual-exponential capacity degradation model shown in the formula 3.3 by using a particle filter algorithm1,p2,p3,p4The method comprises the following specific steps:
a. initializing algorithm related parameters: the number of particles N, process noise, measurement noise and a state initial value;
b. particle initialization: initializing particles according to the initial state values, wherein the weights of the particles are all
Figure FDA00032816399900000310
c. Importance sampling: calculating the particle value at the current moment according to a state transition equation shown as a formula 3.3;
d. weight of particles: calculating the weight of each particle and normalizing by taking the capacity value of the training data set as an observed value;
e. resampling: resampling according to the particle weight;
f. repeating the steps c-e until the cycle number is the prediction starting point T;
g. outputting a dual-exponential capacity degradation model parameter p1,p2,p3,p4；
And 3.7, taking the formula 3.3 as a state transition equation, taking the capacity predicted value obtained by the LSTM prediction model as an observed value, and iteratively updating the capacity predicted value by utilizing a particle filter algorithm.
CN201911368020.3A 2019-12-26 2019-12-26 Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF Active CN111103544B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911368020.3A CN111103544B (en) 2019-12-26 2019-12-26 Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911368020.3A CN111103544B (en) 2019-12-26 2019-12-26 Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF

Publications (2)

Publication Number Publication Date
CN111103544A CN111103544A (en) 2020-05-05
CN111103544B true CN111103544B (en) 2021-12-21

Family

ID=70424401

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911368020.3A Active CN111103544B (en) 2019-12-26 2019-12-26 Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF

Country Status (1)

Country Link
CN (1) CN111103544B (en)

Families Citing this family (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111597759B (en) * 2020-05-18 2022-04-26 中车永济电机有限公司 Construction method of residual service life prediction model of IGBT (insulated Gate Bipolar translator) of converter device
CN111639442A (en) * 2020-06-09 2020-09-08 山东交通学院 Method and system for screening influence factors of residual service life of power lithium battery
CN111784010B (en) * 2020-07-09 2022-07-26 哈尔滨工程大学 Method and system for predicting residual service life of electric gate valve
CN111680848A (en) * 2020-07-27 2020-09-18 中南大学 Battery life prediction method based on prediction model fusion and storage medium
CN112038670B (en) * 2020-07-31 2021-09-14 上海捷氢科技有限公司 Proton exchange membrane fuel cell and water state monitoring method and device thereof
CN111707956B (en) * 2020-08-11 2022-08-30 中北大学 Method for predicting health state and residual life of multi-type lithium ion battery pack management system
CN111999648A (en) * 2020-08-20 2020-11-27 浙江工业大学 Lithium battery residual life prediction method based on long-term and short-term memory network
JP7314880B2 (en) * 2020-08-21 2023-07-26 トヨタ自動車株式会社 BATTERY DIAGNOSTIC APPARATUS, METHOD, PROGRAM AND VEHICLE
CN112327165B (en) * 2020-09-21 2021-07-13 电子科技大学 Battery SOH prediction method based on unsupervised transfer learning
CN112285565B (en) * 2020-09-21 2021-07-13 电子科技大学 Method for predicting SOH (State of health) of battery by transfer learning based on RKHS (remote keyless entry) domain matching
KR20220045451A (en) 2020-10-05 2022-04-12 주식회사 엘지에너지솔루션 Apparatus and method for diagnosing state of battery
CN112241608B (en) * 2020-10-13 2022-08-26 国网湖北省电力有限公司电力科学研究院 Lithium battery life prediction method based on LSTM network and transfer learning
CN113711458A (en) * 2020-11-27 2021-11-26 东莞新能德科技有限公司 Battery capacity estimation method, electronic device, and storage medium
CN112684363A (en) * 2020-12-18 2021-04-20 北京工业大学 Lithium ion battery health state estimation method based on discharge process
CN112881518B (en) * 2021-01-08 2022-07-12 东冶及策河北能源技术有限公司 Method for predicting residual life of dynamic filter compensator
CN112685961B (en) * 2021-01-12 2022-06-21 武汉大学 Method and system for predicting remaining service life of analog circuit
CN113009368A (en) * 2021-02-24 2021-06-22 武汉科技大学 MFF-based multi-core GPR algorithm-based lithium battery residual life prediction method
CN113176518B (en) * 2021-03-24 2023-08-15 西安理工大学 Lithium battery health state estimation method
CN113065283A (en) * 2021-03-26 2021-07-02 深圳技术大学 Battery life prediction method, system, electronic device and storage medium
US20220357406A1 (en) * 2021-05-04 2022-11-10 Toyota Motor North America, Inc. Electric vehicle battery maintenance
CN113361197B (en) * 2021-06-08 2022-10-25 山东大学 Method and system for predicting remaining service life of lithium battery
CN113406505B (en) * 2021-07-22 2024-07-26 中国第一汽车股份有限公司 Residual life prediction method and device for fuel cell
WO2023052910A1 (en) * 2021-10-01 2023-04-06 Kpit Technologies Limited System and method for estimating state of health and remaining useful life of a battery
CN114295994A (en) * 2021-12-23 2022-04-08 重庆邮电大学 PCA-RVM-based lithium ion battery remaining service life prediction method
CN114325393B (en) * 2021-12-28 2024-07-19 惠州富基能源科技有限公司 Self-adaptive estimation method for SOH (self-adaptive state of charge) of lithium ion battery pack based on PF (power factor) and GPR (power factor)
CN114509690A (en) * 2022-04-19 2022-05-17 杭州宇谷科技有限公司 PCA (principal component analysis) decomposition-based lithium battery cell charging and discharging abnormity detection method and system
CN114839538A (en) * 2022-04-26 2022-08-02 北京工商大学 Method for extracting degradation characteristics of lithium ion battery for estimating residual life
CN115047350B (en) * 2022-06-24 2023-04-18 哈尔滨工业大学 Digital-analog linkage based lithium ion battery remaining service life prediction method
CN115575843A (en) * 2022-10-25 2023-01-06 楚能新能源股份有限公司 Lithium ion battery life prediction method
CN116299005B (en) * 2023-02-07 2023-09-05 江南大学 Power battery health state prediction method based on AAF and deep learning
CN116176353B (en) * 2023-04-26 2023-07-18 深圳艾斯特创新科技有限公司 Monitoring method, device, equipment and storage medium for high-performance battery

