CN108896924A - The charge states of lithium ion battery estimation method merged based on depth confidence network and Kalman filtering - Google Patents
The charge states of lithium ion battery estimation method merged based on depth confidence network and Kalman filtering Download PDFInfo
- Publication number
- CN108896924A CN108896924A CN201810747134.8A CN201810747134A CN108896924A CN 108896924 A CN108896924 A CN 108896924A CN 201810747134 A CN201810747134 A CN 201810747134A CN 108896924 A CN108896924 A CN 108896924A
- Authority
- CN
- China
- Prior art keywords
- soc
- moment
- dbn
- training
- lithium ion
- 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.)
- Granted
Links
Abstract
Based on the charge states of lithium ion battery estimation method that depth confidence network and Kalman filtering merge, it is related to technical field of battery management.The present invention is to solve the problems, such as that conventional method exists.The present invention from battery actual motion can electric current, voltage and temperature parameter measured directly, building lithium ion battery SOC estimate model.Lithium ion battery SOC is estimated using the depth confidence network in deep learning, depth confidence network is merged with kalman filter method, obtains merging later lithium ion battery SOC estimation model.For compared with prior art, the SOC estimation method based on data-driven method can carry out feature extraction according to the historical data that battery real work obtains, and systematic error be checked in conjunction with the method for filtering, to obtain relatively accurate SOC estimation.
Description
Technical field
The invention belongs to technical field of battery management, and in particular to a kind of charge states of lithium ion battery On-line Estimation side
Method.
Background technique
Lithium ion battery because its specific energy is high, operating voltage is high, temperature range is wide, self-discharge rate is low, have extended cycle life and
Many advantages, such as safety is good is widely used and mobile phone, laptop and electric automobiles, and is gradually extended to military logical
The fields such as letter, navigation, Aeronautics and Astronautics are increasingly becoming the key and support technology of following many key areas.
Battery charge state (State of Charge, SOC) is as an important pass in lithium ion battery management system
Bond parameter, the service efficiency of battery and the use peace of lifting system can be ensured by carrying out accurately estimation to lithium ion battery SOC
Quan Xing.Therefore accurately estimation is carried out to lithium ion battery SOC to have great importance.
Currently used SOC estimation strategy has open circuit voltage method, current integration method, AC impedence method, neural network, card
Kalman Filtering method etc..Open circuit voltage method is that SOC estimates one of strategy and widely used algorithm for estimating earlier.It
Basic principle be that the open-circuit voltage of battery and SOC have fixed relationship, general open-circuit voltage is dull with the increase of SOC
It is incremented by, by measuring the relation curve of open-circuit voltage and SOC, in this way, corresponding to the moment battery for determining open-circuit voltage
SOC value.But open circuit voltage method needs take a substantial amount of time and are tested, therefore cannot answer in actual use situation
With.The basic thought of current integration method is that battery is regarded as to a black box, regardless of internal state change, only to disengaging battery
Electric current integrated, to calculate the electricity of battery consumption or accumulation, the SOC value of the moment battery is obtained by calculation.
This method needs current measurement to have very high precision, and due to the cumulative effect of integral, the error that current measurement introduces will always
Inaccuracy that is cumulative, causing SOC to estimate.Therefore, current integration method is not used alone, it will usually improve in conjunction with other methods
The precision of SOC estimation.AC impedence method measures the corresponding of battery by the interference signal to the application of battery both ends by a small margin, into
And the AC impedance of battery is obtained, the SOC value of battery is estimated according to AC impedance.But the SOC of this method estimation battery
Existing defects.Firstly, the ac impedance spectroscopy variation of battery is obvious in the case where SOC is very big or very little, but for
SOC changes little situation, and the ac impedance spectroscopy of battery varies less, and if measurement accuracy is inadequate, application condition is big.Kalman
Filter method can be applied to real-time measurement, in the biggish electric car environment of current fluctuation, while can provide the mistake of estimated value
Poor range, but this method needs the model by battery.Complicated battery model is computationally intensive and needs largely to be tested,
And simply battery model characterization ability is weak, there are large errors under dynamic operation condition, therefore there are lithium ion batteries for this method
The problem of model modeling problem and identification of Model Parameters.In addition to the above method, there is also the SOC based on data-driven model to estimate
Meter method, such as neural network, support vector machines method etc..But these methods belong to the learning method of shallow-layer, in lithium ion
When battery operating condition complexity, SOC evaluated error is higher.Currently, depth confidence network (Deep Belief Network, DBN)
Every field is widely used in as a kind of deep learning model.Under the dynamic operation condition of lithium ion battery complexity, DBN by
It, being capable of good fitting lithium ion battery SOC variation in its powerful ability in feature extraction and nonlinear fitting ability.Meanwhile
The model is not needed upon the physical chemical mechanism of lithium ion battery, it is only necessary to be built according to the historical data to operate with batteries
Mould, therefore it is suitable for the electro-chemical systems of this complexity of lithium ion battery.
Shown in sum up, conventional method is not high for the estimated accuracy of the SOC of lithium ion battery, and is not suitable for real work
The lithium ion battery of situation.And the methods of Kalman filtering needs to carry out SOC estimation based on lithium ion battery equivalent model, it is equivalent
The problems such as model foundation is chosen there are model and model parameter solves.In addition, traditional data driving method is in complicated dynamic
Under operating condition, SOC evaluated error increases.At the same time, DBN belongs to a kind of model of deep learning, has powerful feature extraction
Ability and nonlinear fitting ability have good adaptability to the complication system of nonlinear and non-Gaussian, therefore in battery status
It has a good application prospect in estimation.
Summary of the invention
The present invention is low to the estimated accuracy of lithium ion battery SOC in actual condition in order to solve conventional method, and is based on
There is modeling problem and identification of Model Parameters in battery model method, thus provide a kind of lithium towards under dynamic operation condition from
Sub- battery charge state estimation method.
Charge states of lithium ion battery estimation method based on depth confidence network and Kalman filtering fusion, including it is following
Step：
Step 1：Voltage, electric current, temperature and the history SOC data of battery in battery testing data are extracted, and by extraction
Data are normalized, the data set after being normalized；
Step 2：Data set after normalization is divided into the No.1 input vector X at k moment according to the following formulak (1), No. two it is defeated
Incoming vector Xk (2)With output vector Yk, the k moment is current time；
Step 3：By the No.1 input vector X at k momentk (1), No. two input vector Xk (2)With output vector YkAccording to 40%
Ratio cut partition with 60% is training set and test set；
Step 4：The parameter of SOC initial estimate model of the random initializtion based on DBN and the SOC observation based on DBN are estimated
The parameter of model is counted, and above two model is trained using training set, two after being trained kind model；
Step 5：By the No.1 input vector X at k-1 moment in test setk-1 (1)As the SOC initial estimate mould based on DBN
The input vector of type, and obtain SOC initial value SOCinit, the k-1 moment is the last moment at current time；
Step 6：By Xk (1)With SOCinitCarry out No. three input vector X that splicing obtains the k momentk (3), by Xk (3)Substitute into base
In the SOC observation estimation model of DBN, the SOC observation Z at k moment is obtainedk；
Step 7：By ZkIt is input in KF algorithm, obtains k moment accurate SOC value SOCk。
It is further comprising the steps of for solving the accurate SOC value of subsequent time after step 7：
By SOCkWith the No.1 input vector X at k+1 momentk+1 (1)Spliced, obtains No. three input vectors at k+1 moment
Xk+1 (3),
By Xk+1 (1)It substitutes into the SOC observation estimation model based on DBN, obtains the SOC observation Z at k+1 momentk+1；
By Zk+1It is input in KF algorithm, obtains k+1 moment accurate SOC value SOCk+1。
In step 2, the No.1 input vector X at k momentk (1), No. two input vector Xk (2)With output vector YkExpression formula
It is as follows respectively：
Xk (1)=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p],
Xk (2)=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p,SOCk-1],
Yk=[SOCk],
Wherein, vkAnd vk-mRespectively indicate voltage value, the i at k moment and k-m momentkAnd ik-nWhen respectively indicating k moment and k-n
Current value, the t at quarterkAnd tk-pRespectively indicate temperature value, the SOC at k moment and k-p momentk-1Indicate the SOC value at k-1 moment, m, n
It is positive integer with p.
The specific method being trained in step 4 to the SOC initial estimate model based on DBN is：
It is randomly provided the weighted value and bias of the SOC initial estimate model based on DBN, completes model initialization；
By the X in training setk (1)As the input of the SOC initial estimate model based on DBN, unsupervised training is carried out, when
Deconditioning when training error is less than 10 ∧ -2；
By the X in training setk (1)As the input of the SOC initial estimate model based on DBN after unsupervised training, by Yk
As the label of the SOC initial estimate model based on DBN after unsupervised training, Training is carried out, when training error is small
Deconditioning when 10 ∧ -3 completes the training of the SOC initial estimate model based on DBN.
K moment accurate SOC value SOC is obtained in step 7kSpecific method be：
Step 7 one：State-space model is established,
The state-space model includes state equation and measurement equation, and two kinds of equations are specific as follows：
State equation：
Measure equation：
Wherein,Indicate SOC estimation, the I at k-1 momentkIndicate electric current, the C at k momentrateIndicate the appearance of battery
Amount, wkIndicate cell process noise, vkIndicate that battery measures noise；
Step 7 two：The SOC estimation at k moment is obtained using the state equation in step 7 one
Z is updated using the measurement equation in step 7 onek；
Step 7 three：The error covariance P at k moment is updated using following formulak,
Pk=APk-1AT+ Q,
The kalman gain K at k moment is updated using following formulak,
Kk=PkHT(HPkHT+R)-1,
Wherein, A indicates that battery parameter, H indicate that the parameter of measurement battery, R and Q indicate that two kinds of different Gausses of value make an uproar
Sound；
Step 7 four：By the updated Z of step 7 twokWithIt is merged according to the following formula, it is accurate to obtain the k moment
SOC value SOCk,
For the estimation problem of lithium ion battery SOC, the present invention can physics ginseng measured directly from battery real work
Number (voltage, electric current, temperature) sets out, and proposes a kind of SOC On-line Estimation method, this method is for lithium ion battery under dynamic operation condition
System has good adaptability, estimates that maximum relative error has well within 6%, and to different types of battery
Robustness.The input of this method be battery can voltage, electric current and temperature measured directly, export current for lithium ion battery
SOC value.This method does not need to establish lithium ion battery equivalent-circuit model, and does not need specific operation and know to model parameter
Not, it is only necessary to which the partial history data of the lithium ion battery can obtain good SOC estimation effect.
Detailed description of the invention
Fig. 1 is the SOC estimation method functional block diagram based on DBN-KF；
Fig. 2 is the SOC initial estimate model based on DBN；
Fig. 3 is that the SOC observation based on DBN estimates model；
Fig. 4 is four groups of lithium ion battery SOC estimated result curve graphs based on DBN-KF, wherein (a) indicates first group of electricity
Pond SOC estimated result (b) indicates first group of SOC evaluated error, (c) indicates second group of battery SOC estimated result, (d) indicates the
Two groups of SOC evaluated errors (e) indicate third group battery SOC estimated result, (f) indicate third group SOC evaluated error, (g) indicate
4th group of battery SOC estimated result (h) indicates the 4th group of SOC evaluated error.
Specific embodiment
Present embodiment is illustrated in conjunction with Fig. 4 and table 1.Present embodiment selects the random floor data collection of NASA PCoE
Four groups of sample data sets carry out battery charge state estimation,
Step 1：Voltage, electric current, temperature and the SOC data of battery in battery testing data are extracted, and by the data of extraction
Carry out [0,1] normalized, the data set after being normalized.
Step 2：Data set after normalization is divided into the No.1 input vector X at k moment according to the following formulak (1), No. two it is defeated
Incoming vector Xk (2)With output vector Yk；The No.1 input vector X at the k momentk (1), No. two input vector Xk (2)With output vector Yk
Expression formula difference it is as follows：
Xk (1)=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p],
Xk (2)=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p,SOCk-1],
Yk=[SOCk],
Wherein, vkAnd vk-mRespectively indicate voltage value, the i at k moment and k-m momentkAnd ik-nWhen respectively indicating k moment and k-n
Current value, the t at quarterkAnd tk-pRespectively indicate temperature value, the SOC at k moment and k-p momentk-1Indicate the SOC value at k-1 moment, m, n
It is positive integer with p.
Step 3：By the No.1 input vector X at k momentk (1), No. two input vector Xk (2)With output vector YkAccording to 40%
Ratio cut partition with 60% is training set and test set；
Step 4：It is randomly provided the weighted value and bias of the SOC initial estimate model based on DBN, it is initial to complete model
Change；By the X in training setk (1)As the input of the SOC initial estimate model based on DBN, unsupervised training is carried out, when training misses
Deconditioning when difference is less than 10 ∧ -2；By the X in training setk (1)As the SOC initial estimate based on DBN after unsupervised training
The input of model, by YkAs the label of the SOC initial estimate model based on DBN after unsupervised training, supervision instruction has been carried out
Practice, the deconditioning when training error is less than 10 ∧ -3, completes the training of the SOC initial estimate model based on DBN；
It is randomly provided weighted value, bias and the network of the SOC observation estimation model based on DBN, it is initial to complete model
Change；By the X in training setk (2)As the input of the SOC observation estimation model based on DBN, unsupervised training is carried out, training is worked as
Deconditioning when error is less than 10 ∧ -2；By the X in training setk (2)As the SOC observation based on DBN after unsupervised training
The input for estimating model, by YkAs the label of the SOC observation estimation model based on DBN after unsupervised training, carry out
Training, the deconditioning when training error is less than 10 ∧ -3 complete the instruction of the SOC observation estimation model based on DBN
Practice.
Step 5：By the No.1 input vector X at k-1 moment in test setk-1 (1)As the SOC initial estimate mould based on DBN
The input vector of type, and obtain SOC initial value SOCinit；
Step 6：By Xk (1)With SOCinitCarry out No. three input vector X that splicing obtains the k momentk (3),
Xk (3)=[Xk (1),SOCinit]=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p,SOCinit]
Wherein, vkAnd vk-mRespectively indicate the voltage value at k moment and k-m moment, ikAnd ik-nWhen respectively indicating k moment and k-n
The current value at quarter, tkAnd tk-pThe temperature value at k moment and k-p moment is respectively indicated, m, n and p are positive integer；
By Xk (3)It substitutes into the SOC observation estimation model based on DBN, obtains the SOC observation Z at k momentk。
Step 7：State-space model is established,
State equation：
Measure equation：
Wherein,Indicate SOC estimation, the I at k-1 momentkIndicate electric current, the C at k momentrateIndicate the appearance of battery
Amount, wkIndicate cell process noise, vkIndicate that battery measures noise；
The SOC estimation at k moment is obtained using state equationZ is updated using measurement equationk；
The error covariance P at k moment is updated using following formulak,
Pk=APk-1AT+ Q,
The kalman gain K at k moment is updated using following formulak,
Kk=PkHT(HPkHT+R)-1,
Wherein, A indicates that battery parameter, H indicate that the parameter of measurement battery, R and Q indicate that two kinds of different Gausses of value make an uproar
Sound；
By updated ZkWithIt is merged according to the following formula, obtains k moment accurate SOC value SOCk,
It is further comprising the steps of for solving the accurate SOC value of subsequent time after step 7：
By SOCkWith the No.1 input vector X at k+1 momentk+1 (1)Spliced, obtains No. three input vectors at k+1 moment
Xk+1 (3),
Xk+1 (3)=[Xk+1 (1),SOCk]=[vk+1,…,vk-m+1,ik+1,...,ik-n+1,tk+1,...,tk-p+1,SOCk]
Wherein, vk+1And vk-m+1Respectively indicate the voltage value at k+1 moment and k-m+1 moment, ik+1And ik-n+1Respectively indicate k+
The current value at 1 moment and k-n+1 moment, tk+1And tk-p+1Respectively indicate the temperature value at k+1 moment and k-p+1 moment；
By Xk+1 (1)It substitutes into the SOC observation estimation model based on DBN, obtains the SOC observation Z at k+1 momentk+1；It will
Zk+1It is input in KF (Kalman filtering fusion) algorithm, obtains k+1 moment accurate SOC value SOCk+1。
The SOC estimated result of four groups of batteries of the random floor data of NASA PCoE is as shown in the table：
1 SOC estimated result evaluation index of table
Present embodiment in order to realize, estimate under dynamic operation condition by battery SOC, can be measured directly from battery actual motion
Electric current, voltage and temperature parameter set out, and building lithium ion battery SOC estimates model.In view of lithium ion battery is with non-thread
Property etc. characteristics, therefore lithium ion battery SOC is estimated using the depth confidence network in deep learning.In view of measurement
There is measurement error in the process, therefore depth confidence network is merged with kalman filter method, obtains merging later
Lithium ion battery SOC estimates model.For compared with prior art, the SOC estimation method based on data-driven method can be with
Feature extraction is carried out according to the historical data that battery real work obtains, checks systematic error in conjunction with the method for filtering, thus
Obtain relatively accurate SOC estimation.
Claims (8)
1. based on the charge states of lithium ion battery estimation method that depth confidence network and Kalman filtering merge, feature exists
In including the following steps：
Step 1：Voltage, electric current, temperature and the history SOC data of battery in battery testing data are extracted, and by the data of extraction
It is normalized, the data set after being normalized；
Step 2：Data set after normalization is divided into the No.1 input vector X at k moment according to the following formulak (1), No. two input to
Measure Xk (2)With output vector Yk, the k moment is current time；
Step 3：By the No.1 input vector X at k momentk (1), No. two input vector Xk (2)With output vector YkIt is divided into training set
And test set；
Step 4：The parameter of SOC initial estimate model of the random initializtion based on DBN and SOC observation based on DBN estimate mould
The parameter of type, and above two model is trained using training set, two after being trained kind model；
Step 5：By the No.1 input vector X at k-1 moment in test setk-1 (1)As the SOC initial estimate model based on DBN
Input vector, and obtain SOC initial value SOCinit, the k-1 moment is the last moment at current time；
Step 6：By Xk (1)With SOCinitCarry out No. three input vector X that splicing obtains the k momentk (3), by Xk (3)It substitutes into and is based on DBN
SOC observation estimation model in, obtain the k moment SOC observation Zk；
Step 7：By ZkIt is input in KF algorithm, obtains k moment accurate SOC value SOCk。
2. the charge states of lithium ion battery according to claim 1 merged based on depth confidence network and Kalman filtering
Estimation method, which is characterized in that further comprising the steps of for solving the accurate SOC value of subsequent time after step 7：
By SOCkWith the No.1 input vector X at k+1 momentk+1 (1)Spliced, obtains No. three input vector X at k+1 momentk+1 (3),
By Xk+1 (1)It substitutes into the SOC observation estimation model based on DBN, obtains the SOC observation Z at k+1 momentk+1；
By Zk+1It is input in KF algorithm, obtains k+1 moment accurate SOC value SOCk+1。
3. the lithium ion battery according to claim 1 or 2 merged based on depth confidence network and Kalman filtering is charged
Method for estimating state, which is characterized in that in step 2, the No.1 input vector X at k momentk (1), No. two input vector Xk (2)With it is defeated
Outgoing vector YkExpression formula difference it is as follows：
Xk (1)=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p],
Xk (2)=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p,SOCk-1],
Yk=[SOCk],
Wherein, vkAnd vk-mRespectively indicate voltage value, the i at k moment and k-m momentkAnd ik-nRespectively indicate k moment and k-n moment
Current value, tkAnd tk-pRespectively indicate temperature value, the SOC at k moment and k-p momentk-1Indicate the SOC value at k-1 moment, m, n and p are equal
For positive integer.
4. the lithium ion battery according to claim 1 or 2 merged based on depth confidence network and Kalman filtering is charged
Method for estimating state, which is characterized in that the specific method that the SOC initial estimate model based on DBN is trained in step 4
For：
It is randomly provided the weighted value and bias of the SOC initial estimate model based on DBN, completes model initialization；
By the X in training setk (1)As the input of the SOC initial estimate model based on DBN, unsupervised training is carried out, when training misses
Deconditioning when difference is less than 10 ∧ -2；
By the X in training setk (1)As the input of the SOC initial estimate model based on DBN after unsupervised training, by YkAs
The label of the SOC initial estimate model based on DBN after unsupervised training carries out Training, when training error is less than 10
Deconditioning when ∧ -3 completes the training of the SOC initial estimate model based on DBN.
5. the lithium ion battery according to claim 1 or 2 merged based on depth confidence network and Kalman filtering is charged
Method for estimating state, which is characterized in that the specific side that the SOC observation estimation model based on DBN is trained in step 4
Method is：
It is randomly provided weighted value, bias and the network of the SOC observation estimation model based on DBN, completes model initialization；
By the X in training setk (2)As the input of the SOC observation estimation model based on DBN, unsupervised training is carried out, training is worked as
Deconditioning when error is less than 10 ∧ -2；
By the X in training setk (2)As the input of the SOC observation estimation model based on DBN after unsupervised training, by YkMake
The label that model is estimated for the SOC observation based on DBN after unsupervised training, carries out Training, works as training error
Deconditioning when less than 10 ∧ -3 completes the training of the SOC observation estimation model based on DBN.
6. the lithium ion battery according to claim 1 or 2 merged based on depth confidence network and Kalman filtering is charged
Method for estimating state, which is characterized in that k moment accurate SOC value SOC is obtained in step 7kSpecific method be：
Step 7 one：State-space model is established,
The state-space model includes state equation and measurement equation, and two kinds of equations are specific as follows：
State equation：
Measure equation：
Wherein,Indicate SOC estimation, the I at k-1 momentkIndicate electric current, the C at k momentrateIndicate capacity, the w of batteryk
Indicate cell process noise, vkIndicate that battery measures noise；
Step 7 two：The SOC estimation at k moment is obtained using the state equation in step 7 one
Z is updated using the measurement equation in step 7 onek；
Step 7 three：The error covariance P at k moment is updated using following formulak,
Pk=APk-1AT+ Q,
The kalman gain K at k moment is updated using following formulak,
Kk=PkHT(HPkHT+R)-1,
Wherein, A indicates that battery parameter, H indicate that the parameter of measurement battery, R and Q indicate the different Gaussian noise of two kinds of values；
Step 7 four：By the updated Z of step 7 twokWithIt is merged according to the following formula, obtains k moment accurate SOC value
SOCk,
7. the charge states of lithium ion battery according to claim 1 merged based on depth confidence network and Kalman filtering
Estimation method, which is characterized in that No. three input vector X at k momentk (3)Expression formula it is as follows：
Xk (3)=[Xk (1),SOCinit]=[vk,…,vk-m,ik,...,ik-n,tk,...,tk-p,SOCinit]
Wherein, vkAnd vk-mRespectively indicate the voltage value at k moment and k-m moment, ikAnd ik-nRespectively indicate k moment and k-n moment
Current value, tkAnd tk-pThe temperature value at k moment and k-p moment is respectively indicated, m, n and p are positive integer.
8. the charge states of lithium ion battery according to claim 2 merged based on depth confidence network and Kalman filtering
Estimation method, which is characterized in that No. three input vector X at k+1 momentk+1 (3)Expression formula it is as follows：
Xk+1 (3)=[Xk+1 (1),SOCk]=[vk+1,…,vk-m+1,ik+1,...,ik-n+1,tk+1,...,tk-p+1,SOCk]
Wherein, vk+1And vk-m+1Respectively indicate the voltage value at k+1 moment and k-m+1 moment, ik+1And ik-n+1Respectively indicate the k+1 moment
With the current value at k-n+1 moment, tk+1And tk-p+1The temperature value at k+1 moment and k-p+1 moment is respectively indicated, m, n and p are positive
Integer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810747134.8A CN108896924B (en) | 2018-07-09 | 2018-07-09 | Lithium ion battery state-of-charge estimation method based on fusion of deep belief network and Kalman filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810747134.8A CN108896924B (en) | 2018-07-09 | 2018-07-09 | Lithium ion battery state-of-charge estimation method based on fusion of deep belief network and Kalman filtering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108896924A true CN108896924A (en) | 2018-11-27 |
| CN108896924B CN108896924B (en) | 2020-08-04 |
Family
ID=64348734
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810747134.8A Active CN108896924B (en) | 2018-07-09 | 2018-07-09 | Lithium ion battery state-of-charge estimation method based on fusion of deep belief network and Kalman filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108896924B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109581223A (en) * | 2018-11-29 | 2019-04-05 | 吉林大学 | The DIE Temperature estimation method of Li-ion batteries piles based on Kalman filtering |
