CN108549036A - Ferric phosphate lithium cell life-span prediction method based on MIV and SVM models - Google Patents
Ferric phosphate lithium cell life-span prediction method based on MIV and SVM models Download PDFInfo
- Publication number
- CN108549036A CN108549036A CN201810416011.6A CN201810416011A CN108549036A CN 108549036 A CN108549036 A CN 108549036A CN 201810416011 A CN201810416011 A CN 201810416011A CN 108549036 A CN108549036 A CN 108549036A
- Authority
- CN
- China
- Prior art keywords
- miv
- variable
- svm
- training
- lithium cell
- 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
Landscapes
- Secondary Cells (AREA)
Abstract
The invention discloses a kind of ferric phosphate lithium cell method for predicting residual useful life based on MIV and SVM models, by obtaining influence importance of the input variable to output with MIV algorithms, then most important variable is filtered out as input variable, avoids unessential independent variable being introduced into the training and test process of prediction model.The new training set and test set for only including preferred variable is obtained after variable is preferred, prediction model is trained using preferred training set and SVM, since SVM uses structural risk minimization as optiaml ciriterion, globally optimal solution can be obtained, can effectively improve forecasting efficiency and precision in conjunction with the prediction model trained by the preferred data set only comprising variables such as cycle-index, resistance.
Description
Technical field
The present invention relates to ferric phosphate lithium cell life prediction field more particularly to a kind of phosphoric acid based on MIV and SVM models
Lithium iron battery life-span prediction method.
Background technology
Ferric phosphate lithium cell is because it is widely used in many occasions there are many advantage, instead of traditional lead
Batteries such as storage, ni-Cd, and applying in various fields such as agricultural, communications industry, industry, the life with the mankind become it is close can not
Point.But there is also the puzzlements of life problems for ferric phosphate lithium cell.Several factors influence its charging and discharging capabilities, such as inside battery
Being lost of material, unreasonable occupation mode etc..These can all make the health status (State of Health, SOH) of battery
It gradually degenerates, SOH is used to indicate that the storage capacity of battery, is the parameter for describing battery performance state.If ignored
Its degenerative process can damage relevant equipment, can bring about great losses to the lives and properties of the mankind under serious situation.It is close several
Year, on the one hand related researcher starts to develop better battery, on the other hand also begin to that research is unfolded to the SOH of battery.Electricity
The monitoring of pond SOH can not only monitor the degenerate state of battery, additionally it is possible to prevent the generation of failure disaster accident.It is presently relevant
Researcher the material of battery and production technique are improved in all its bearings, but in practice, this problem does not have
Have to obtain basic solution, therefore assessment and prediction work are carried out to the SOH of battery, it can be long from largely guarantee battery
Phase, reliable work, prevent the generation of accident.Patil et al. using classification and returns the side being combined on the basis of algorithm
Method carrys out the SOH of real-time estimation lithium ion battery, but the method is possible to introduce irrelevant variable, leads to the precision for reducing model.
The present invention screens input variable with MIV algorithms on the basis of SVM models, eliminates the influence of irrelevant variable, carries
The high precision of model.
Invention content
The object of the present invention is to provide a kind of methods in energy Accurate Prediction ferric phosphate lithium cell service life.
The present invention provides a kind of ferric phosphate lithium cell life-span prediction methods based on MIV and SVM models, realize process
For:
Step 1:Capacity data of the ferric phosphate lithium cell under different variables is obtained, and data are divided into two groups, one group is
Training set, another group is test set;Wherein, the different variable include at least ferric phosphate lithium cell charge cutoff total voltage,
Electric discharge cut-off total voltage, cycle-index and internal resistance;The capacity data is the battery capacity of ferric phosphate lithium cell;
Step 2:The training parameter of BP networks in MIV algorithms is configured, determines the capacity data of ferric phosphate lithium cell
In each variable MIV values, and be ranked up by order of magnitude, set MIV absolute value threshold values, and MIV absolute values are less than
The variable of MIV absolute value threshold values is rejected, and the variable that MIV absolute values are more than MIV absolute value threshold values is retained, and is realized to input variable
Screening, and training set and test set after being screened;
Step 3:The parameter of SVM is configured, using the training set after screening as the input of SVM, training is predicted
Model;
Step 4:It is tested using test set, and verifies the accuracy of prediction model.
Wherein, the step of setting network training parameter includes:
According to the connection weight ω of input variable, input layer and implicit interlayerijAnd hidden layer threshold value a calculating hidden layers are defeated
Go out H, calculation formula is:
F is general hidden layer excitation function, and calculation formula is:
L is the number of nodes of hidden layer in formula;
H, connection weight ω are exported according to hidden layerjkThe prediction that BP neural network is calculated with output layer threshold value b exports O, and
By prediction output as battery can discharge capacity, calculation formula is:
Wherein, after the step of network training parameter is set, including:
Network training is carried out using the newff functions in BP neural network, the neuron number of hidden layer is set as 18, output
The neuron number of layer is set as 1, and the transfer function of hidden layer neuron is set as tansig, and the transfer function of output layer is set as
The training function of purelin, backpropagation are set as traingdm, then initialize BP networks, and the interval times that display is arranged are
50, learning rate 0.0000001, factor of momentum 0.9, iterations 100, error range target is 0.00004;
After completing BP neural network training, each variable in training set is added and subtracted 10% respectively, forms two new instructions
Practice collection, and two new training sets are emulated using built network, obtains two simulation results;
Two simulation results are subjected to mathematic interpolation, resulting value is influence value of the independent variable to output, and by data set
Number seeks the average of influence value, so that it may obtain the corresponding MIV values of the independent variable;
It repeats this step and calculates the MIV values of each independent variable, be ranked up by order of magnitude, set MIV absolute values
Threshold value, and the variable by MIV absolute values less than MIV absolute value threshold values is rejected, and is retained MIV absolute values and is more than MIV absolute value threshold values
Variable, remaining MIV absolute values be more than MIV absolute value threshold values variable form new training set and test set.
Wherein, input of the training set after Variable Selection as SVM models further includes that training set is normalized
The step of processing.
Wherein, SVM models are indicated with following formula:
s.t.yi(ωTxi+b)≥1-ξi
ξi>=0, i=1,2 ..., m. (4)
ω is normal vector in formula, determines the direction of hyperplane;B is displacement item;ξiFor slack variable;C is constant;(xi,yi)
For training sample.
Wherein, using the training set after screening as the input of SVM, trained the step of obtaining prediction model, includes:
The rough selection of SVM algorithm parameter c and g are carried out, SVMcgForRegress functions, c are selectedminIt is set as -8, cmaxIf
It is 8, gminFor -8, gmaxIt is 8;
It is finely selected again according to the result figure of selection, cminIt is set as -3, cmaxIt is set as 3, gminFor -3, gmaxIt is for 3, v
4, cstepIt is 0.5, gstepIt is 0.5, msestepIt is 0.01;
SVM network trainings are carried out according to the training set after selected parameter and normalization, obtain SVM prediction models.
Wherein, it to test set normalized, is predicted using obtained SVM prediction models, by prediction result and not
It is compared using the result of MIV algorithms, obtains prediction error and prediction error comparison diagram, carry out validity and accuracy
Verification.
The present invention is ranked up input variable importance with Mean Impact Value (MIV) algorithm, eliminates to the battery longevity
Life influences smaller variable's attribute, in conjunction with the ferric phosphate lithium cell Life Prediction Model trained using support vector machines, carries
High forecasting efficiency and precision of prediction.
Description of the drawings
Fig. 1 is a kind of flow of the ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
Schematic diagram.
Fig. 2 is to optimize in a kind of ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
Preceding battery pack can discharge capacity prediction comparison diagram.
Fig. 3 is to optimize in a kind of ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
Preceding battery pack can discharge capacity prediction difference figure.
Fig. 4 is to optimize in a kind of ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
Preceding SVM prediction-error images.
Fig. 5 is to optimize in a kind of ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
Battery pack afterwards can discharge capacity prediction comparison diagram.
Fig. 6 is to optimize in a kind of ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
Battery pack afterwards can discharge capacity prediction difference figure.
Fig. 7 is to optimize in a kind of ferric phosphate lithium cell life-span prediction method of the SVM models based on MIV provided by the invention
SVM prediction-error images afterwards.
Specific implementation mode
Further more detailed description is made to technical scheme of the present invention With reference to embodiment.Obviously, it is retouched
The embodiment stated is only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention,
The every other embodiment that those of ordinary skill in the art are obtained under the premise of not making creative work, should all belong to
The scope of protection of the invention.
The present invention provides a kind of ferric phosphate lithium cell method for predicting residual useful life based on MIV and SVM models, this method
The step of include:
Step 1:Capacity data of the ferric phosphate lithium cell under different variables is obtained, and data are divided into two groups, one group is
Training set, another group is test set;Wherein, the different variable include at least ferric phosphate lithium cell charge cutoff total voltage,
Electric discharge cut-off total voltage, cycle-index and internal resistance;The capacity data is the battery capacity of ferric phosphate lithium cell;
Step 2:The training parameter of BP networks in MIV algorithms is configured, determines the capacity data of ferric phosphate lithium cell
In each variable MIV values, and be ranked up by order of magnitude, set MIV absolute value threshold values, and MIV absolute values are less than
The variable of MIV absolute value threshold values is rejected, and the variable that MIV absolute values are more than MIV absolute value threshold values is retained, and is realized to input variable
Screening, and training set and test set after being screened;
Step 3:The parameter of SVM is configured, using the training set after screening as the input of SVM, training is predicted
Model;
Step 4:It is tested using test set, and verifies the accuracy of prediction model.
Wherein, the step of setting network training parameter includes:
According to the connection weight ω of input variable, input layer and implicit interlayerijAnd hidden layer threshold value a calculating hidden layers are defeated
Go out H, calculation formula is:
F is general hidden layer excitation function, and calculation formula is:
L is the number of nodes of hidden layer in formula;
H, connection weight ω are exported according to hidden layerjkThe prediction that BP neural network is calculated with output layer threshold value b exports O, and
By prediction output as battery can discharge capacity, calculation formula is:
After network training parameter is set, network training, the god of hidden layer are carried out using the newff functions in BP neural network
18 are set as through first number, the neuron number of output layer is set as 1, and the transfer function of hidden layer neuron is set as tansig, output
The transfer function of layer is set as purelin, and the training function of backpropagation is set as traingdm, then initializes BP networks, setting
The interval times of display are 50, learning rate 0.0000001, factor of momentum 0.9, iterations 100, error range target
It is 0.00004.
After completing BP neural network training, each variable in training set is added and subtracted 10% respectively, forms two new instructions
Practice collection, and two new training sets are emulated using built network, obtains two simulation results.
Two simulation results are subjected to mathematic interpolation, resulting value is influence value of the independent variable to output, and by data set
Number seeks the average of influence value, so that it may obtain the corresponding MIV values of the independent variable.
It repeats abovementioned steps and calculates the MIV values of each independent variable, be ranked up by order of magnitude, setting MIV is absolute
It is worth threshold value, and MIV absolute values is less than to the variable rejecting of MIV absolute value threshold values, retains MIV absolute values and be more than MIV absolute value thresholds
The variable of value, the variable that remaining MIV absolute values are more than MIV absolute value threshold values form new training set and test set.
Input of the training set as SVM models after Variable Selection, further includes that training set is normalized
The step of.
SVM models are indicated with following formula:
s.t.yi(ωTxi+b)≥1-ξi
ξi>=0, i=1,2 ..., m. (4)
ω is normal vector in formula, determines the direction of hyperplane;B is displacement item;ξiFor slack variable;C is constant;(xi,yi)
For training sample.
Using the training set after screening as the input of SVM, trained the step of obtaining prediction model, includes:
The rough selection of SVM algorithm parameter c and g are carried out, SVMcgForRegress functions, c are selectedminIt is set as -8, cmaxIf
It is 8, gminFor -8, gmaxIt is 8;
It is finely selected again according to the result figure of selection, cminIt is set as -3, cmaxIt is set as 3, gminFor -3, gmaxIt is for 3, v
4, cstepIt is 0.5, gstepIt is 0.5, msestepIt is 0.01;
SVM network trainings are carried out according to the training set after selected parameter and normalization, obtain SVM prediction models.
After obtaining SVM prediction models, to test set normalized, predicted using obtained SVM prediction models, it will
Prediction result and the result of unused MIV algorithms are compared, and are obtained prediction error and prediction error comparison diagram, are carried out effective
The verification of property and accuracy.
Embodiment 1:
Experimental data used in this example comes from Electronics Co., Ltd. of Shenzhen.The LiFePO4 of the experiment model
Battery correlation rating data is as follows:Specified monomer capacity 120Ah, specified charge cutoff voltage 3.65V, nominal discharge blanking voltage
Battery is composed in series battery pack by 2.5V.Input parameter be electric discharge cut-off total voltage, charge cutoff total voltage, cycle-index with
And internal resistance, output parameter are the active volume of battery pack.
When carrying out preferred without using four variables of MIV algorithms pair, SVM is trained using complete training set and predicts mould
Type, the model are as follows to the prediction case and actual conditions comparing result of test set:
Fig. 2 be battery pack can discharge capacity predict comparison diagram, Fig. 3 be battery pack can discharge capacity prediction difference figure, Fig. 4 is
SVM prediction-error images.
By Fig. 2, Fig. 3 and Fig. 4 it is found that can achieve the purpose that predict that battery can discharge capacity substantially using SVM networks.But
That the relative error after predicting is larger, maximum can discharge capacity prediction error reached 2.2Ah, maximum relative error reaches
2.7% or so.
By research, the quantity and independent variable parameter of training sample all can be to the performances of the SVM prediction models trained
It is presented with large effect.Under normal circumstances, input variable is all that researcher is pre- according to the knowledge of profession and abundant experience
It is first selected, but in practical applications, the selection of input variable is difficult to determine in advance, therefore is easy to some are inessential
Independent variable be introduced into network, reduce the estimated performance of model, thus training prediction model during to input
Independent variable parameter is preferably of great significance.
MIV algorithms can effectively measure weighing factor of the input unit to output unit, and by the algorithm, we obtain four
The MIV weights of a battery variable parameter, as shown in table 1.
1 input variable MIV weights of table
Electric discharge cut-off total voltage/V | Charge cutoff total voltage/V | Cycle-index | Internal resistance/m Ω |
0.0057 | -0.0025 | 0.2159 | 0.8612 |
Each input variable is different to the weighing factor of result as shown in Table 1, the selection result show cycle-index and
Internal resistance is to influence most important two factors of battery life.
After Variable Selection, we only retain training set and two variables of cycle-index and internal resistance in test set, profit
SVM prediction models are trained with the training set after screening, and estimated performance analysis is carried out using the test the set pair analysis model after screening.
Fig. 5 be carry out variable optimal screening after battery pack can discharge capacity prediction comparison diagram, Fig. 6 be optimize after battery pack can discharge
Capacity prediction difference figure, Fig. 7 are the SVM prediction model prediction-error images trained after optimizing.
Table 2 optimizes front and back prediction error comparison
Error | Before optimization | After optimization |
Root-mean-square error | 0.2630 | 0.1620 |
Mean error | 0.0016 | 0.0012 |
By Fig. 5, Fig. 6 and Fig. 7 it is found that after being screened to variable parameter by MIV algorithms, maximum relative error is predicted
It is 1.4%, is greatly improved than before, the fitting degree of actual value and predicted value is higher, and table 2 is pre- twice before and after optimizing
Survey root-mean-square error and the average relative error comparison of result, it can be seen that the model prediction accuracy after optimization has larger carry
It rises, it was demonstrated that the validity of this paper institutes extracting method.
The present invention is directed to the life prediction problem of ferric phosphate lithium cell, is primarily based on MIV algorithms and optimizes sieve to variable
Choosing, avoids during unessential variable is introduced into the training and prediction of prediction model, the prediction trained after optimization
Model has excellent performance, advantageously accounts for the problems such as battery life predicting period is long, of high cost and precision is not high.
It these are only embodiments of the present invention, be not intended to limit the scope of the invention, it is every to utilize the present invention
Equivalent structure or equivalent flow shift made by specification and accompanying drawing content is applied directly or indirectly in other relevant technologies
Field is included within the scope of the present invention.
Claims (7)
1. a kind of ferric phosphate lithium cell method for predicting residual useful life based on MIV and SVM models, it is characterised in that:
Step 1:Capacity data of the ferric phosphate lithium cell under different variables is obtained, and data are divided into two groups, one group is training
Collection, another group is test set;Wherein, the different variables include at least the charge cutoff total voltage of ferric phosphate lithium cell, electric discharge
End total voltage, cycle-index and internal resistance;The capacity data is the battery capacity of ferric phosphate lithium cell;
Step 2:The training parameter of BP networks in MIV algorithms is configured, is determined each in the capacity data of ferric phosphate lithium cell
The MIV values of a variable, and be ranked up by order of magnitude, MIV absolute value threshold values are set, and MIV absolute values is exhausted less than MIV
Variable to being worth threshold value is rejected, and the variable that MIV absolute values are more than MIV absolute value threshold values is retained, and realizes the screening to input variable,
And the training set after being screened and test set;
Step 3:The parameter of SVM is configured, using the training set after screening as the input of SVM, training obtains prediction mould
Type;
Step 4:It is tested using test set, and verifies the accuracy of prediction model.
2. the ferric phosphate lithium cell method for predicting residual useful life according to claim 1 based on MIV and SVM models, special
Sign is:
Be arranged network training parameter the step of include:
According to the connection weight ω of input variable, input layer and implicit interlayerijAnd hidden layer threshold value a calculates hidden layer and exports H,
Calculation formula is:
F is general hidden layer excitation function, and calculation formula is:
L is the number of nodes of hidden layer in formula;
H, connection weight ω are exported according to hidden layerjkThe prediction that BP neural network is calculated with output layer threshold value b exports O, and will be pre-
Survey output as battery can discharge capacity, calculation formula is:
3. the ferric phosphate lithium cell method for predicting residual useful life according to claim 1 based on MIV and SVM models, special
Sign is:
After the step of network training parameter is set, including:
Network training is carried out using the newff functions in BP neural network, the neuron number of hidden layer is set as 18, output layer
Neuron number is set as 1, and the transfer function of hidden layer neuron is set as tansig, and the transfer function of output layer is set as purelin,
The training function of backpropagation is set as traingdm, then initializes BP networks, and the interval times that display is arranged are 50, learning rate
It is 0.0000001, factor of momentum 0.9, iterations 100, error range target is 0.00004;
After completing BP neural network training, each variable in training set is added and subtracted 10% respectively, forms two new training sets,
And emulate two new training sets using built network, obtain two simulation results;
Two simulation results are subjected to mathematic interpolation, resulting value is influence value of the independent variable to output, and is asked by number of data sets
Take the average of influence value, so that it may obtain the corresponding MIV values of the independent variable;
It repeats this step and calculates the MIV values of each independent variable, be ranked up by order of magnitude, set MIV absolute value threshold values,
And the variable that MIV absolute values are less than to MIV absolute value threshold values is rejected, and the change that MIV absolute values are more than MIV absolute value threshold values is retained
Amount, the variable that remaining MIV absolute values are more than MIV absolute value threshold values form new training set and test set.
4. the ferric phosphate lithium cell method for predicting residual useful life according to claim 1 based on MIV and SVM models, special
Sign is:
Input of the training set as SVM models after Variable Selection, further includes the step that training set is normalized
Suddenly.
5. the ferric phosphate lithium cell method for predicting residual useful life according to claim 1 based on MIV and SVM models, special
Sign is:
SVM models are indicated with following formula:
ω is normal vector in formula, determines the direction of hyperplane;B is displacement item;ξiFor slack variable;C is constant;(xi,yi) it is instruction
Practice sample.
6. the ferric phosphate lithium cell method for predicting residual useful life according to claim 5 based on MIV and SVM models, special
Sign is:
Using the training set after screening as the input of SVM, trained the step of obtaining prediction model, includes:
The rough selection of SVM algorithm parameter c and g are carried out, SVMcgForRegress functions, c are selectedminIt is set as -8, cmaxIt is set as 8,
gminFor -8, gmaxIt is 8;
It is finely selected again according to the result figure of selection, cminIt is set as -3, cmaxIt is set as 3, gminFor -3, gmaxFor 3, v 4, cstep
It is 0.5, gstepIt is 0.5, msestepIt is 0.01;
SVM network trainings are carried out according to the training set after selected parameter and normalization, obtain SVM prediction models.
7. the ferric phosphate lithium cell method for predicting residual useful life according to claim 1 based on MIV and SVM models, special
Sign is:
It to test set normalized, is predicted using obtained SVM prediction models, prediction result and unused MIV is calculated
The result of method is compared, and is obtained prediction error and prediction error comparison diagram, is carried out the verification of validity and accuracy.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810416011.6A CN108549036A (en) | 2018-05-03 | 2018-05-03 | Ferric phosphate lithium cell life-span prediction method based on MIV and SVM models |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810416011.6A CN108549036A (en) | 2018-05-03 | 2018-05-03 | Ferric phosphate lithium cell life-span prediction method based on MIV and SVM models |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108549036A true CN108549036A (en) | 2018-09-18 |
Family
ID=63513471
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810416011.6A Pending CN108549036A (en) | 2018-05-03 | 2018-05-03 | Ferric phosphate lithium cell life-span prediction method based on MIV and SVM models |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108549036A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109031152A (en) * | 2018-10-15 | 2018-12-18 | 重庆科技学院 | Lead-acid accumulator residual capacity flexible measurement method based on neural network transparence |
CN109613440A (en) * | 2019-01-17 | 2019-04-12 | 清华-伯克利深圳学院筹备办公室 | Stage division, device, equipment and the storage medium of battery |
CN109740809A (en) * | 2018-12-27 | 2019-05-10 | 北京五维星宇科技有限公司 | The equipment cluster method for predicting residual useful life and terminal to be developed based on single dress service life |
CN110007660A (en) * | 2019-04-10 | 2019-07-12 | 河北工业大学 | A kind of equivalent thermal stress online soft sensor method of steam turbine of thermal power plant group transient state |
CN110443377A (en) * | 2019-06-24 | 2019-11-12 | 南方电网调峰调频发电有限公司信息通信分公司 | A kind of support vector machines life of storage battery prediction technique based on immune algorithm optimization |
CN110555226A (en) * | 2019-04-03 | 2019-12-10 | 太原理工大学 | method for predicting residual life of lithium iron phosphate battery based on EMD and MLP |
CN110568359A (en) * | 2019-09-04 | 2019-12-13 | 太原理工大学 | lithium battery residual life prediction method |
CN110850315A (en) * | 2019-11-29 | 2020-02-28 | 北京邮电大学 | Method and device for estimating state of charge of battery |
CN113094989A (en) * | 2021-04-07 | 2021-07-09 | 贵州大学 | Unmanned aerial vehicle battery life prediction method based on random configuration network |
CN113391209A (en) * | 2021-05-26 | 2021-09-14 | 江苏小牛电动科技有限公司 | Method, device and system for predicting health state of battery and battery |
CN113589175A (en) * | 2021-08-23 | 2021-11-02 | 上海智能新能源汽车科创功能平台有限公司 | Fuel cell service life prediction method and system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060212277A1 (en) * | 2005-03-16 | 2006-09-21 | Terry Hansen | Apparatus and method for estimating battery state of charge |
CN104156791A (en) * | 2014-08-29 | 2014-11-19 | 哈尔滨工业大学 | Lithium ion battery residual life predicting method based on LS-SVM probability ensemble learning |
CN105808914A (en) * | 2014-12-31 | 2016-07-27 | 北京航天测控技术有限公司 | Method and device for predicting life of satellite lithium ion battery |
CN107765190A (en) * | 2017-12-11 | 2018-03-06 | 太原理工大学 | A kind of life-span prediction method of long-life fast charging type ferric phosphate lithium cell |
-
2018
- 2018-05-03 CN CN201810416011.6A patent/CN108549036A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060212277A1 (en) * | 2005-03-16 | 2006-09-21 | Terry Hansen | Apparatus and method for estimating battery state of charge |
CN104156791A (en) * | 2014-08-29 | 2014-11-19 | 哈尔滨工业大学 | Lithium ion battery residual life predicting method based on LS-SVM probability ensemble learning |
CN105808914A (en) * | 2014-12-31 | 2016-07-27 | 北京航天测控技术有限公司 | Method and device for predicting life of satellite lithium ion battery |
CN107765190A (en) * | 2017-12-11 | 2018-03-06 | 太原理工大学 | A kind of life-span prediction method of long-life fast charging type ferric phosphate lithium cell |
Non-Patent Citations (4)
Title |
---|
于洋 等: "基于LS-SVM算法动力电池SOC估计方法的研究", 《电源技术研究与设计》 * |
张金国 等: "基于MIV的BP神经网络磷酸铁锂电池寿命预测", 《电源技术研究与设计》 * |
聂铭 等: "基于MIV-SVM的烤烟评吸质量预测模型", 《中国烟草学报》 * |
赵振江: "基于SMO-SVM算法的变压器故障诊断", 《煤矿机械》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109031152A (en) * | 2018-10-15 | 2018-12-18 | 重庆科技学院 | Lead-acid accumulator residual capacity flexible measurement method based on neural network transparence |
CN109740809A (en) * | 2018-12-27 | 2019-05-10 | 北京五维星宇科技有限公司 | The equipment cluster method for predicting residual useful life and terminal to be developed based on single dress service life |
CN109740809B (en) * | 2018-12-27 | 2021-03-23 | 北京五维星宇科技有限公司 | Equipment cluster residual life prediction method and terminal based on single-mounted life evolution |
CN109613440A (en) * | 2019-01-17 | 2019-04-12 | 清华-伯克利深圳学院筹备办公室 | Stage division, device, equipment and the storage medium of battery |
CN109613440B (en) * | 2019-01-17 | 2021-06-08 | 清华-伯克利深圳学院筹备办公室 | Battery grading method, device, equipment and storage medium |
CN110555226A (en) * | 2019-04-03 | 2019-12-10 | 太原理工大学 | method for predicting residual life of lithium iron phosphate battery based on EMD and MLP |
CN110007660B (en) * | 2019-04-10 | 2020-06-16 | 河北工业大学 | Online soft measurement method for transient equivalent thermal stress of steam turbine set of thermal power plant |
CN110007660A (en) * | 2019-04-10 | 2019-07-12 | 河北工业大学 | A kind of equivalent thermal stress online soft sensor method of steam turbine of thermal power plant group transient state |
CN110443377A (en) * | 2019-06-24 | 2019-11-12 | 南方电网调峰调频发电有限公司信息通信分公司 | A kind of support vector machines life of storage battery prediction technique based on immune algorithm optimization |
CN110443377B (en) * | 2019-06-24 | 2022-11-01 | 南方电网调峰调频发电有限公司信息通信分公司 | Support vector machine storage battery life prediction method based on immune algorithm optimization |
CN110568359A (en) * | 2019-09-04 | 2019-12-13 | 太原理工大学 | lithium battery residual life prediction method |
CN110568359B (en) * | 2019-09-04 | 2021-11-23 | 太原理工大学 | Lithium battery residual life prediction method |
CN110850315A (en) * | 2019-11-29 | 2020-02-28 | 北京邮电大学 | Method and device for estimating state of charge of battery |
CN113094989A (en) * | 2021-04-07 | 2021-07-09 | 贵州大学 | Unmanned aerial vehicle battery life prediction method based on random configuration network |
CN113391209A (en) * | 2021-05-26 | 2021-09-14 | 江苏小牛电动科技有限公司 | Method, device and system for predicting health state of battery and battery |
CN113589175A (en) * | 2021-08-23 | 2021-11-02 | 上海智能新能源汽车科创功能平台有限公司 | Fuel cell service life prediction method and system |
CN113589175B (en) * | 2021-08-23 | 2024-04-12 | 上海智能新能源汽车科创功能平台有限公司 | Fuel cell life prediction method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108549036A (en) | Ferric phosphate lithium cell life-span prediction method based on MIV and SVM models | |
Guo et al. | Online estimation of SOH for lithium-ion battery based on SSA-Elman neural network | |
CN105277896B (en) | Lithium battery method for predicting residual useful life based on ELM MUKF | |
CN111584952B (en) | Method and system for online evaluation of electrochemical cells of energy storage power station | |
CN107290679B (en) | The Intelligentized battery method for detecting health status of charging pile is shared for electric car | |
CN111474490B (en) | Rapid screening method for batteries used in echelon | |
CN108037463A (en) | A kind of lithium ion battery life-span prediction method | |
CN107870306A (en) | A kind of lithium battery charge state prediction algorithm based under deep neural network | |
CN107093775A (en) | A kind of method for evaluating consistency and device of cascaded structure battery pack | |
CN107436411A (en) | Battery SOH On-line Estimation methods based on fractional order neural network and dual-volume storage Kalman | |
CN112149873B (en) | Low-voltage station line loss reasonable interval prediction method based on deep learning | |
Zhang et al. | Intelligent state of charge estimation of battery pack based on particle swarm optimization algorithm improved radical basis function neural network | |
CN108960321B (en) | Battery fault prediction method for large lithium battery energy storage power station | |
CN106777786A (en) | A kind of lithium ion battery SOC estimation method | |
CN111027625A (en) | Battery screening method based on SAE and K-means clustering algorithm | |
CN110909802A (en) | Improved PSO (particle swarm optimization) based fault classification method for optimizing PNN (portable network) smoothing factor | |
CN104332665A (en) | Battery grouping method | |
CN110888077A (en) | Accelerated lithium ion battery life evaluation method based on ARIMA time sequence | |
CN111639882B (en) | Deep learning-based electricity risk judging method | |
CN112791997A (en) | Method for cascade utilization and screening of retired battery | |
CN111487532B (en) | Retired battery screening method and system based on analytic hierarchy process and entropy method | |
CN116449218B (en) | Lithium battery health state estimation method | |
CN107330573A (en) | A kind of state evaluating method and device of photovoltaic system key equipment | |
CN112036083B (en) | Similar product residual life prediction method and system | |
CN113917336A (en) | Lithium ion battery health state prediction method based on segment charging time and GRU |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180918 |
|
RJ01 | Rejection of invention patent application after publication |