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107238800A (en) * 2017-06-12 2017-10-10 北京理工大学 A kind of remaining probable life Forecasting Methodology of the electrokinetic cell system based on Method Using Relevance Vector Machine and particle filter
WO2018162023A2 (en) * 2017-03-06 2018-09-13 Volvo Truck Corporation A battery state of power estimation method and a battery state monitoring system
CN109633474A (en) * 2018-11-14 2019-04-16 江苏大学 A kind of lithium ion battery residual life prediction technique
CN109993270A (en) * 2019-03-27 2019-07-09 东北大学 Lithium ion battery residual life prediction technique based on grey wolf pack optimization LSTM network

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150349385A1 (en) * 2014-04-01 2015-12-03 Medtronic, Inc. Method and System for Predicting Useful Life of a Rechargeable Battery

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018162023A2 (en) * 2017-03-06 2018-09-13 Volvo Truck Corporation A battery state of power estimation method and a battery state monitoring system
CN107238800A (en) * 2017-06-12 2017-10-10 北京理工大学 A kind of remaining probable life Forecasting Methodology of the electrokinetic cell system based on Method Using Relevance Vector Machine and particle filter
CN109633474A (en) * 2018-11-14 2019-04-16 江苏大学 A kind of lithium ion battery residual life prediction technique
CN109993270A (en) * 2019-03-27 2019-07-09 东北大学 Lithium ion battery residual life prediction technique based on grey wolf pack optimization LSTM network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"基于LSTM循环神经网络的电池SOC预测方法";耿攀等;《上海海事大学学报》;20190930;第120-126页 *

Also Published As

Publication number Publication date
CN111103544A (en) 2020-05-05

Similar Documents

Publication Publication Date Title
CN111103544B (en) Lithium ion battery remaining service life prediction method based on long-time and short-time memory LSTM and particle filter PF
CN111090047B (en) Lithium battery health state estimation method based on multi-model fusion
CN109061504A (en) Same type difference lithium ion battery remaining life prediction technique and system
CN113052464B (en) Method and system for evaluating reliability of battery energy storage system
CN112986831A (en) Lithium ion battery life prediction method based on correlation coefficient particle filtering
CN112684363A (en) Lithium ion battery health state estimation method based on discharge process
CN113219357B (en) Battery pack health state calculation method and system and electronic equipment
CN113917334B (en) Battery health state estimation method based on evolution LSTM self-encoder
CN103983920B (en) A kind of method of the model of the electrokinetic cell setting up electric vehicle
CN114290960B (en) Method and device for acquiring battery health of power battery and vehicle
CN112989690B (en) Multi-time-scale state-of-charge estimation method for lithium battery of hybrid electric vehicle
CN116381541B (en) Health assessment method and system for energy storage lithium battery system
CN104569844A (en) Valve control seal type lead-acid storage battery health condition monitoring method
CN110703120A (en) Lithium ion battery service life prediction method based on particle filtering and long-and-short time memory network
CN114167284B (en) Lithium battery RUL prediction method and equipment based on BMS big data and integrated learning
CN111142025A (en) Battery SOC estimation method and device, storage medium and electric vehicle
CN113657360A (en) Lithium battery health state estimation method, device, equipment and readable storage medium
CN116609676B (en) Method and system for monitoring state of hybrid energy storage battery based on big data processing
CN114114038A (en) Lithium battery SOC and available capacity joint estimation method under full-life and full-temperature conditions
CN114839538A (en) Method for extracting degradation characteristics of lithium ion battery for estimating residual life
CN110503301B (en) Cox proportional risk model-based electric vehicle conduction charging safety influence factor evaluation method
CN117825970A (en) Battery degradation analysis method, device, equipment and storage medium
CN116679214A (en) Lithium ion battery SOH evaluation method, system, equipment and medium integrating data driving and degradation models
CN116819364A (en) Battery health determination method, application method, electronic device and storage medium
CN118311459A (en) Battery life prediction method, device, storage medium and equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant