CN115542168A - Lithium battery residual service life prediction method based on fusion data driving model - Google Patents
Lithium battery residual service life prediction method based on fusion data driving model Download PDFInfo
- Publication number
- CN115542168A CN115542168A CN202211149963.9A CN202211149963A CN115542168A CN 115542168 A CN115542168 A CN 115542168A CN 202211149963 A CN202211149963 A CN 202211149963A CN 115542168 A CN115542168 A CN 115542168A
- Authority
- CN
- China
- Prior art keywords
- lithium battery
- constant
- model
- battery capacity
- curve
- 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
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]
- G01R31/385—Arrangements for measuring battery or accumulator variables
-
- 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]
- G01R31/385—Arrangements for measuring battery or accumulator variables
- G01R31/387—Determining ampere-hour charge capacity or SoC
-
- 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]
- G01R31/385—Arrangements for measuring battery or accumulator variables
- G01R31/387—Determining ampere-hour charge capacity or SoC
- G01R31/388—Determining ampere-hour charge capacity or SoC involving voltage measurements
-
- 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
- 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
- G06N3/08—Learning methods
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for predicting the remaining service life of a lithium battery based on a fusion data driving model, which comprises the following steps: extracting the constant-current charging time, the constant-voltage charging time, the vertical slope at the corner of the constant-current charging curve and the vertical slope at the corner of the constant-current discharging curve as health characteristics; training a chaotic sparrow-extreme learning machine model and a least square support vector regression model by taking the 4 health characteristics as input and corresponding battery capacity as output; and respectively obtaining the predicted values of the battery capacity by using the two trained models, performing weighted fusion to obtain a final predicted value of the battery capacity of the lithium battery, and finally obtaining the remaining service life of the lithium battery by combining a lithium battery capacity curve. The invention adopts the CSSA-ELM-LSSVR fusion algorithm, can fully utilize the CSSA-ELM to extract the whole trend of the lithium battery degradation process, and utilizes the LSSVR to obtain local nonlinear characteristics, thereby realizing accurate prediction of the residual service life of the lithium battery and having better robustness.
Description
Technical Field
The invention belongs to the technical field of batteries, and particularly relates to a method for predicting the residual service life of a lithium battery based on a fusion data driving model.
Background
In recent years, lithium ion batteries having the outstanding advantages of high output voltage, high energy density, good cycle performance, small self-discharge, no memory effect, environmental friendliness and the like have a role of core energy storage components in various fields such as new energy automobiles, smart power grids, aerospace aircrafts, medical treatment, communication, consumer electronics and the like.
However, the lithium ion battery has complex charge and discharge reactions and a harsh practical use environment, and along with the use of the battery, the problems of gradual aging and fading of the performance inside the battery and the like occur under different working conditions (such as temperature, voltage, different charge and discharge currents and the like). If the service life degradation process of the lithium ion battery is neglected, the battery is overloaded and used for a long time, functional failure or damage of electric equipment can be caused if the battery is light, huge economic loss is caused if the battery is heavy, and even serious safety accidents can be caused.
The safety and reliability of lithium ion batteries are bottleneck problems that restrict their rapid development. The efficient management of the lithium ion battery is one of important ways for guaranteeing the safe operation of the electric equipment, and is also a research hotspot problem in the current energy field. Therefore, a prediction method of remaining useful life of the lithium battery RUL is urgently needed.
Disclosure of Invention
The invention aims to provide a lithium battery remaining service life prediction method based on a fusion data driving model, and the lithium battery remaining service life prediction method can be accurately realized.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a method for predicting the remaining service life of a lithium battery based on a fusion data driven model adopts a CSSA-ELM and LSSVR fusion algorithm based on health feature extraction to realize accurate prediction of the remaining service life of the lithium battery, and specifically comprises the following steps:
and acquiring the capacity data of the lithium battery, and extracting characteristics according to a constant current-constant voltage (CC-CV) charging voltage curve and a Constant Current (CC) discharging voltage curve. The method specifically comprises the following steps: acquiring battery capacity data and drawing a capacity curve, a CC-CV charging voltage curve and a CC discharging voltage curve according to the NASA battery data set and the MIT battery data set, wherein the lithium battery capacity curve is a battery capacity-cycle period curve; extracting a constant-current charging time length HI1, a constant-voltage charging time length HI2 and a vertical slope HI3 at a corner of the constant-current charging curve according to a CC-CV charging voltage curve to be used as health indexes; and extracting a vertical slope HI4 at the corner of the constant current discharge curve as a health index according to the CC discharge voltage curve.
And respectively performing off-line training on the battery degradation process based on a chaotic sparrow algorithm-extreme learning machine (CSSA-ELM) and a Least Squares Support Vector Regression (LSSVR) algorithm. The method specifically comprises the following steps: for the CSSA-ELM algorithm, CSSA is used for optimizing the initial weight omega and the bias b of the ELM, a fitness function is designed according to the Mean Square Error (MSE) of a training set, and the ELM is constructed and trained by using the obtained optimal parameters; for the LSSVR algorithm, the LSSVR model is trained using the health indicators and the capacity sequences as inputs and outputs.
And predicting the residual service life of the lithium battery based on a CSSA-ELM-LSSVR fusion algorithm. The method comprises the following specific steps: and (3) using a data fusion method, combining the respective advantages of the ELM and the LSSVR, reasonably distributing the weight of the predicted values of the ELM and the LSSVR, and predicting the residual service life of the lithium battery based on a CSSA-ELM-LSSVR fusion algorithm.
To verify the validity of the prediction model, the RUL prediction was subjected to error analysis. The RUL prediction error, the root mean square error and the decision coefficient may be used to evaluate the model performance.
Further, the method also comprises the following steps:
after the characteristics of the CC-CV charging voltage curve and the CC discharging voltage curve are extracted, the correlation between the health indicator HIs and the capacity is quantitatively analyzed using the Pearson Correlation Coefficient (PCC).
Compared with the prior art, the invention has the following advantages and beneficial effects:
according to the lithium battery remaining service life prediction method based on the fusion data driving model, a CSSA-ELM-LSSVR fusion algorithm is adopted, the CSSA-ELM can be fully utilized to extract the overall trend of the lithium battery degradation process, and the LSSVR is utilized to obtain local nonlinear characteristics, so that the lithium battery remaining service life can be accurately predicted, and meanwhile, the lithium battery remaining service life prediction method has good robustness.
In addition, the method extracts four health indexes of the constant-current charging time, the constant-voltage charging time, the vertical slope at the corner of the constant-current charging curve and the vertical slope at the corner of the constant-current discharging curve, and can predict the residual service life of the lithium battery more accurately.
Drawings
Fig. 1 is a frame diagram of a method for predicting remaining service life of a lithium ion battery according to an embodiment of the present invention;
FIG. 2 is a graph of the charging voltage over different cycle periods for a battery pack of NASA data set B0005 used in an embodiment of the invention;
FIG. 3 is a graph of the charging voltage of the MIT14 battery cell at different cycle periods for the MIT data set used in an embodiment of the present invention;
FIG. 4 is a graph of the discharge voltage over different cycle periods for a battery pack of NASA data set B0005 used in an embodiment of the invention;
fig. 5 is a discharge voltage curve of an MIT 14-set battery at different cycle periods of an MIT data set used by an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The method for predicting the remaining service life of the lithium battery based on the fusion data driving model, disclosed by the embodiment of the invention, as shown in figure 1, comprises the following steps:
step 1: acquiring capacity data of the lithium battery, and performing characteristic extraction according to a CC-CV charging voltage curve and a CC discharging voltage curve;
step 2: respectively performing off-line training on the battery degradation process based on CSSA-ELM and LSSVR algorithms to obtain an optimal training model;
step 3: predicting the residual service life of the lithium battery based on a CSSA-ELM-LSSVR fusion algorithm;
step 4: to verify the validity of the prediction model, the RUL prediction was subjected to error analysis.
Specifically, in Step 1, the obtained NASA data set and MIT data set are first screened, and accordingly, three groups of batteries, namely B0005, B0006 and B0018, in the NASA data set and CH14 batteries of the lot "2018-02-20" in the MIT data set are randomly selected as experimental data sets. Four health indexes of constant-current charging time, constant-voltage charging time, vertical slope at a corner of a constant-current charging curve and vertical slope at a corner of a constant-current discharging curve are extracted, 60% of data sequences are randomly selected to serve as a training set, and the rest are test sets.
In Step 2, the CSSA-ELM algorithm and the LSSVR algorithm are respectively adopted in the embodiments of the present invention to perform training of a single model. It can be understood that ELM is a learning algorithm of a single hidden layer feedforward neural network, has simple model and strong learning ability, and is widely applied in the field of health state prediction; the CSSA is a novel colony intelligent optimization algorithm inspired by sparrow foraging and anti-predation behaviors, can avoid being trapped in local optima, and has the characteristics of high search precision, high convergence rate, good stability, strong robustness and the like. Since the selection of weights and offsets in the ELM is randomly generated, embodiments of the present invention use CSSA to optimize the initial weights and offsets of the ELM. The LSSVR is widely applied to the RUL prediction as a classic kernel-based learning method with the advantages of low calculation cost and strong generalization performance.
In Step 3, from the perspective of learning efficiency, the ELM has the advantages of simple structure, fast learning speed and strong generalization ability, and can fully excavate the overall degradation trend of the degradation process. Meanwhile, the LSSVR algorithm has the advantages of strong nonlinear mapping and strong small sample learning capability, and can extract the local nonlinear characteristics of the degradation process. However, the single data driving prediction method generally has the problem of poor robustness. Therefore, the method for using the fusion data driving model in the embodiment of the invention combines the respective advantages of the two algorithms in Step 2, reasonably distributes the weight of the predicted values of the ELM and the LSSVR, provides a new CSSA-ELM-LSSVR fusion algorithm, and predicts the residual service life of the lithium battery.
Finally, in Step 4, in order to verify the validity of the model, the RUL prediction error E is adopted rul Root mean square error RMSE and coefficient of determination R 2 The model performance was evaluated.
The method for predicting the remaining service life of the lithium ion battery based on the CSSA-ELM and LSSVR fusion algorithm extracted by the health characteristics provided by the embodiment of the invention adopts the CSSA-ELM-LSSVR fusion algorithm, can realize accurate prediction of the remaining service life, and has better robustness.
On the basis of the above embodiment, the feature extraction is performed according to the CC-CV charging voltage curve and the CC discharging voltage curve, and includes:
acquiring battery capacity data and drawing a capacity curve, a CC-CV charging voltage curve and a CC discharging voltage curve according to the NASA battery data set and the MIT battery data set;
extracting a constant-current charging time HI1, a constant-voltage charging time HI2 and a vertical slope HI3 at a corner of the constant-current charging curve according to a CC-CV charging voltage curve to be used as health indexes;
and extracting a vertical slope HI4 at the corner of the constant current discharge curve as a health index according to the CC discharge voltage curve.
Specifically, in the simulation experiment process, three groups of battery data (random 60% data is used as a training set, and the rest is used as a test set) of B0005, B0006 and B0018 in NASA and battery data (random 60% data is used as a training set, and the rest is used as a test set) of CH14 in MIT are selected for carrying out the simulation experiment respectively. Taking the B0005 battery and the MIT14 battery as examples, the CC-CV charging voltage curves and the CC discharging voltage curves of the 40 th, 80 th, 120 th, 160 th different cycle periods and the 100 th, 200 th, 300 th, 400 th different cycle periods are extracted, respectively, as shown in fig. 2 to 5.
As can be seen from fig. 2 and 3, as the number of cycles increases, the charging voltage curve moves to the left, the charging time in the CC mode gradually decreases, the charging duration in the CV mode increases, and the vertical slope at the corner of the curve in the CC charging mode gradually increases, so that the constant-current charging duration HI1, the constant-voltage charging duration HI2, and the vertical slope HI3 at the corner of the constant-current charging curve are selected as the health indicators in the embodiment of the present invention;
as can be seen from fig. 4 and 5, as the number of cycles increases, the voltage of the battery decreases faster and faster, the slope of the discharge curve tends to be stable at the end of the CC discharge mode, and as the battery ages, the vertical slope of the corner of the CC discharge curve decreases, so that the vertical slope HI4 at the corner of the constant-current discharge curve is selected as the health indicator in the embodiment of the present invention.
Specifically, in the simulation experiment process, 60% of data sequences are randomly selected as a training set, and the rest are test sets. Setting a prediction starting point T, and then converting the original data interval into [0,1] by using a min-max normalization method, wherein the normalization formula is as follows:
wherein, y i And y norm Representing raw data and normalized data, respectively. y is max And y min The maximum and minimum values in the raw data.
On the basis of the above embodiment, the method further includes:
after the extraction of the characteristics of the CC-CV charge curve and the CC discharge curve, the correlation between the health indicator HIs and the capacity was quantitatively analyzed using the Pearson Correlation Coefficient (PCC). The calculation formula of PCC is as follows:
wherein x i 、
Respectively the extracted health index sequence and the mean value, Q, of the sequence i 、
The extracted capacity sequence and the mean value of the capacity sequence are respectively shown, n represents the dimensionality of the extracted sequence, the numerator of PCC is the covariance of the health index and the capacity, and the denominator is the standard deviation of the health index and the standard deviation of the capacity. The PCC ranges from-1 to 1, wherein the closer the PCC is to 1, the better the correlation between HIs and capacity is, and 0 is irrelevant.
On the basis of the embodiment, the battery degradation process is respectively trained off line based on CSSA-ELM and LSSVR algorithms, and the method comprises the following steps:
for the CSSA-ELM algorithm, CSSA is used for optimizing the initial weight omega and the bias b of the ELM, a fitness function is designed according to the Mean Square Error (MSE) of a training set, and the ELM is constructed and trained by using the obtained optimal parameters to obtain an optimal CSSA-ELM model;
and for the LSSVR algorithm, the health index and the capacity sequence are used as input and output, and an LSSVR model is trained to obtain the optimal LSSVR model.
From the content of the above embodiment, the method provided by the embodiment of the present invention adopts the CSSA-ELM algorithm and the LSSVR algorithm to perform single model prediction, respectively.
1、CSSA-ELM
ELM is a learning algorithm of a single hidden layer feedforward neural network, and comprises three layers: an input layer, a hidden layer, and an output layer. Suppose there are N samples (x) i ，t i ) Wherein x is i ＝[x i1 ，x i2 ，…，x in ] T ∈R n ，t i ＝[t i1 ，t i2 ，…，t im ] T ∈R mm Representing a data set as
The ELM algorithm is expressed as:
wherein l is the number of neurons in the hidden layer, n is the number of training samples, x i ＝[x i1 ，x i2 ，…，x in ] T ∈R n For input, w i ＝[w i1 ，w i2 ，…，w it ] T Representing the weight of the connection of the input layer to the hidden layer, beta i ＝[β i1 ，β i2 ，…，β im ] T Representing the connection weight of the hidden layer to the output layer, b i For the bias of the ith hidden layer unit, g (-) represents the hidden layer's activation function.
Equation (3) is abbreviated as:
T＝Gβ (4)
wherein
Randomly generating weights w i And bias b i And determining the output weight:
wherein G + ＝(G T G) -1 G T And represents the generalized inverse of matrix G.
Because the weights and the offsets in the ELM are generated randomly, the chaotic sparrow algorithm (CSSA) is added in the embodiment of the invention to optimize the weights and the offsets of the ELM. Firstly, a chaotic map is used for initializing a population position, and then, in order to avoid reduction of population diversity in an iteration process, a Gaussian mutation operator is adopted to enhance local search capability. The CSSA procedure is as follows:
1) Tent mapping:
the iteration of Tent mapping is described as:
wherein z (k) belongs to (0, 1) is a system value of the kth iteration, and mu belongs to (0, 2) is a chaotic control parameter of the Tent mapping function.
According to the Tent mapping, the initialization of the CSSA is realized by the following means:
first, the ith sparrow z of the jth dimension is divided into ij (i =1,2, \8230;, N; j =1,2, \8230;, D) is normalized to z by equation (6) ij (0)：
Wherein lb and ub are each z ij Lower and upper bounds.
Then, using Tent mapping to get z ij (k + 1) is as follows:
z is represented by formula (9) ij (k + 1) from the original domain [0,1]]Transformation to a new z ij ：
z ij ＝lb+z ij (k+1)(ub-lb) (9)
2) Gaussian variation:
in order to avoid trapping in local optimum and keep the diversity of sparrow populations in the iterative process, a Gaussian mutation operator is introduced to enhance the local search capability, and the Gaussian mutation formula is as follows:
mutation(z)＝z(1+rand(1，D)) (10)
on the basis of the above, in the embodiment of the invention, the CSSA is added to optimize the initial weight ω and the bias b in the ELM algorithm during single-model prediction. The fitness function is designed according to the Mean Square Error (MSE) of a training set:
fitness＝arg min(MSE train_set ) (11)
wherein
y i And
respectively representing the actual value and the predicted value of the sample i, N representing the total number of samples, MSE train_set The smaller the correlation between the output of the ELM and the original capacity.
Specifically, the CSSA-ELM algorithm comprises the following implementation steps:
(1) And (5) initializing. Initializing population size N and finding number P N Number of scouts R N The dimension D of the target function, the upper bound ub and the lower bound lb of the initial omega and b, the maximum iteration number T and the solving precision epsilon.
(2) Tent mapping and translation. Tent mapping and transforming the initial value x using equations (7) - (9) i ＝(x i，ω ，x i，b )(i＝1，2，…，N)。
(3) The ELM model is trained. According to equations (3) - (5), the ELM model is trained using ω and b of equation (9) as initialization parameters.
(4) Calculating the fitness of each sparrow by using the formula (11), and selecting the current optimal fitness f best And its corresponding position
And the current worst fitness f worst And its corresponding position
Then, the positions of the seeker and the follower are updated according to equations (12) and (13).
Wherein
Showing the position of the ith sparrow in the jth dimension at the current iteration time t.
Is the optimal location of the current finder,
is the worst position of the population. And the alpha epsilon (0, 1) is a random number, and Q is a random number which obeys standard normal distribution. L is a 1 × D matrix, where each element is 1.A is a 1 × D matrix in which each element is randomly assigned a value of 1 or-1 + ＝A T (AA T ) -1 。R∈[0，1]And ST ∈ [0.5,1 ]]Respectively an early warning value and a safety threshold value. When R is less than ST, the finder can search widely to guide the population to obtain higher fitness without predators or other dangers; when R is larger than or equal to ST, sparrows are detected to find predators, the population immediately makes anti-predation behaviors, and all sparrows rapidly migrate to other safe areas.
(5) Randomly selecting a scout R according to 10-20% of the sparrows N The scout position is updated by equation (14).
Wherein
Is the global optimal position of the sparrow population, beta is a step length control parameter which follows standard normal distribution, and K is the ∈ [ -1,1]Is a random number. e is a small constant, avoiding a denominator of 0. When in use
Meanwhile, the sparrows are detected to be at the edge of the colony and are easy to be attacked by predators; when in use
Meanwhile, the detected sparrows are in the center of the population and need to be close to other sparrows for anti-predation.
(6) The mutation was performed by using the formula (10).
(7) When the maximum iteration N or the solving precision epsilon is met, outputting the optimal parameters
Otherwise, repeating the step 2 to the step 7 until the termination condition is met.
(8) And (5) constructing a training ELM by using the optimal parameters in the step 7, and then outputting the result.
2、LSSVR
The embodiment of the invention trains the LSSVR model by using the health index and the capacity sequence as input and output, and can output the capacity estimation value corresponding to the new period when the health index of the new period is extracted as the input of the LSSVR model. The LSSVR regression model is expressed as:
where N represents the total number of samples, γ i Represents the weight vector, K (x) i X) is a kernel function and b is a bias term.
On the basis of the embodiment, the method for predicting the residual service life of the lithium battery based on the CSSA-ELM-LSSVR fusion algorithm comprises the following steps:
and (3) using a data fusion method, combining the respective advantages of the ELM and the LSSVR, reasonably distributing the weight of the predicted values of the ELM and the LSSVR, and predicting the residual service life of the lithium battery based on a CSSA-ELM-LSSVR fusion algorithm.
From the above description, it can be seen that the embodiments of the present invention combine the advantages of ELM and LSSVR, and provide a new CSSA-ELM-LSSVR algorithm to implement RUL prediction. And weighting the prediction of the CSSA-ELM and LSSVR models to obtain a prediction result given by the fusion model. The fusion model is represented as:
wherein
And
predicted results, ω, for CSSA-ELM and LSSVR, respectively 1 And ω 2 Is a corresponding weight and satisfies omega 1 +ω 2 ＝1。
It will be appreciated that the prediction accuracy of the individual different models is different, the greater the variance of the prediction error, the lower the prediction accuracy of the individual model, and the lower its importance in the fused model. The above weight calculation formula is as follows:
wherein
And
covariance of prediction error respectively
And
it can be seen that the weight ω of the CSSA-ELM model 1 Prediction error covariance with LSSVR model
In direct proportion and vice versa.
On the basis of the above embodiment, the error analysis of the prediction result includes:
error analysis is carried out on the prediction result, and in order to verify the effectiveness of the model, the RUL is adopted to predict the error E rul Root mean square error RMSE and coefficient of determination R 2 The model performance was evaluated.
From the content of the above embodiments, in order to verify the validity of the prediction model, the embodiments of the present invention perform error analysis on the prediction results, and respectively use the RUL prediction error E rul Root mean square error RMSE and coefficient of determination R 2 And evaluating the performance of the model, wherein an evaluation index formula is as follows:
E rul ＝|RUL T -RUL P | (19)
wherein RUL T To predict the RUL of the starting point T, RUL P RUL predicted for the fusion model. Q k In order to be of a practical capacity,
in order to predict the capacity value,
is the average value of the capacity. And N is the number of test sample cycles. E rul The smaller the value of RMSE, R 2 The larger the value of (b), the better the RUL prediction performance.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions described in the foregoing embodiments or substitute equivalent technical features, and the modifications or substitutions do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be covered in the scope of the claims of the present invention.
Claims (4)
1.A method for predicting the remaining service life of a lithium battery based on a fusion data driving model is characterized by comprising the following steps:
acquiring capacity data of the lithium battery, and drawing a capacity curve of the lithium battery, and a constant current-constant voltage charging voltage curve and a constant current discharging voltage curve of the lithium battery in different cycle periods; wherein, the lithium battery capacity curve is a battery capacity-cycle period curve;
extracting the constant-current charging time, the constant-voltage charging time and the vertical slope at the corner of the constant-current charging curve according to a constant-current-constant-voltage charging voltage curve to be used as health characteristics; extracting a vertical slope at a corner of a constant current discharge curve according to the constant current discharge voltage curve as a health characteristic;
respectively training the lithium battery degradation process on a training set based on a chaotic sparrow-extreme learning machine and a least square support vector regression model by taking the 4 health characteristics as input and corresponding battery capacity as output so as to obtain a trained chaotic sparrow-extreme learning machine model and a trained least square support vector regression model;
and respectively obtaining battery capacity predicted values by using the two trained models, weighting the two battery capacity predicted values to obtain a final lithium battery capacity predicted value, and finally obtaining the remaining service life of the lithium battery by combining a lithium battery capacity curve.
2. The method for predicting the remaining service life of the lithium battery based on the fusion data driven model as claimed in claim 1, wherein after the health features are extracted, a pearson correlation coefficient is adopted to perform quantitative analysis on the correlation between the health features and the battery capacity, and the health features with high correlation are selected as model training input.
3. The method for predicting the remaining service life of the lithium battery based on the fusion data driving model according to claim 1, wherein the training of the lithium battery degradation process based on the chaotic sparrow-extreme learning machine and the least square support vector regression model is respectively as follows:
for the chaotic sparrow-extreme learning machine, optimizing the initial weight omega and the bias b of the extreme learning machine by using a chaotic sparrow algorithm, designing a fitness function according to a mean square error, and constructing and training the extreme learning machine by using the obtained optimal parameters to obtain an optimal chaotic sparrow-extreme learning machine model reflecting the degradation process of the lithium battery;
for the least squares support vector regression model, the health index and the battery capacity sequence are used as input and output, and the least squares support vector regression model is trained to obtain the optimal least squares support vector regression model.
4. The method for predicting the remaining service life of the lithium battery based on the fusion data driving model as claimed in claim 1, wherein the trained two models are used to respectively obtain predicted values of battery capacity, and the two predicted values of battery capacity are weighted to obtain a final predicted value f of battery capacity of the lithium battery, wherein the predicted value f of battery capacity of the lithium battery is:
in the formula (I), the compound is shown in the specification,
and
predicted values of battery capacity, omega, of lithium battery of chaotic sparrow-extreme learning machine model and least square support vector regression model respectively 1 And ω 2 For the corresponding weight, the calculation method is as follows:
in the formula (I), the compound is shown in the specification,
and
respectively are the covariance of the prediction errors of the chaos sparrow-extreme learning machine and the least square support vector regression model on the training set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211149963.9A CN115542168A (en) | 2022-09-21 | 2022-09-21 | Lithium battery residual service life prediction method based on fusion data driving model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211149963.9A CN115542168A (en) | 2022-09-21 | 2022-09-21 | Lithium battery residual service life prediction method based on fusion data driving model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115542168A true CN115542168A (en) | 2022-12-30 |
Family
ID=84727404
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211149963.9A Pending CN115542168A (en) | 2022-09-21 | 2022-09-21 | Lithium battery residual service life prediction method based on fusion data driving model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115542168A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116432531A (en) * | 2023-04-17 | 2023-07-14 | 北方工业大学 | Bearing residual service life prediction method based on improved nuclear extreme learning machine |
| CN118050640A (en) * | 2023-11-30 | 2024-05-17 | 湖北工业大学 | Area characteristic and data driving-based lithium battery RUL prediction method and system |
-
2022
- 2022-09-21 CN CN202211149963.9A patent/CN115542168A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116432531A (en) * | 2023-04-17 | 2023-07-14 | 北方工业大学 | Bearing residual service life prediction method based on improved nuclear extreme learning machine |
| CN118050640A (en) * | 2023-11-30 | 2024-05-17 | 湖北工业大学 | Area characteristic and data driving-based lithium battery RUL prediction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Khalid et al. | Prediction of li-ion battery state of charge using multilayer perceptron and long short-term memory models | |
| CN110888058B (en) | Algorithm based on power battery SOC and SOH joint estimation | |
| WO2022253038A1 (en) | Method and system for predicting state of health of lithium battery on basis of elastic network, and device and medium | |
| CN111366848A (en) | Battery health state prediction method based on PSO-ELM algorithm | |
| CN113253116A (en) | Lithium ion battery state of charge estimation method and storage medium | |
| CN115542168A (en) | Lithium battery residual service life prediction method based on fusion data driving model | |
| CN111948560A (en) | Lithium battery health state estimation method based on multi-factor evaluation model | |
| CN107132490B (en) | Method for estimating state of charge of lithium battery pack | |
| CN115632179B (en) | Intelligent quick charging method and system for lithium ion battery | |
| CN110007235A (en) | A kind of accumulator of electric car SOC on-line prediction method | |
| CN110007240A (en) | A kind of lithium ion battery residual life prediction technique | |
| CN113466710B (en) | SOC and SOH collaborative estimation method for energy storage battery in receiving-end power grid containing new energy | |
| CN116449218B (en) | Lithium battery health state estimation method | |
| CN116794547A (en) | Lithium ion battery residual service life prediction method based on AFSA-GRU | |
| Li et al. | A hybrid framework for predicting the remaining useful life of battery using Gaussian process regression | |
| CN116774088A (en) | Lithium ion battery health state estimation method based on multi-objective optimization | |
| CN113901707A (en) | SOH time sequence prediction method for vehicle lithium ion power battery | |
| CN113884936B (en) | ISSA coupling DELM-based lithium ion battery health state prediction method | |
| CN115236526A (en) | Method and device for predicting residual charging time, storage medium and vehicle | |
| CN116027198A (en) | Lithium battery health state estimation method based on combined weighted domain countermeasure network | |
| CN113671378B (en) | Fractional order theory-based lithium ion battery modeling and parameter identification method | |
| CN117289167A (en) | Battery remaining life prediction method, device and medium based on multiple neural network | |
| CN117169743A (en) | Battery health state estimation method and device based on partial data and model fusion | |
| CN116774045A (en) | Lithium battery health state prediction method based on HHO-SVR | |
| CN116796624A (en) | Self-adaptive incremental battery RUL prediction method combining multi-source degradation characteristics |
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 |