| CN110133524A (en) * | 2019-04-19 | 2019-08-16 | 顺丰科技有限公司 | Battery state of charge calculation method, device, server and medium |
| CN110135527A (en) * | 2019-06-12 | 2019-08-16 | 哈尔滨工业大学 | A kind of dynamical unmanned plane charge states of lithium ion battery estimating system and method |
| CN111200141A (en) * | 2020-01-14 | 2020-05-26 | 东北电力大学 | Proton exchange membrane fuel cell performance prediction and optimization method based on deep belief network |
| CN111624495A (en) * | 2020-04-28 | 2020-09-04 | 合肥工业大学 | Lithium battery SOC interval estimation method and system based on EKF (extended Kalman Filter) optimization by deep belief network |
| CN111948539A (en) * | 2019-05-17 | 2020-11-17 | 天津科技大学 | Kalman filtering lithium ion battery SOC estimation method based on deep reinforcement learning |
| CN112083331A (en) * | 2020-08-09 | 2020-12-15 | 昆明理工大学 | Fusion method for improving lithium ion battery state of charge estimation precision |
| CN112327165A (en) * | 2020-09-21 | 2021-02-05 | 电子科技大学 | Battery SOH prediction method based on unsupervised transfer learning |
| CN113495213A (en) * | 2020-03-20 | 2021-10-12 | 北京京东乾石科技有限公司 | Electric quantity prediction model construction method, electric quantity prediction model carrying equipment deployment method and electric quantity prediction model carrying equipment deployment device |
| CN114167295A (en) * | 2021-11-30 | 2022-03-11 | 北京理工大学 | Lithium ion battery SOC estimation method and system based on multi-algorithm fusion |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6534954B1 (en) * | 2002-01-10 | 2003-03-18 | Compact Power Inc. | Method and apparatus for a battery state of charge estimator |
| CN1601295A (en) * | 2004-10-25 | 2005-03-30 | 清华大学 | Estimation for accumulator loading state of electric vehicle and carrying out method thereof |
| CN106354991A (en) * | 2016-07-29 | 2017-01-25 | 宁波飞拓电器有限公司 | Emergency lamp battery SOC estimation method based on deep learning CKF |
| CN106908736A (en) * | 2017-03-17 | 2017-06-30 | 哈尔滨工业大学 | Based on the lithium battery method for predicting residual useful life that depth confidence net and Method Using Relevance Vector Machine are merged |
-
2018
- 2018-07-09 CN CN201810747134.8A patent/CN108896924B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6534954B1 (en) * | 2002-01-10 | 2003-03-18 | Compact Power Inc. | Method and apparatus for a battery state of charge estimator |
| CN1601295A (en) * | 2004-10-25 | 2005-03-30 | 清华大学 | Estimation for accumulator loading state of electric vehicle and carrying out method thereof |
| CN106354991A (en) * | 2016-07-29 | 2017-01-25 | 宁波飞拓电器有限公司 | Emergency lamp battery SOC estimation method based on deep learning CKF |
| CN106908736A (en) * | 2017-03-17 | 2017-06-30 | 哈尔滨工业大学 | Based on the lithium battery method for predicting residual useful life that depth confidence net and Method Using Relevance Vector Machine are merged |
Non-Patent Citations (2)
| Title |
|---|
| JIE YU 等: "Remote correction analysis of SOC accuracy based on deep belief network", 《 2017 CHINESE AUTOMATION CONGRESS (CAC)》 * |
| 候朋飞 等: "基于深度学习和量子遗传算法的电池SoC估算方法研究", 《微型机与应用》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109581223A (en) * | 2018-11-29 | 2019-04-05 | 吉林大学 | The DIE Temperature estimation method of Li-ion batteries piles based on Kalman filtering |
| CN109581223B (en) * | 2018-11-29 | 2020-08-11 | 吉林大学 | Kalman filtering based core temperature estimation method of lithium ion battery pack |
| CN110133524A (en) * | 2019-04-19 | 2019-08-16 | 顺丰科技有限公司 | Battery state of charge calculation method, device, server and medium |
| CN110133524B (en) * | 2019-04-19 | 2022-07-08 | 丰翼科技(深圳)有限公司 | Battery charge state calculation method, device, server and medium |
| CN111948539A (en) * | 2019-05-17 | 2020-11-17 | 天津科技大学 | Kalman filtering lithium ion battery SOC estimation method based on deep reinforcement learning |
| CN110135527A (en) * | 2019-06-12 | 2019-08-16 | 哈尔滨工业大学 | A kind of dynamical unmanned plane charge states of lithium ion battery estimating system and method |
| CN111200141A (en) * | 2020-01-14 | 2020-05-26 | 东北电力大学 | Proton exchange membrane fuel cell performance prediction and optimization method based on deep belief network |
| CN111200141B (en) * | 2020-01-14 | 2022-10-28 | 东北电力大学 | Proton exchange membrane fuel cell performance prediction and optimization method based on deep belief network |
| CN113495213A (en) * | 2020-03-20 | 2021-10-12 | 北京京东乾石科技有限公司 | Electric quantity prediction model construction method, electric quantity prediction model carrying equipment deployment method and electric quantity prediction model carrying equipment deployment device |
| CN111624495A (en) * | 2020-04-28 | 2020-09-04 | 合肥工业大学 | Lithium battery SOC interval estimation method and system based on EKF (extended Kalman Filter) optimization by deep belief network |
| CN111624495B (en) * | 2020-04-28 | 2022-08-02 | 合肥工业大学 | Lithium battery SOC interval estimation method and system based on EKF (extended Kalman Filter) optimization by deep belief network |
| CN112083331A (en) * | 2020-08-09 | 2020-12-15 | 昆明理工大学 | Fusion method for improving lithium ion battery state of charge estimation precision |
| CN112327165A (en) * | 2020-09-21 | 2021-02-05 | 电子科技大学 | Battery SOH prediction method based on unsupervised transfer learning |
| CN112327165B (en) * | 2020-09-21 | 2021-07-13 | 电子科技大学 | Battery SOH prediction method based on unsupervised transfer learning |
| CN114167295A (en) * | 2021-11-30 | 2022-03-11 | 北京理工大学 | Lithium ion battery SOC estimation method and system based on multi-algorithm fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108896924B (en) | 2020-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108896924A (en) | The charge states of lithium ion battery estimation method merged based on depth confidence network and Kalman filtering | |
| CN109031153B (en) | Online health state estimation method for lithium ion battery | |
| Wang et al. | An online method to simultaneously identify the parameters and estimate states for lithium ion batteries | |
| Zhang et al. | Lithium-ion battery pack state of charge and state of energy estimation algorithms using a hardware-in-the-loop validation | |
| CN106772064B (en) | A kind of health state of lithium ion battery prediction technique and device | |
| CN104569835B (en) | A kind of method of the state-of-charge for the electrokinetic cell for estimating electric automobile | |
| CN103941195B (en) | Method for battery SOC estimation based on small model error criterion expanding Kalman filter | |
| CN111581904B (en) | Lithium battery SOC and SOH collaborative estimation method considering cycle number influence | |
| CN105807226B (en) | Lithium ion battery SOC Forecasting Methodologies and device based on equivalent-circuit model | |
| CN105510829B (en) | A kind of Novel lithium ion power battery SOC methods of estimation | |
| CN106291381B (en) | A kind of method of Combined estimator electrokinetic cell system state-of-charge and health status | |
| CN105572596B (en) | Lithium battery SOC estimation method and system | |
| CN106772067B (en) | The method of Multiple Time Scales estimated driving force battery charge state and health status | |
| CN105929340B (en) | A method of battery SOC is estimated based on ARIMA | |
| CN110346734A (en) | A kind of lithium-ion-power cell health status evaluation method based on machine learning | |
| CN110208703A (en) | The method that compound equivalent-circuit model based on temperature adjustmemt estimates state-of-charge | |
| CN105319515A (en) | A combined estimation method for the state of charge and the state of health of lithium ion batteries | |
| CN105717460A (en) | Power battery SOC estimation method and system based on nonlinear observer | |
| CN109669132A (en) | A kind of battery charge state estimation method based on variation Bayesian filter | |
| CN106093782A (en) | The least square method supporting vector machine SOC method of estimation of dynamic modeling | |
| Xiong et al. | Data-driven state-of-charge estimator for electric vehicles battery using robust extended Kalman filter | |
| CN109143097A (en) | It is a kind of meter and temperature and cycle-index lithium ion battery SOC estimation method | |
| CN106291393A (en) | A kind of method for ONLINE RECOGNITION battery model parameter | |
| Qiuting et al. | State of health estimation for lithium-ion battery based on D-UKF | |
| Liu et al. | Battery degradation model and multiple-indicators based lifetime estimator for energy storage system design and operation: Experimental analyses of cycling-induced aging |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |