CN116609678A - Lithium ion battery residual life prediction method based on improved mixed model - Google Patents
Lithium ion battery residual life prediction method based on improved mixed model Download PDFInfo
- Publication number
- CN116609678A CN116609678A CN202310424744.5A CN202310424744A CN116609678A CN 116609678 A CN116609678 A CN 116609678A CN 202310424744 A CN202310424744 A CN 202310424744A CN 116609678 A CN116609678 A CN 116609678A
- Authority
- CN
- China
- Prior art keywords
- lithium ion
- neural network
- ion battery
- voltage
- current
- 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
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 title claims abstract description 53
- 229910001416 lithium ion Inorganic materials 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 46
- 241000287127 Passeridae Species 0.000 claims abstract description 43
- 230000015654 memory Effects 0.000 claims abstract description 27
- 230000001537 neural effect Effects 0.000 claims abstract description 20
- 238000010845 search algorithm Methods 0.000 claims abstract description 20
- 230000008569 process Effects 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 12
- 238000010277 constant-current charging Methods 0.000 claims abstract description 10
- 238000007600 charging Methods 0.000 claims abstract description 9
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 230000007547 defect Effects 0.000 claims abstract description 4
- 230000002708 enhancing effect Effects 0.000 claims abstract description 4
- 238000013528 artificial neural network Methods 0.000 claims description 24
- 230000008859 change Effects 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 17
- 238000013527 convolutional neural network Methods 0.000 claims description 14
- 239000012634 fragment Substances 0.000 claims description 12
- 230000002431 foraging effect Effects 0.000 claims description 9
- 125000004122 cyclic group Chemical group 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 8
- 238000010280 constant potential charging Methods 0.000 claims description 7
- 230000007787 long-term memory Effects 0.000 claims description 7
- 230000006403 short-term memory Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000012300 Sequence Analysis Methods 0.000 claims description 5
- 230000009286 beneficial effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000010220 Pearson correlation analysis Methods 0.000 claims description 3
- 238000012352 Spearman correlation analysis Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 2
- 238000001914 filtration Methods 0.000 claims description 2
- 230000006872 improvement Effects 0.000 claims description 2
- 238000013507 mapping Methods 0.000 claims description 2
- 238000013178 mathematical model Methods 0.000 claims description 2
- 238000012795 verification Methods 0.000 claims description 2
- 230000002123 temporal effect Effects 0.000 claims 1
- 238000010281 constant-current constant-voltage charging Methods 0.000 abstract 1
- 238000010219 correlation analysis Methods 0.000 abstract 1
- 238000005065 mining Methods 0.000 abstract 1
- 230000007246 mechanism Effects 0.000 description 5
- 230000032683 aging Effects 0.000 description 3
- 230000015556 catabolic process Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000006731 degradation reaction Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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]
- G01R31/378—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC] specially adapted for the type of battery or accumulator
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
Abstract
The invention relates to a lithium ion battery residual life prediction method based on an improved hybrid model. Comprising the following steps: acquiring charging voltage, current and time data of the lithium ion battery in the constant current charging and constant voltage charging processes; extracting the ratio of voltage to time and the ratio of current to time according to the acquired data, and analyzing the correlation degree of the ratio and the battery capacity by adopting a Pierson and Szelman correlation analysis method; constructing a lithium ion battery residual life prediction data set, and dividing the lithium ion battery residual life prediction data set into a training set and a testing set; establishing a convolutional neural network-long-short-term memory network hybrid model, optimizing parameters of the hybrid model by adopting an improved sparrow search algorithm, and achieving the purposes of effectively mining data information and enhancing the residual service life prediction capability; the lithium ion battery residual life prediction method based on the improved hybrid model provided by the invention has higher accuracy and good adaptability to battery residual life prediction, and the defect of unstable output of a single model is overcome.
Description
Technical Field
The invention belongs to the technical field of lithium ion batteries, and particularly relates to a method for predicting the residual life of a lithium ion battery based on an improved hybrid model.
Background
The lithium ion battery is widely applied to the fields of wearable equipment, electric automobiles, power sources, battery energy storage systems and the like because of the advantages of long service life, wide working temperature range, good safety performance and environmental protection. However, lithium ion batteries are subject to continuous aging during recycling, which presents challenges to the safety and economy of battery use. In order to ensure safe and stable operation and efficient operation of the battery system, it is necessary to establish a reliable battery management system. Among them, the prediction of the remaining life of a lithium ion battery is one of the core problems of a battery management system.
In recent years, scholars at home and abroad have achieved many results regarding the research on the prediction of the residual service life of lithium ion batteries. The main prediction methods are divided into two categories: a mechanism model-based approach and a data-driven approach. The method based on the mechanism model relies on the aging mechanism of the lithium ion battery, comprehensively considers the internal structure and the physical and chemical reaction of the battery, and establishes a dynamic model through the operation condition and the failure mechanism of the battery to predict the residual service life. Compared with the method based on the mechanism model, the method based on the data driving does not consider the physical and chemical reaction process of the lithium ion battery, can be regarded as a black box, and directly excavates the information of the degradation trend of the reaction battery from the battery data. In recent years, the rapid development of machine learning algorithms provides new opportunities for predicting the remaining useful life of lithium ion batteries.
The key problem of residual service life prediction is to obtain factors which can represent the battery aging state most from lithium ion battery operation data and construct an accurate prediction model based on the factors. Considering the difficulty of online extraction of capacity loss values, there is an urgent need to find health factors that are highly correlated to capacity and can characterize battery degradation.
The convolutional neural network can enhance the learning ability of the lithium ion battery, effectively dig rules among data, and extract deep characteristic information of the battery in a convolutional form to form a characteristic vector; removing redundant information through a pooling layer to optimize the complexity of the neural network, and inputting the pooling layer result into a long-term and short-term memory network; the long-term and short-term memory network selectively reserves the afferent information of the convolutional neural network, and finally predicts the residual service life by utilizing the abundant time sequence analysis capability.
Disclosure of Invention
The embodiment of the invention aims to provide a lithium ion battery residual life prediction method based on an improved hybrid model so as to accurately predict the residual service life of a lithium ion battery.
In order to solve the technical problems, the invention adopts the technical proposal that,
step 1, carrying out charge and discharge cycles on a lithium ion battery for a plurality of times, and obtaining constant voltage, constant current and time data in each charge and discharge cycle;
and 2, dividing the charging voltage of the obtained constant current charging data into 4 equidistant fragments, determining the starting voltage and the ending voltage of each fragment, and dividing the difference value of the ending voltage and the starting voltage by the time of the two fragments of voltage to obtain the change rate ratio. For the obtained constant voltage charging data, dividing the charging current in the process into 2 equidistant fragments, determining the initial current and the end current of each fragment, and dividing the difference value between the end current and the initial current by the time of the two fragments of current to obtain the change rate ratio;
and step 3, carrying out pearson and spearman correlation analysis on the calculated change rate ratio of the voltage segment and the current segment and the battery capacity data. The method is used as a residual service life data set of the lithium ion battery, and is preprocessed, and the processed data set is divided into a training set and a verification set;
step 4, building a convolutional neural network-long-short-term memory neural network mixed model, and introducing an improved sparrow search algorithm to optimize the convolutional neural network-long-term memory neural network mixed model;
and step 5, training the optimized convolutional neural network-long-short-term memory cyclic neural network hybrid model by using a training set, and evaluating the estimation effect of the model by using a testing set to test the prediction precision of the model.
In one embodiment of the present invention, the step 1 includes the following sub-steps:
step 1.1, experimental data are derived from a public data set provided by PCoE of the United states of America and aerospace bureau, and constant current charging data of four LG Chem 18650 lithium ion batteries (B5, B6, B7 and B18) are selected. And 4 batteries are subjected to cyclic charge and discharge experiments in an environment of room temperature (24 ℃), and are charged to a cut-off voltage (4.2V) with a constant current of 1.5A in the charging process, and then are continuously charged to a current falling of 20mA by a constant voltage model.
In one embodiment of the present invention, the step 2 includes the following sub-steps:
step 2.1, when extracting the voltage segment data of the constant current charging process, the set 4 voltage segments are [3.8-3.9], [3.9-4.0], [4.0-4.1], [4.1-4.2], and the change rate ratio is defined as
Wherein delta is t1 Time required for starting from the start voltage to the end voltage;
2.2, when extracting the current segment data of the constant voltage charging process, the set 2 current segments are [1.5-1.0], [1.0-0.5], and the change rate ratio is defined as
Wherein delta is t2 The time required to start and end the current from the start current;
in one embodiment of the present invention, the step 3 includes the following sub-steps:
step 3.1, analyzing the relativity between the voltage and current change rate ratio obtained in the step 3 and the battery capacity by using a Pierson and Szelman formula, and setting two variables a= (a) 1 ,a 2 ,…,a 3 ) And b= (b) 1 ，b 2 ，…，b n ). Pearson formula is
Wherein: e is a mathematical expectation. The absolute value of P is approximately 1, indicating a higher correlation of HF with capacity.
The Spearman formula is
Wherein:is the average of a and b.
In one embodiment of the present invention, the step 4 includes the following sub-steps:
in the step 4.1, in the convolutional neural network-long-short-term memory neural network mixed model, the convolutional neural network has the function of enhancing the learning capacity of the lithium ion battery, effectively excavating rules among data, and extracting deep characteristic information of the battery in a convolutional form to form a characteristic vector; removing redundant information through a pooling layer to optimize the complexity of the neural network, inputting the pooling layer result into a long-short-term memory neural network, selectively reserving the afferent information of the convolutional neural network, and finally predicting the residual service life of the lithium ion battery by utilizing the abundant time sequence analysis capability of the convolutional neural network;
in the step 4.2, in the sparrow searching algorithm, the sparrow population is divided into three categories of discoverers, joiners and alerters, and the location updating formula of the discoverers is as follows:
the subscriber location update formula is:
the alerter location update formula is:
and 4.2, introducing a reverse learning strategy and a Levy flight algorithm to improve the sparrow search algorithm. The main idea of generating populations using reverse learning strategies is to generate their reverse populations from randomly generated populations, from which a preferred population is selected as the next generation population. The reverse learning strategy selects the initial individuals of the population as the individuals closer to the optimal solution such that each individual is further from the optimal solution in order to increase the convergence rate of all the individuals of the population.For reversing individual information
Wherein: u is the upper bound of the foraging range; l is the lower bound of the foraging range; k is any constant between [0,1 ].
After the sparrow search algorithm introduces the reverse learning strategy, the improved location update of the discoverer is as follows
The Levy flight strategy can expand the foraging range, is beneficial to increasing population diversity, and can well solve the defect that the traditional SSA algorithm falls into a local optimal solution. The Levy flight strategy step formula s is
After the Levy flight strategy is introduced, the improved location update of the enrollees is as follows
Step 4.3, setting the sparrow population N as 30, setting the maximum iteration number M as 100, setting the producer duty ratio PD as 20%, setting the guard duty ratio SD as 10%, setting the safety threshold ST as 0.8, and setting the parameter dimension dim as 4. The mean square error MSE between the actual value and the predicted value is used as a fitness function.
Wherein:as a predicted value, C i Is a true value.
And sorting sparrows according to the fitness function, and selecting the optimal sparrows. And (3) introducing a reverse learning strategy and a Levy flight algorithm to update the positions of a finder and a joiner, comparing the iterated optimal value with the stored optimal value, and judging whether the iteration termination condition is met. If not, continuously updating the position of the sparrow according to the improved formula, and starting the next optimization; if yes, outputting the optimal network parameters.
In one embodiment of the present invention, the step 5 includes the following sub-steps:
step 5.1, preprocessing training set data, then inputting a convolutional neural network-long-short-term memory cyclic neural network mixed model, and training the model by taking battery capacity as output of the mixed model;
and 5.2, absolute error AE, root mean square error RMSE and average absolute error MAE are introduced to evaluate the model precision.
The invention has the beneficial effects that
1. The strong feature extraction capability of the convolutional neural network and the time sequence analysis capability of the long-term and short-term memory network are effectively combined, and the unstable memory output of a single model is avoided
2. The degradation trend of the battery can be observed at any time by taking the slope of the voltage segment as a health factor, the health condition is not required to be estimated, and complete charge and discharge data are not required to be acquired.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to an embodiment of the present invention;
fig. 2 is a graph showing capacity fade curves of four lithium ion batteries according to an embodiment of the present invention;
fig. 3 is a voltage variation curve during a constant current charging process of different lithium ion batteries according to an embodiment of the present invention;
fig. 4 is a graph showing a current change during a constant voltage charging process of different lithium ion batteries according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a convolutional neural network-long-short-term memory neural network hybrid model according to an embodiment of the present invention;
fig. 6 is a comparison of the residual life prediction method of the lithium ion battery based on the improved hybrid model provided by the embodiment of the invention with other methods.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In one embodiment, as shown in fig. 1, a method for predicting remaining life of a lithium ion battery based on an improved hybrid model includes the steps of:
1. and (3) carrying out multiple charge and discharge cycles on the lithium ion battery, wherein experimental data are derived from a public data set provided by PCoE of the space agency, and constant current charge data of four LG Chem 18650 lithium ion batteries (B5, B6, B7 and B18) are selected. And 4 batteries are subjected to cyclic charge and discharge experiments in an environment of room temperature (24 ℃), and are charged to a cut-off voltage (4.2V) with a constant current of 1.5A in the charging process, and then are continuously charged to a current falling of 20mA by a constant voltage model. The operating parameters are shown in table 1.
Table 1 selected battery operating parameters
2. The capacity fade trend for 4 cells is shown in fig. 2. As shown in FIG. 3, the voltage change condition in the constant current charging process of the battery is shown, the charging voltage interval is divided into four equidistant parts, and when the voltage segment data in the constant current charging process is extracted, the set 4 voltage segments are [3.8-3.9], [3.9-4.0], [4.0-4.1], [4.1-4.2], and the change rate ratio is defined as
Wherein delta is t1 Time required for starting from the start voltage to the end voltage;
as shown in FIG. 4, the current change condition in the constant voltage charging process of the battery is shown, when the current segment data of the constant voltage charging process is extracted, the set 2 current segments are [1.5-1.0], [1.0-0.5], and the change rate ratio is defined as
Wherein delta is t2 The time required to start and end the current from the start current;
3. for the ratio of the voltage to the current change rate obtained in the step 3, the correlation between the ratio and the battery capacity is analyzed by using the Pierson and Szelman formulas, and two variables a= (a) are set 1 ,a 2 ,…,a 3 ) And b= (b) 1 ，b 2 ，…，b n )。PearsoThe formula of n is
Wherein: e is a mathematical expectation. The absolute value of P is approximately 1, indicating a higher correlation of HF with capacity.
The Spearman formula is
Wherein:is the average of a and b.
The rate of change and the capacity of the different voltage-current intervals are shown in table 2.
TABLE 2 correlation of the rate of change with capacity for different voltage (current) intervals
4. In the convolutional neural network-long-short-term memory neural network mixed model, the convolutional neural network has the function of enhancing the learning capacity of the lithium ion battery, effectively excavating rules among data, and extracting deep characteristic information of the battery in a convolutional form to form a characteristic vector; and removing redundant information through a pooling layer to optimize the complexity of the neural network, inputting the pooling layer result into the long-short-term memory neural network, extracting the data characteristics of the lithium ion battery by adopting a one-dimensional convolutional neural network by adopting a mixed model, wherein the mathematical model is as follows
Wherein: f (·) is the activation function;is the firstThe j-th mapping of the i-layer; />A j-th map for the i-1 th layer; m is the number of input features; />Is a trainable convolution kernel; />A bias term. However, convolutional neural networks suffer from the disadvantage that the time-sequential nature of lithium ion battery data cannot be effectively utilized.
The long-term and short-term memory neural network has the function of selectively reserving the afferent information of the convolutional neural network and finally predicting the residual service life of the lithium ion battery by utilizing the abundant time sequence analysis capability; the long-term and short-term memory neural network is updated through a gating structure and is divided into a forgetting gate, an input gate and an output gate. Wherein the forget gate indicates the degree to which the last level unit was forgotten; the input gate and an activation function jointly control the range of the entering information, and the state of the unit is continuously updated; the output gate controls the current unit filtering degree. The calculation process at the time t is as follows
Wherein: f (f) t 、i t 、o t The calculation results of the three gates are obtained; wf, wi, wc are inputs x for the corresponding gate and time t t And intermediate output h t-1 A weight matrix of the product; bf. bi and bc are bias vectors of the corresponding gates; sigma and tanh are activation functions.
5. Introducing an improved sparrow search algorithm to optimize the parameters of the mixed model, so as to obtain a mixed model optimized based on the improved sparrow search algorithm, and further predicting the residual service life of the lithium ion battery;
in the sparrow searching algorithm, the sparrow population is divided into three categories of discoverers, joiners and alerters, and the location updating formula of the discoverers is as follows:
wherein: t is the current iteration number; i.e MAX For maximum number of iterations, a has a value of (0, 1]Between them;the position of the ith sparrow in the j-th dimension; q is a random number and accords with normal distribution; l is a matrix with all elements 1. When R is 2 In < ST, it is stated that the finder can forge in the area; when R is 2 When the price is not less than ST, the natural enemy exists, and the person needs to leave immediately.
The subscriber location update formula is:
wherein:representing the position with the worst population fitness in the t-th iteration; />Representing the optimal position of the population fitness at the t+1st iteration; a and L are matrices of the same dimension, the element value is 1 or-1 and satisfies A + ＝A T (AA T ) -1 . When i is larger than n/2, the adaptability of the joiner is poor, and the joiner needs to fly to other positions; when i is less than or equal to n/2, the adaptability of the user is good, and the user can find food and live.
The alerter location update formula is:
wherein:is a global optimal position; beta is a step control parameter conforming to a normal distribution of (0, 1); k is a constant, and the value range is [ -1,1]M is a constant set to avoid zero denominator; f (f) i And f g Is the global current optimal and worst fitness value. When f i ＞f g When f, the sparrow is located at the position possibly attacked by natural enemy i ＝f g When the sparrow observes natural enemies, the sparrow needs to fly to other sparrows to avoid the natural enemies.
The improved sparrow search algorithm improves the sparrow search algorithm by introducing a reverse learning strategy and a Levy flight algorithm. The main idea of generating populations using reverse learning strategies is to generate their reverse populations from randomly generated populations, from which a preferred population is selected as the next generation population. The reverse learning strategy selects the initial individuals of the population as the individuals closer to the optimal solution such that each individual is further from the optimal solution in order to increase the convergence rate of all the individuals of the population.
For reversing individual information
Wherein: u is the upper bound of the foraging range; l is the lower bound of the foraging range; k is any constant between [0,1 ].
After the sparrow search algorithm introduces the reverse learning strategy, the improved location update of the discoverer is as follows
The Levy flight strategy can expand the foraging range, is beneficial to increasing population diversity, and can well solve the defect that the traditional SSA algorithm falls into a local optimal solution. The Levy flight strategy step formula s is
After the Levy flight strategy is introduced, the improved location update of the enrollees is as follows
5. Setting the sparrow population N as 30, the maximum iteration number M as 100, the producer duty ratio PD as 20%, the alerter duty ratio SD as 10%, the safety threshold ST as 0.8 and the parameter dimension dim as 4. The mean square error MSE between the actual value and the predicted value is used as a fitness function.
Wherein:as a predicted value, C i Is a true value.
And sorting sparrows according to the fitness function, and selecting the optimal sparrows. And (3) introducing a reverse learning strategy and a Levy flight algorithm to update the positions of a finder and a joiner, comparing the iterated optimal value with the stored optimal value, and judging whether the iteration termination condition is met. If not, continuously updating the position of the sparrow according to the improved formula, and starting the next optimization; if yes, outputting the optimal network parameters.
6. The training set data is preprocessed and then used as input of a convolutional neural network-long-short-term memory cyclic neural network mixed model, the battery capacity is used as output of the mixed model, the model is trained, and absolute error AE, root mean square error RMSE and average absolute error MAE are introduced to evaluate model accuracy. Wherein AE is the absolute value of the difference between predicted life and real life, and RMSE, MAE are defined as
7. And comparing the mixed model optimized by the sparrow search algorithm with the convolutional neural network-long-short-term memory network mixed model, the long-short-term memory network model optimized by the sparrow search algorithm and the convolutional neural network-long-term memory network mixed model optimized by the sparrow search algorithm, wherein all experimental batteries are provided with two starting points, and the setting conditions of the starting points are shown in table 3.
TABLE 3 starting point settings for different batteries
The prediction results are shown in fig. 6 by taking the B5 battery starting points as 80 and 100 as examples, and the evaluation indexes are shown in table 4.
TABLE 4 residual Life prediction results for different methods
M1, M2, M3 and M4 are respectively a convolutional neural network-long-short-term memory network mixed model, a long-short-term memory network model optimized by a sparrow search algorithm, a convolutional neural network-long-term memory network mixed model optimized by the sparrow search algorithm and the lithium ion battery residual life prediction method based on the improved mixed model.
In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
The foregoing description is only of the preferred embodiments of the present invention and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.
Claims (7)
1. A lithium ion battery residual life prediction method based on an improved hybrid model is characterized by comprising the following steps of: the method comprises the following steps:
step 1, carrying out charge and discharge cycles on a lithium ion battery for a plurality of times, and obtaining constant voltage, constant current and time data in each charge and discharge cycle;
step 2, for the obtained constant current charging data, dividing the charging voltage in the process into 4 equidistant fragments, determining the starting voltage and the ending voltage of each fragment, and dividing the difference value between the ending voltage and the starting voltage by the time of the two fragments of voltage to obtain a change rate ratio; for the obtained constant voltage charging data, dividing the charging current in the process into 2 equidistant fragments, determining the initial current and the end current of each fragment, and dividing the difference value between the end current and the initial current by the time of the two fragments of current to obtain the change rate ratio;
step 3, carrying out pearson and spearman correlation analysis on the calculated change rate ratio of the voltage segment and the current segment and the battery capacity data, and taking the pearson and spearman correlation analysis as a residual service life data set of the lithium ion battery; preprocessing, namely dividing the processed data set into a training set and a verification set;
step 4, building a convolutional neural network-long-short-term memory neural network mixed model, and introducing an improved sparrow search algorithm to optimize the convolutional neural network-long-term memory neural network mixed model;
step 5, training the optimized convolutional neural network-long-short-term memory cyclic neural network hybrid model by using a training set, and evaluating the estimation effect of the model by using a testing set to test the prediction precision of the model;
and 6, comparing the output result of the mixed model with other methods.
2. The method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to claim 1, wherein the step 1 comprises the following sub-steps:
step 1.1, experimental data are derived from a public data set provided by PCoE of the space agency, and constant current charging data of four LG Chem 18650 lithium ion batteries (B5, B6, B7 and B18) are selected; and 4 batteries are subjected to cyclic charge and discharge experiments in an environment of room temperature (24 ℃), and are charged to a cut-off voltage (4.2V) with a constant current of 1.5A in the charging process, and then are continuously charged to a current falling of 20mA by a constant voltage model.
3. The method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to claim 1, wherein the step 2 comprises the following sub-steps:
step 2.1, when extracting voltage segment data of a constant current charging process, the set 4 voltage segments are [3.8-3.9], [3.9-4.0], [4.0-4.1], [4.1-4.2], and the change rate ratio is defined as:
wherein delta is t1 Time required for starting from the start voltage to the end voltage;
2.2, when extracting current segment data of a constant voltage charging process, setting 2 current segments as [1.5-1.0], [1.0-0.5], wherein the ratio of the change rates is defined as:
wherein delta is t2 The time required to start and end the current from the start current.
4. The method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to claim 1, wherein the step 3 comprises the following sub-steps:
step 3.1,For the ratio of the voltage to the current change rate obtained in the step 3, the correlation between the ratio and the battery capacity is analyzed by using the Pierson and Szelman formulas, and two variables a= (a) are set 1 ,a 2 ,…,a 3 ) And b= (b) 1 ，b 2 ，…，b n ). The Pearson formula is:
wherein: e is a mathematical expectation. The absolute value of P is approximately 1, indicating that the higher the correlation of HF with capacity;
the Spearman formula is:
wherein:is the average of a and b.
5. The method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to claim 1, wherein the step 4 comprises the following sub-steps:
in the step 4.1, in the convolutional neural network-long-short-term memory neural network mixed model, the convolutional neural network has the function of enhancing the learning capacity of the lithium ion battery, effectively excavating rules among data, and extracting deep characteristic information of the battery in a convolutional form to form a characteristic vector; and removing the complexity of the redundant information optimizing neural network through the pooling layer, inputting the pooling layer result into the long-short-term memory neural network, extracting the data characteristics of the lithium ion battery by adopting a one-dimensional convolutional neural network by adopting a mixed model, wherein the mathematical model is as follows:
wherein: f (·) is the activation function;a j-th mapping for the i-th layer; />A j-th map for the i-1 th layer; m is the number of input features; />Is a trainable convolution kernel; />A bias term. However, convolutional neural networks suffer from the disadvantage that the temporal characteristics of lithium ion battery data cannot be effectively utilized;
the long-term and short-term memory neural network has the function of selectively reserving the afferent information of the convolutional neural network and finally predicting the residual service life of the lithium ion battery by utilizing the abundant time sequence analysis capability; the long-term and short-term memory neural network is updated through a gating structure and is divided into a forgetting gate, an input gate and an output gate, wherein the forgetting gate indicates the forgetting degree of a unit at the last level; the input gate and an activation function jointly control the range of the entering information, and the state of the unit is continuously updated; the output door controls the filtering degree of the current unit; the calculation process at the time t is as follows:
wherein: f (f) t 、i t 、o t The calculation results of the three gates are obtained; wf, wi, wc are inputs x for the corresponding gate and time t t And intermediate output h t-1 A weight matrix of the product; bf. bi and bc are bias vectors of the corresponding gates; sigma and tanh are activation functions;
in the step 4.2, in the sparrow searching algorithm, the sparrow population is divided into three categories of discoverers, joiners and alerters, and the location updating formula of the discoverers is as follows:
the subscriber location update formula is:
the alerter location update formula is:
step 4.3, introducing a reverse learning strategy and a Levy flight algorithm to improve a sparrow search algorithm; the main idea of generating the population by using the reverse learning strategy is to generate a reverse population according to the randomly generated population, and select a better population as a next generation population; the reverse learning strategy selects initial individuals of the population as individuals closer to the optimal solution, so that each individual is further away from the optimal solution, and the convergence speed of all the individuals of the population is improved;the individual information is expressed as the reverse:
wherein: u is the upper bound of the foraging range; l is the lower bound of the foraging range; k is any constant between [0,1 ];
after the sparrow search algorithm introduces a reverse learning strategy, the improved position of the discoverer is expressed as:
the Levy flight strategy can expand the foraging range, is beneficial to increasing population diversity, and can well solve the defect that the traditional SSA algorithm falls into a local optimal solution; the Levy flight strategy step formula s is:
after the Levy flight strategy is introduced, the improvement position of the joiner is expressed as:
step 4.4, setting the sparrow population N as 30, setting the maximum iteration number M as 100, setting the producer duty ratio PD as 20%, setting the guard duty ratio SD as 10%, setting the safety threshold ST as 0.8, and setting the parameter dimension dim as 4; using the mean square error MSE between the actual and predicted values as a fitness function, the MSE is expressed as:
wherein:as a predicted value, C i Is a true value;
sorting sparrows according to the fitness function, and selecting the optimal sparrows; the reverse learning strategy and the Levy flight algorithm are introduced to update the positions of the discoverer and the joiner, the iterative optimal value is compared with the previous stored optimal value, and whether the iteration termination condition is met is judged; if not, continuously updating the position of the sparrow according to the improved formula, and starting the next optimization; if yes, outputting the optimal network parameters.
6. The method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to claim 1, wherein the step 5 comprises the following sub-steps:
step 5.1, preprocessing training set data, then inputting a convolutional neural network-long-short-term memory cyclic neural network mixed model, and training the model by taking battery capacity as output of the mixed model;
and 5.2, absolute error AE, root mean square error RMSE and average absolute error MAE are introduced to evaluate the model precision.
7. The method for predicting the remaining life of a lithium ion battery based on an improved hybrid model according to claim 1, wherein said step 6 comprises the sub-steps of:
and 6.1, comparing the mixed model optimized by the sparrow search algorithm with a convolutional neural network-long-short-term memory network mixed model, a long-short-term memory network model optimized by the sparrow search algorithm and a convolutional neural network-long-term memory network mixed model optimized by the sparrow search algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310424744.5A CN116609678A (en) | 2023-04-19 | 2023-04-19 | Lithium ion battery residual life prediction method based on improved mixed model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310424744.5A CN116609678A (en) | 2023-04-19 | 2023-04-19 | Lithium ion battery residual life prediction method based on improved mixed model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116609678A true CN116609678A (en) | 2023-08-18 |
Family
ID=87673700
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310424744.5A Pending CN116609678A (en) | 2023-04-19 | 2023-04-19 | Lithium ion battery residual life prediction method based on improved mixed model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116609678A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117236200A (en) * | 2023-11-16 | 2023-12-15 | 北京航空航天大学 | Method for optimizing quick charge strategy of aerocar battery based on data-driven reduced-order model |
-
2023
- 2023-04-19 CN CN202310424744.5A patent/CN116609678A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117236200A (en) * | 2023-11-16 | 2023-12-15 | 北京航空航天大学 | Method for optimizing quick charge strategy of aerocar battery based on data-driven reduced-order model |
| CN117236200B (en) * | 2023-11-16 | 2024-02-02 | 北京航空航天大学 | Method for optimizing quick charge strategy of aerocar battery based on data-driven reduced-order model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109993270B (en) | Lithium ion battery residual life prediction method based on gray wolf group optimization LSTM network | |
| CN112348271A (en) | Short-term photovoltaic power prediction method based on VMD-IPSO-GRU | |
| CN111999649A (en) | XGboost algorithm-based lithium battery residual life prediction method | |
| CN111999648A (en) | Lithium battery residual life prediction method based on long-term and short-term memory network | |
| CN114861527A (en) | Lithium battery life prediction method based on time series characteristics | |
| CN110007235A (en) | A kind of accumulator of electric car SOC on-line prediction method | |
| CN113094822A (en) | Method and system for predicting residual life of mechanical equipment | |
| CN116609678A (en) | Lithium ion battery residual life prediction method based on improved mixed model | |
| CN116736133A (en) | Early prediction method for capacity degradation track of lithium ion battery in full life cycle | |
| CN112149896A (en) | Attention mechanism-based mechanical equipment multi-working-condition fault prediction method | |
| CN116794547A (en) | Lithium ion battery residual service life prediction method based on AFSA-GRU | |
| Eleftheriadis et al. | Hyperband optimization of stacked bidirectional long short-term memory neural network for the state of charge estimation | |
| CN114460481A (en) | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism | |
| CN117825965A (en) | State prediction method and system for lithium ion battery | |
| CN113033898A (en) | Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network | |
| CN117368784A (en) | Battery health state prediction method based on feature compression | |
| CN112014757A (en) | Battery SOH estimation method integrating capacity increment analysis and genetic wavelet neural network | |
| CN117312835A (en) | Fuel cell performance prediction method based on mixed probability data driving model | |
| CN117007974A (en) | Solid-state battery SOC estimation method based on model fusion | |
| CN116646929A (en) | PSO-CNN-BILSTM-based short-term wind power prediction method | |
| CN115758187A (en) | Coal mine mechanical motor state prediction method based on K-CNN-N-GRU | |
| Zhang et al. | Modeling of electric vehicle batteries using rbf neural networks | |
| CN115619028A (en) | Clustering algorithm fusion-based power load accurate prediction method | |
| Zhou et al. | Remaining useful life prediction and state of health diagnosis of lithium-ion batteries with multiscale health features based on optimized CatBoost algorithm | |
| CN115542168A (en) | Lithium battery residual service life prediction method based on fusion data driving model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |