CN109633449A - Mining service life of lithium battery prediction technique and management system based on grey vector machine - Google Patents
Mining service life of lithium battery prediction technique and management system based on grey vector machine Download PDFInfo
- Publication number
- CN109633449A CN109633449A CN201811396523.7A CN201811396523A CN109633449A CN 109633449 A CN109633449 A CN 109633449A CN 201811396523 A CN201811396523 A CN 201811396523A CN 109633449 A CN109633449 A CN 109633449A
- Authority
- CN
- China
- Prior art keywords
- prediction
- model
- rvm
- dgm
- capacity
- 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.)
- Withdrawn
Links
Abstract
The invention discloses a kind of mining service life of lithium battery prediction technique and management system based on grey vector machine.The life-span prediction method establishes grey RVM regressive prediction model as training sample using mining capacity of lithium ion battery；DGM is established according to training sample data, using the predicted value of DGM as input, original training sample as output, training obtains RVM regressive prediction model；Capacity short-term forecast is done using DGM (1,1), and using predicted value as the input of RVM regressive prediction model, obtains the short-term regression forecasting result and prediction probability value of capacity, and introduce metabolic processes, updates training sample data；Correlation is judged using grey correlation analysis, and RVM model is dynamically updated with this result, obtains new associated vector, to obtain this method long-term trend prediction result.The present invention obtains the life prediction precision of more accurately mining lithium battery by acquiring Li-Battery monitor data in real time.
Description
Technical field
The present invention relates to a kind of mining service life of lithium battery prediction technique and management system based on grey vector machine, belong to lithium
Technical field of battery management.
Background technique
Lithium battery is with monomer operating voltage is high, small in size, light-weight, energy density is high, service life cycle is long, puts certainly
The advantages that electric current is small, memory-less effect, high pollution-free and cost performance, thus each row such as be widely used in communication, traffic, mining
Industry.Lithium battery includes battery core and protection two parts of circuit, and the protection circuit function of large-scale lithium battery is powerful, also known as management system
System, effect mainly ensure the uniformity of each economize on electricity tankage, the mistake for being diagnosed to be the battery problem in time, preventing battery
Charging and overdischarge, the state-of-charge for accurately obtaining battery etc..
Since lithium battery price is higher, scientifical use lithium battery is prolonged its service life, for reducing production cost, mentioning
High-environmental level all has realistic meaning.But due to the presence of noise, measurement error etc., existing lithium battery management system pair
The prediction of service life of lithium battery is also in low-level state.
Summary of the invention
It is an object of that present invention to provide a kind of service life of lithium battery prediction technique and management system based on grey vector machine, energy
The life prediction enough significantly improved to lithium battery is horizontal.
To achieve the above object, a kind of mining service life of lithium battery prediction technique based on grey vector machine, including following step
It is rapid:
The first step selects prediction model DGM (1,1), is defined as follows:
x(1)(k+1)=β1x(1)(k)+β2；
By carrying out simulation analysis to mining lithium battery cycle life test data, DGM (1,1) is to GM (1,1) model
Further precision, improves the stability of prediction to a certain extent.
Second step selects mining cycle life of lithium ion battery capacity sample data as initial training data, by sample
All data are converted to the number between [- 1,1], eliminate the quantity between cycle period number and capacity by normalized
Grade difference；
Third step, initialization RVM model parameter: Selection of kernel function gaussian kernel function, K (x, xi)=exp (- | | x-xi||2/
r2), carry out EM interative computation, noise variance σ2=0.1var (x), condition of convergence δ take 0.1, and weight w is set as
Wherein r is bandwidth；
4th step establishes predictive equation according to prediction model in the first step
β is solved with DGM (1,1)1And β2；Original non-negative training data sequence isIt is primary
Accumulating generation sequence are as follows:
WhereinBy X(1)It substitutes into the formula of the first step, obtains:
Y=B β
Wherein β=(β1,β2)T, to join sequence,
Then DGM differential equation x(1)(k+1)=β1x(1)(k)+β2Least-squares estimation parameter column meet β=(BTB)- 1BTY, and then can be calculated β1And β2；
It takesThe then estimated value of one-accumulate formation sequence are as follows:
Reduction can obtain DGM (1,1) prediction model:
It is iterated to calculate by the DGM (1,1) of foundation, updates original training data；
5th step establishes RVM regressive prediction model
Input data by (1, the 1) model of DGM in third step to the predicted value of original training data as RVM model,
Output data of the original training data as RVM obtains RVM regression model using EM iterative algorithm training RVM model；
6th step, lithium battery capacity prediction
Predicted value is inputted by the trend prediction of setting step-length to battery capacity by third step using DGM (1,1) model algorithm
In the RVM regression model that middle training obtains, the prediction result and probable range of battery capacity are obtained；
7th step, prediction terminate judgement
Judge whether battery capacity prediction value is greater than the capacity failure threshold values of setting, if more than the capacity failure valve of setting
Value, goes to the 8th step and continues to predict；If the capacity for being less than setting predicts threshold values, prediction terminates, and capacity is predicted to tie
Fruit and its confidence interval are converted to RUL value and corresponding confidence interval, and compare with actual RUL, to verify herein
The validity of method；
8th step, correlation analysis
Using metabolic method, the prediction result of the battery capacity in the 6th step is updated into original training data, is obtained
New training data；Short-term forecast is carried out in new training data input DGM (1,1) model algorithm；Finally with grey correlation point
Analysis method analyzes the degree of association between new training data and original training data；If the degree of association between the two is larger, it is greater than setting
Value returns to the 6th step and continues to predict；Conversely, jumping to the 5th step re -training RVM regression model, new RVM model is obtained, and
Continue to predict.
Meanwhile the present invention also provides a kind of lithium battery management system, which includes the above-mentioned lithium battery longevity
Order prediction technique.
The present invention selects battery capacity as original training data, establishes Grey Models of Dynamic Prediction, and what is generated is pre-
Measured value sets corresponding failure threshold as RVM mode input data, and it is related to original training data to obtain grey correlation distribution
Property judgement, carry out the judgement that prediction target is completed, and then capacity predicted value and prediction technique be converted into confidence interval.
Compared with prior art, the present invention predicts service life of lithium battery more accurate, can according to the prediction result of this method
To advanced optimize management system, energy content of battery utilization rate is maximized, battery is effectively extended.
Detailed description of the invention
Fig. 1 is flow chart of the present invention.
Specific embodiment
The following further describes the present invention with reference to the drawings:
A kind of mining service life of lithium battery prediction technique based on grey vector machine, comprising the following steps:
The first step selects prediction model DGM (1,1), is defined as follows:
x(1)(k+1)=β1x(1)(k)+β2；
By carrying out simulation analysis to mining lithium battery cycle life test data, DGM (1,1) is to GM (1,1) model
Further precision, improves the stability of prediction to a certain extent.
Second step selects mining cycle life of lithium ion battery capacity sample data as initial training data, by sample
All data are converted to the number between [- 1,1], eliminate the quantity between cycle period number and capacity by normalized
Grade difference；
Third step, initialization RVM model parameter: Selection of kernel function gaussian kernel function, K (x, xi)=exp (- | | x-xi||2/
r2), carry out EM interative computation, noise variance σ2=0.1var (x), condition of convergence δ take 0.1, and weight w is set as
Wherein r is bandwidth；
4th step establishes predictive equation according to prediction model in the first step
β is solved with DGM (1,1)1And β2；Original non-negative training data sequence isIt is primary
Accumulating generation sequence are as follows:
WhereinBy X(1)It substitutes into the formula of the first step, obtains:
Y=B β
Wherein β=(β1,β2)T, to join sequence,
Then DGM differential equation x(1)(k+1)=β1x(1)(k)+β2Least-squares estimation parameter column meet β=(BTB)- 1BTY, and then can be calculated β1And β2；
It takesThe then estimated value of one-accumulate formation sequence are as follows:
Reduction can obtain DGM (1,1) prediction model:
It is iterated to calculate by the DGM (1,1) of foundation, updates original training data；
5th step establishes RVM regressive prediction model
Input data by (1, the 1) model of DGM in third step to the predicted value of original training data as RVM model,
Output data of the original training data as RVM obtains RVM regression model using EM iterative algorithm training RVM model；
6th step, lithium battery capacity prediction
Predicted value is inputted by the trend prediction of setting step-length to battery capacity by third step using DGM (1,1) model algorithm
In the RVM regression model that middle training obtains, the prediction result and probable range of battery capacity are obtained；
7th step, prediction terminate judgement
Judge whether battery capacity prediction value is greater than the capacity failure threshold values of setting, if more than the capacity failure valve of setting
Value, goes to the 8th step and continues to predict；If the capacity for being less than setting predicts threshold values, prediction terminates, and capacity is predicted to tie
Fruit and its confidence interval are converted to RUL value and corresponding confidence interval, and compare with actual RUL, to verify herein
The validity of method；
8th step, correlation analysis
Using metabolic method, the prediction result of the battery capacity in the 6th step is updated into original training data, is obtained
New training data；Short-term forecast is carried out in new training data input DGM (1,1) model algorithm；Finally with grey correlation point
Analysis method analyzes the degree of association between new training data and original training data；If the degree of association between the two is larger, it is greater than setting
Value returns to the 6th step and continues to predict；Conversely, jumping to the 5th step re -training RVM regression model, new RVM model is obtained, and
Continue to predict.
Grey Incidence Analysis described in 8th step uses slope grey Relational Analysis Method, and this method is in traditional ash
Improvement on color association analysis method, resolution ratio is higher, mining capacity of lithium ion battery degradation trend is suitble to analyze, specific formula
It is as follows:
Assuming that two data sequencesWithDegree of association coefficient of relationship calculation formula between the two are as follows:
Wherein, Δ xk=xk+1-xk, Δ yk=yk+1-yk,
Finally obtain the degree of association between two data sequences:
Bandwidth r described in third step is core parameter, and the sparsity and accuracy of decision model, bandwidth is smaller, it is related to
Amount is more intensive, fitting precision is higher, while the complexity of model also increases, and calculates time growth, it is also possible to cause model excessively quasi-
It closes, model is made to lose sparsity, so bandwidth is suitable according to selection the characteristics of mining lithium battery, the preferred r=of bandwidth
5。
The number of iterations of EM iterative algorithm described in third step is more, calculates more accurate, but the number of iterations increases meter
Burden is calculated, so the application loop iteration number takes 1200.
The present invention also provides a kind of mining lithium battery management system, which includes any of the above-described described
Service life of lithium battery prediction technique.
Claims (5)
1. a kind of mining service life of lithium battery prediction technique based on grey vector machine, which comprises the following steps:
The first step selects prediction model DGM (1,1), is defined as follows:
x(1)(k+1)=β1x(1)(k)+β2；
Second step selects mining cycle life of lithium ion battery capacity sample data as initial training data, by sample normalizing
All data are converted to the number between [- 1,1] by change processing, and the quantity eliminated between cycle period number and capacity is differential
Not；
Third step, initialization RVM model parameter: Selection of kernel function gaussian kernel function, K (x, xi)=exp (- | | x-xi||2/r2),
Carry out EM interative computation, noise variance σ2=0.1var (x), condition of convergence δ take 0.1, and weight w is set asWherein r
For bandwidth；
4th step establishes predictive equation according to prediction model in the first step
β is solved with DGM (1,1)1And β2；Original non-negative training data sequence isIts one-accumulate
Formation sequence are as follows:
WhereinBy X(1)It substitutes into the formula of the first step, obtains:
Y=B β
Wherein β=(β1,β2)T, to join sequence,
Then DGM differential equation x(1)(k+1)=β1x(1)(k)+β2Least-squares estimation parameter column meet β=(BTB)-1BTY, into
And it can be calculated β1And β2；
It takesThe then estimated value of one-accumulate formation sequence are as follows:
Reduction can obtain DGM (1,1) prediction model:
It is iterated to calculate by the DGM (1,1) of foundation, updates original training data；
5th step establishes RVM regressive prediction model
Input data by (1, the 1) model of DGM in third step to the predicted value of original training data as RVM model is original
Output data of the training data as RVM obtains RVM regression model using EM iterative algorithm training RVM model；
6th step, lithium battery capacity prediction
Predicted value is inputted in third step and instructed by the trend prediction of setting step-length by battery capacity using DGM (1,1) model algorithm
In the RVM regression model got, the prediction result and probable range of battery capacity are obtained；
7th step, prediction terminate judgement
Judge whether battery capacity prediction value is greater than the capacity failure threshold values of setting, if more than the capacity failure threshold values of setting, turns
Continue to predict to the 8th step；If the capacity for being less than setting predicts threshold values, prediction terminates, and by capacity prediction result and its
Confidence interval is converted to RUL value and corresponding confidence interval, and compares with actual RUL, to verify context of methods
Validity；
8th step, correlation analysis
Using metabolic method, the prediction result of the battery capacity in the 6th step is updated into original training data, is obtained new
Training data；Short-term forecast is carried out in new training data input DGM (1,1) model algorithm；Finally with grey correlation analysis side
Method analyzes the degree of association between new training data and original training data；If the degree of association between the two is larger, it is greater than the set value, returns
The 6th step is returned to continue to predict；Conversely, jumping to the 5th step re -training RVM regression model, new RVM model is obtained, and continue
Prediction.
2. mining service life of lithium battery prediction technique according to claim 1, which is characterized in that grey described in the 8th step is closed
Connection analysis method is slope grey Relational Analysis Method, and formula is as follows:
Assuming that two data sequencesWithDegree of association coefficient of relationship calculation formula between the two are as follows:
Wherein, Δ xk=xk+1-xk, Δ yk=yk+1-yk,
Finally obtain the degree of association between two data sequences:
3. mining service life of lithium battery prediction technique according to claim 2, which is characterized in that bandwidth r described in third step
=5.
4. service life of lithium battery prediction technique according to claim 3, which is characterized in that EM interative computation described in third step
Largest loop the number of iterations take 1200.
5. a kind of mining lithium battery management system, which is characterized in that include claim 1 to any lithium of claim 4
Battery life predicting method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811396523.7A CN109633449A (en) | 2018-11-22 | 2018-11-22 | Mining service life of lithium battery prediction technique and management system based on grey vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811396523.7A CN109633449A (en) | 2018-11-22 | 2018-11-22 | Mining service life of lithium battery prediction technique and management system based on grey vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109633449A true CN109633449A (en) | 2019-04-16 |
Family
ID=66068819
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811396523.7A Withdrawn CN109633449A (en) | 2018-11-22 | 2018-11-22 | Mining service life of lithium battery prediction technique and management system based on grey vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109633449A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110187280A (en) * | 2019-05-20 | 2019-08-30 | 天津大学 | A method of the lithium battery remaining life probabilistic forecasting based on gray model |
| CN110187281A (en) * | 2019-05-22 | 2019-08-30 | 天津大学 | The method of lithium battery health status estimation based on charging stage health characteristics |
| CN110398697A (en) * | 2019-07-23 | 2019-11-01 | 北京工业大学 | A kind of lithium ion health status estimation method based on charging process |
| CN111273352A (en) * | 2020-01-15 | 2020-06-12 | 中国煤炭地质总局勘查研究总院 | Intelligent detection method and device for geological structure and electronic equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103033761A (en) * | 2012-12-17 | 2013-04-10 | 哈尔滨工业大学 | Lithium ion battery residual life forecasting method of dynamic gray related vector machine |
-
2018
- 2018-11-22 CN CN201811396523.7A patent/CN109633449A/en not_active Withdrawn
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103033761A (en) * | 2012-12-17 | 2013-04-10 | 哈尔滨工业大学 | Lithium ion battery residual life forecasting method of dynamic gray related vector machine |
Non-Patent Citations (1)
| Title |
|---|
| 姜玉叶: "矿用锂离子电池状态监测与寿命预测研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110187280A (en) * | 2019-05-20 | 2019-08-30 | 天津大学 | A method of the lithium battery remaining life probabilistic forecasting based on gray model |
| CN110187281A (en) * | 2019-05-22 | 2019-08-30 | 天津大学 | The method of lithium battery health status estimation based on charging stage health characteristics |
| CN110187281B (en) * | 2019-05-22 | 2021-06-04 | 天津大学 | Lithium battery health state estimation method based on charging stage health characteristics |
| CN110398697A (en) * | 2019-07-23 | 2019-11-01 | 北京工业大学 | A kind of lithium ion health status estimation method based on charging process |
| CN111273352A (en) * | 2020-01-15 | 2020-06-12 | 中国煤炭地质总局勘查研究总院 | Intelligent detection method and device for geological structure and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109633449A (en) | Mining service life of lithium battery prediction technique and management system based on grey vector machine | |
| CN110398697B (en) | Lithium ion health state estimation method based on charging process | |
| CN109993270A (en) | Lithium ion battery residual life prediction technique based on grey wolf pack optimization LSTM network | |
| CN111832220A (en) | Lithium ion battery health state estimation method based on codec model | |
| CN113064093B (en) | Method and system for jointly estimating state of charge and state of health of energy storage battery | |
| CN106055775B (en) | A kind of service life of secondary cell prediction technique that particle filter is combined with mechanism model | |
| CN112415414A (en) | Method for predicting remaining service life of lithium ion battery | |
| CN112051511A (en) | Power battery state of health estimation method and system based on multichannel technology | |
| CN110442941A (en) | It is a kind of to be tracked and RUL prediction technique based on the battery status for improving particle filter and process noise features fusion algorithm | |
| CN112269134A (en) | Battery SOC and SOH joint estimation method based on deep learning | |
| CN113702843B (en) | Lithium battery parameter identification and SOC estimation method based on suburb optimization algorithm | |
| CN109815995A (en) | Lithium battery method for predicting residual useful life under the conditions of a kind of missing observations | |
| CN112684363A (en) | Lithium ion battery health state estimation method based on discharge process | |
| CN111426957A (en) | SOC estimation optimization method for power battery under simulated vehicle working condition | |
| CN114660464A (en) | Lithium ion battery state of charge estimation method | |
| Navas et al. | Modelling Li-ion batteries using equivalent circuits for renewable energy applications | |
| Kong et al. | A novel parameter adaptive method for state of charge estimation of aged lithium batteries | |
| CN115308608A (en) | All-vanadium redox flow battery voltage prediction method, device and medium | |
| Zhou et al. | Remaining useful life prediction with probability distribution for lithium-ion batteries based on edge and cloud collaborative computation | |
| Schmitt et al. | State-of-health estimation by virtual experiments using recurrent decoder–encoder based lithium-ion digital battery twins trained on unstructured battery data | |
| CN113376541A (en) | Lithium ion battery health state prediction method based on CRJ network | |
| CN113011101A (en) | Control method and system for energy storage participating in frequency modulation auxiliary service optimization | |
| CN117407665A (en) | Retired battery time sequence data missing value filling method based on generation countermeasure network | |
| Yu et al. | SOH estimation method for lithium-ion battery based on discharge characteristics | |
| CN112580211A (en) | Lead-acid storage battery SOH estimation method based on SA and ANN algorithm |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190416 |
|
| WW01 | Invention patent application withdrawn after publication |