CN109031147A - A kind of SOC estimation method of ferric phosphate lithium cell group - Google Patents
A kind of SOC estimation method of ferric phosphate lithium cell group Download PDFInfo
- Publication number
- CN109031147A CN109031147A CN201810953759.XA CN201810953759A CN109031147A CN 109031147 A CN109031147 A CN 109031147A CN 201810953759 A CN201810953759 A CN 201810953759A CN 109031147 A CN109031147 A CN 109031147A
- Authority
- CN
- China
- Prior art keywords
- neural network
- connection weight
- soc
- estimation method
- soc estimation
- 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
Landscapes
- Secondary Cells (AREA)
Abstract
The present invention provides a kind of methods using BP neural network estimation ferric phosphate lithium cell group SOC.BP neural network is using 1 hidden layer, 3 inputs, 1 network structure exported, the number of hidden nodes 11.Evaluation method includes the following steps: to be adjusted the connection weight adjustment link of BP neural network, it is inserted into an inertia coeffeicent, connection weight adjustment is improved using smooth weighted calculation, when being attached weighed value adjusting calculating, increase former connection weight item, and assigns heavier weight.The present invention changes the parameter of BP neural network in such a way that current integration is modified, reduces the influence that ferric phosphate lithium cell pool-size is decayed to SOC estimation precision.
Description
Technical field
The application belongs to electrochemical energy storage monitoring technical field, is related to a kind of SOC estimation method of ferric phosphate lithium cell group,
More particularly to a kind of SOC estimation method of ferric phosphate lithium cell group based on BP neural network.
Background technique
In use, ferric phosphate lithium cell group remaining capacity directly influences the decision of energy management and running, and mistake
It fills or over-discharge then directly affects the service life of ferric phosphate lithium cell group, state-of-charge (the State of of ferric phosphate lithium cell group
Charge, SOC) become electric energy management an important reference indicator.Since the SOC estimation of ferric phosphate lithium cell group is related to complexity
Electrochemical reaction, in the operating condition, voltage, electric current constantly change energy storage ferric phosphate lithium cell group, SOC accurately be estimated as
For the problem of technical staff's headache.Meanwhile there is capacity attenuation in the long-time service of ferric phosphate lithium cell group, influence SOC's
Accurate estimation.
The method of existing prediction SOC has very much, such as charge-discharge test method, open circuit voltage method, current integration method, Kalman
Filter method etc. may only use, and calculate however, these methods have respective limitation under the conditions of some specific
Process complexity is cumbersome, and accuracy is poor.The available reliable SOC of charge-discharge test method, but special charging/discharging apparatus is needed,
Time is long, at high cost, and is off-line measurement, is unable to get real time data.Open circuit voltage method must stand for a long time battery, when
The voltage measured when voltage reaches relatively stable is only effectively.Current integration method is the discharge current pair by calculating lithium battery
The integral of certain time calculates the SOC of battery, can on-line measurement, and without establishing complicated SOC model, but estimation result
Influenced error occur by factors such as battery temperature, charge and discharge ratio, cell degradations, estimation error can be with the accumulation of time
It is continuously increased, and the self-discharge phenomenon of initial time also will affect the estimation result of SOC.Kalman filtering method is based on lowest mean square
Poor principle describes state migration procedure using the state equation of system, passes through state equation using SOC as the quantity of state of system
Transfer specific description each moment between state correlation function, to realize that SOC is estimated.However it requires battery model
It is higher, and need to do a large amount of matrix operation, calculation amount is very big, and traditional Kalman filtering method is only suitable for linear system, because
This is not suitable for that nonlinear lithium iron phosphate battery system is presented in operation.
To solve the above-mentioned problems, Many researchers use BP neural network to estimate the SOC of ferric phosphate lithium cell.BP
Neural network is Back Propagation neural network, is the multilayer feedforward neural network by Back Propagation Algorithm training.
The Chinese invention patent application of Publication No. CN105823989A discloses a kind of improved RBF neural
Battery SOC prediction technique establishes power battery SOC prediction model by RBF neural method, with the output of neural network
Battery SOC establishes Optimized model as evaluation index, then using artificial fish-swarm algorithm respectively to the width vector of neural network,
The weight of center vector and output neuron optimizes calculating.
The Chinese invention patent application of Publication No. CN107037373A, disclose a kind of electric power storage neural network based
Pond remaining capacity prediction technique, including initial model is predicted using neural network model building remaining capacity；Obtain battery
Multiple groups voltage, electric current and remaining capacity data, using voltage and current data as input sample, remaining capacity data are as expectation
Output determines the hidden node of neural network using L1/2 regularization method after being input in remaining capacity prediction initial model
Number, then remaining capacity prediction initial model is trained, obtain multiple remaining capacity forecast value revision models；Selection one is to the phase
Hope the smallest model of error of output as final remaining capacity prediction model；By the electric current and electricity of the battery for needing to predict
Pressure value is input in remaining capacity prediction model, the residual electric quantity of obtained battery.
The Chinese invention patent application of Publication No. CN105974327A discloses a kind of based on neural network and UKF
Lithium battery group SOC prediction technique passes through the history charge and discharge data and corresponding SOC data to electric automobile lithium battery group
It is analyzed, establishes BP neural network, related data is normalized, according to the data after normalized to BP mind
It is trained through network.
These prediction techniques in the prior art, do not account for the nonlinear characteristic of inside battery complexity, need to pass through instruction
Practice a large amount of sample data, changes that flatter area efficiency is lower in SOC, it is low in the pixel accuracy that SOC is changed greatly.
Summary of the invention
To solve the above-mentioned problems, the SOC for the ferric phosphate lithium cell group based on BP neural network that the present invention provides a kind of
Evaluation method.The present invention constructs ferric phosphate lithium cell in the prediction model of SOC, to the company of BP neural network using BP neural network
It connects weighed value adjusting step to be improved, is inserted into an inertia coeffeicent, the connection weight tune is improved using smooth weighted calculation
Perfect square formula increases the weight of former connection weight item, the original connection weight item when carrying out the connection weight adjustment calculating
Refer to a preceding calculated result.
Preferably, the BP neural network can be using 1 hidden layer, 3 inputs, 1 network structure exported, hidden layer section
Points are 11.By largely testing and calculating test, above-mentioned network structure is selected, so that network structure becomes succinctly, to count
Calculation amount greatly reduces, and single-chip microcontroller can be easy to use to be calculated, greatly reduce equipment cost.
Preferably, the input vector matrix of the input layer of the BP neural network is defined as X (k), the input layer
Connection weight matrix to the hidden layer is defined as W1 (k), and the input layer is defined as B1 (k) to hidden layer connection threshold values,
Hidden layer vector matrix is defined as H (k), and the connection weight matrix of the hidden layer to the output layer is defined as W2 (k), the hidden layer
It is defined as B2 (k) to output layer connection threshold values, the output vector matrix of the output layer is defined as Y (k), and wherein k is indicated
Matrix order, the output calculation formula of the BP neural network are as follows:
Preferably, the connection weight adjustment moment matrix of the input layer to the hidden layer is defined as W1C(k), the hidden layer arrives
The output layer connection weight adjustment moment matrix geography is W2C(k), the formula of each connection weight adjustment is respectively as follows:
Wherein, the value range of a+b=1, a are 0.60-0.95.
It is furthermore preferred that a=0.9, b=0.1.
Preferably, further BP neural network is adjusted using current integration modification method.
Preferably, the current integration modification method using U-Ih slope of a curve as judge battery pack whether be full of or
The foundation of emptying.
Preferably, it is started to charge from battery pack emptying state, and current value is integrated, according to specific sampling rule
Data sample is obtained during the charging process, until battery pack reaches full state, after obtaining data sample, therefrom selected part
Sample carries out relearning for BP neural network, updates connection weight, and threshold value remains unchanged, and BP neural network is repaired in completion
Just.
Preferably, sample is selected in such a way that subregion is chosen.
Preferably, respectively in charging stage and discharge regime using SOC value as partitioning standards, two regions are all averagely divided
For 24 zonules, totally 48 zonules, only allow to select a sample in each zonule.
By using above-mentioned technical proposal, the present invention has following beneficial effect: changing by using smooth weighted calculation
Into the connection weight adjustment mode, it is suppressed that concussion caused by weighed value adjusting is conducive to the continuity of original learning outcome, and
Calculated result can be will affect to avoid the instantaneous mutation for acquiring signal in calculating process；Meanwhile BP is optimized by current integration
The model of neural network reduces the influence in ferric phosphate lithium cell group due to capacity attenuation to SOC estimation precision, so as to improve
The SOC estimation precision of ferric phosphate lithium cell group, and improve estimation efficiency.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description
Obviously, or recognize in embodiment or example through the invention.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, institute in being described below to the embodiment of the present invention
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without creative efforts, can also implement according to the present invention
The content of example and these attached drawings obtain other attached drawings.
Fig. 1 is the BP neural network schematic diagram of the embodiment of the present invention；
Fig. 2 is the ferric phosphate lithium cell group charging curve of the embodiment of the present invention；With
Fig. 3 is the ferric phosphate lithium cell group discharge curve of the embodiment of the present invention.
Related explanation, in Fig. 1, x1, x2, x3 are input, h1, h2, h3 ..., h11 be hidden layer；Y1 is output；In Fig. 2,
Horizontal axis is that charging current integrates Ah, and the longitudinal axis is cell voltage；In Fig. 3, horizontal axis is that discharge current integrates Ah, and the longitudinal axis is battery electricity
Pressure.
Specific embodiment
Below with reference to attached drawing, further detailed description is carried out to a specific embodiment of the invention.
The present invention is based on the SOC estimation method of the ferric phosphate lithium cell group of BP neural network the following steps are included:
Step 1 constructs BP neural network.
Ferric phosphate lithium cell group to be assessed is selected, the ferric phosphate lithium cell group is monitored with online real-time mode.Online
The related technical parameter of real-time measurement, comprising: battery voltage U(t), battery pack current I(t), battery pack internal resistance R(t) and, battery pack
Temperature T(t).The ferric phosphate lithium cell group rated capacity C0It can be directly obtained from related manufacturing firm, construct relation formula
(1), (2):
C(t)Indicate the capacity of ferric phosphate lithium cell group on-line real time monitoring.f 1Indicate the function of variable U, I, R, T.
Ferric phosphate lithium cell group is subjected to charge and discharge under the conditions of 0.1C, 0.2C, 0.3C, 0.4C, 0.5C respectively, and to reality
Test data and carry out after calculating analysis the internal resistance R(t it is found that same battery pack) close to a steady state value, therefore can estimating in SOC
Ignore in calculation.Same ferric phosphate lithium cell group rated capacity C0 is rating number, therefore above-mentioned relation formula (1) can be reduced to close
It is formula (4):
The SOC estimation process it is found that ferric phosphate lithium cell group is derived by above-mentioned theory, is exactly one according to on-line real time monitoring
Obtained battery voltage U(t), battery pack current I(t), battery pack temperature T(t) by series of computation and obtain its SOC ginseng
Several processes.
It is analyzed based on above-mentioned theory, establishes a kind of neural network of supervised learning, i.e., built according to above-mentioned functional relation
It founds various matrix functions and forms signal forward-propagating, output error passes through BP neural network mould of the hidden layer to the layer-by-layer anti-pass of input layer
Type.
As shown in Figure 1, being the BP neural network schematic diagram of the embodiment of the present invention.The network structure use 1 hidden layer, 3
A input and 1 output, wherein the number of hidden nodes is 11.
In the present embodiment, the input vector matrix of input layer is defined as X(k), input layer to hidden layer connection weight matrix
It is defined as W1(k), input layer is defined as B1(k to hidden layer connection threshold values), hidden layer vector matrix is defined as H(k), hidden layer to output
Layer connection weight matrix is defined as W2(k), hidden layer is defined as B2(k to output layer connection threshold values), output layer output vector matrix
It is defined as Y(k).K representing matrix order.
The output calculation formula of the BP neural network passes through relation formula (5), (6):
In the prior art, the adjustment of the connection weight of BP neural network generallys use gradient descent method, using this method be easy by
Gradually weaken the accumulation of history learning experience, convergence slowly, is easily trapped into the defect of local minimum, poor robustness.Due to BP mind
It is directly proportional to receptance function derivative through network connection weighed value adjusting amplitude.It is made in error curve smooth region since derivative is less than normal
Small at weighed value adjusting amplitude, error reduces slow；In the big region of error curve slope, since derivative is larger, weight tune is caused
Whole picture degree is big, and weighed value adjusting concussion is larger, it is difficult to restrain.In order to enhance the robustness of system, the present invention is to BP neural network
Connection weight set-up procedure is improved, and an inertia coeffeicent is inserted into, and improves connection weight adjustment using smooth weighted calculation,
When being attached weighed value adjusting calculating, increase the weight of former connection weight item, former connection weight item primary calculating knot before referring to
Fruit.In the present embodiment, input layer is defined as W to hidden layer connection weight adjustment moment matrix1C(k), hidden layer is to output layer connection weight
Value adjustment moment matrix is defined as W2C(k), improved calculation formula is relation formula (7), (8):
Wherein, the value range of a+b=1, a are 0.60-0.95, and the preferably value of 0.9, b is preferably 0.1.Above-mentioned assignment
Purpose is the weight in order to increase last calculated result, is avoided because this input fluctuation causes result big fluctuation occur.
Based on above-mentioned improved relation formula (7), (8), the weight as shared by original BP neural network connection weight is larger,
In the big region of error curve slope, even if connection weight adjustment amplitude is larger, but since the connection weight is in relation formula
Weight is small, therefore reduces the influence to last calculated result, it is suppressed that concussion caused by weighed value adjusting；Simultaneously as former
There is the weight of BP neural network connection weight big, be conducive to the continuity of original learning outcome, is not in showing for " being no sooner learned than it is forgotten "
As.Further, since the case where there are current breaks in the operational process of battery, this will lead to voltage and is also mutated therewith, therefore, increase
The weight for adding former connection weight item can also avoid the instantaneous mutation for acquiring signal in calculating process from will affect calculated result.
Step 2 corrects principle according to current integration and optimizes the BP neural network.During Neural Network Self-learning, by
More long using the time in battery itself the reason of decay, error can be bigger, it is therefore desirable to using the algorithm pair of a non-neural network
As a result it is modified, therefore step 2 is the regular amendment to step 1.It the so-called regularly time, can set according to the actual situation
It is fixed, for example can be one week or one month etc., it is contemplated that the life cycle of battery and itself attenuation characteristic, the present invention are preferably
One month or 30 days, efficiency and benefit can be taken into account.
In actual use due to ferric phosphate lithium cell group, with the increase for using the time, there are capacity attenuations
, especially there is the battery pack of large capacity charge and discharge and the frequent phenomenon of charge and discharge in problem, attenuation is especially serious in use.
Different ferric phosphate lithium cell groups, even same producer is with a batch of product, there is also individual differences.Relaxation phenomenon and
The presence of body difference, making one to be fixedly connected with weight and connect the BP neural network of threshold values will appear error after long-term use and gets over
Carry out bigger situation, new sample can only be obtained, and relearn by implementing on-line amending in battery pack periodic maintenance,
It just can guarantee that BP neural network can be always maintained at computational accuracy.
Real-time online amendment uses the current integration correcting mode directly related with SOC.As shown in Figures 2 and 3, by right
It is found that in charge or discharge by the region of completion, charge and discharge U-Ih curve is oblique for the research of ferric phosphate lithium cell group charge-discharge characteristic
Rate can become larger rapidly.
The U-Ih slope of curve is indicated with parameter K, (60s) there are relation formula (9) within a shorter period.
Current integration correcting mode using U-Ih slope of a curve as the foundation for judging whether battery pack is full of or is vented,
And started to charge from battery pack emptying state, and current value is integrated, during the charging process according to specific sampling rule
Data sample is obtained, the specific sampling is regular, refers to according under different size of current condition, when according to different fixations
Between be spaced and be sampled, until battery pack reaches full state.After obtaining data sample, therefrom selected part sample carries out BP
Neural network relearns, and updates connection weight and connection threshold values, completes the amendment to BP neural network.
The described current integration amendment specifically includes the following steps:
Step 21: by the integral in ferric phosphate lithium cell group maintenance to charging and discharging currents on a timeline, establishing new
Sample group, the sample in new samples group and step 1 here is different, new samples here be in cell decay by
It reacquires, constantly updates according to current integration mode.
Wherein, the SOC value of learning sample is obtained, it is thus necessary to determine that the upper and lower bound of the ferric phosphate lithium cell pool-size.
One slope upper limit parameter KH, KH are set and are set as 10.When ferric phosphate lithium cell group is under charged state and K
When > KH, battery capacity is upper limit SH, that is, corresponds to SOC=100, that is, corresponding SOC is exactly 100% when the upper limit；Work as phosphoric acid
Lithium iron battery group is under discharge condition and when K > KH, and ferric phosphate lithium cell pool-size is lower limit SL, that is, corresponds to SOC=0,
Namely lower limit when corresponding SOC be exactly 0%.
In the ferric phosphate lithium cell group maintenance stage, after electric discharge reaches ferric phosphate lithium cell pool-size lower limit SL first, stop
It discharges and stands 1 hour or more, ferric phosphate lithium cell group is transferred to charged state, record voltage U, the electric current I, temperature at the moment
T is respectively labeled as U0、I0、T0.Starting to carry out current integration on a timeline by starting point of SL, current accumulation amount is indicated with QI,
That is QI=I × t, enables C0.01=C0 × 0.01, and reach C in QI value0.01Integral multiple when, store corresponding voltage U, electric current I and
Current accumulation QI, temperature T, until K > KH, i.e., stopping when battery capacity reaches SH records the QI value at the moment, is labeled as
QI100.By the current accumulation in the charging stage obtain 100 groups or so real-time parameter (due to battery possible capacity decay,
When charging reaches actual battery pool-size upper limit SH, perhaps do not reach C0, therefore the not necessarily ideal feelings of sampled data
101 groups under condition).By the QI value in real-time recording parameters divided by QI100, new SOC value can be obtained, and then establish new packet
U containing voltage, electric current I, temperature T and SOC value sample set, constitute charging stage sample set.
It will be transferred to discharge condition after battery pack standing 1 hour or more again, identical method obtains about 100 when according to charging
The real-time parameter of group discharge regime, constitutes discharge regime sample set.
Step 22: subregion samples in sample group.
Since the samples selection requirement of BP neural network is as representative as possible, a cell is concentrated on to avoid sample
Domain, the BP neural network after causing training is accurate in the region, and big in remaining domain error, therefore chosen using subregion
Mode selects sample, is respectively all averagely divided into two regions using SOC value as partitioning standards in charging stage and discharge regime
24 zonules, totally 48 zonules, only allow to select a sample in each zonule.Due to charging stage sample and put
Electric stage total sample number is much larger than zonule number 48 close to 200 groups, therefore can guarantee alternative in each zonule
Sample size is greater than 1, is not in the case where respective regions sample lacks.Sample mode is chosen using subregion and obtains training use
Sample set, for BP neural network re -training, the neural network after training will more accurately calculate the SOC of entire battery pack
Value.
Better embodiment of the invention is described above, it is intended to so that spirit of the invention is more clear and convenient for managing
Solution, is not meant to limit the present invention, all within the spirits and principles of the present invention, modification, replacement, the improvement made should all
Within the protection scope that appended claims of the invention is summarized.
Claims (10)
1. a kind of ferric phosphate lithium cell SOC estimation method, using the prediction mould of BP neural network building ferric phosphate lithium cell group SOC
Type, it is characterised in that:
In the connection weight adjustment to the BP neural network, connection weight adjustment, insertion are improved using smooth weighted calculation
One inertia coeffeicent, increases the weight of former connection weight, and the original connection weight item refers to a preceding calculated result.
2. SOC estimation method according to claim 1, it is characterised in that:
The BP neural network uses Sigmoid using 1 hidden layer, 3 inputs and 1 network structure exported, the hidden layer
Transfer function, the number of hidden nodes are 11.
3. SOC estimation method according to claim 2, it is characterised in that:
The input vector matrix of the input layer of the BP neural network is defined as X (k), the input layer to the hidden layer
Connection weight matrix is defined as W1 (k), and the input layer is defined as B1 (k), hidden layer vector matrix to hidden layer connection threshold values
It is defined as H (k), the connection weight matrix of the hidden layer to the output layer is defined as W2 (k), the hidden layer to the output layer
Connection threshold values is defined as B2 (k), and the output vector matrix of the output layer is defined as Y (k), and wherein k representing matrix order, described
The output calculation formula of BP neural network are as follows:
4. SOC estimation method according to claim 3, it is characterised in that:
The connection weight adjustment moment matrix of the input layer to the hidden layer is defined as W1C(k), the hidden layer is to the output layer
It is W that connection weight, which adjusts moment matrix geography,2C(k), the formula of each connection weight adjustment is respectively as follows:
Wherein, the value range of a+b=1, a are 0.60-0.95.
5. SOC estimation method according to claim 4, it is characterised in that: a=0.9, b=0.1.
6. SOC estimation method according to claim 5, it is characterised in that:
Also BP neural network is adjusted using current integration modification method.
7. SOC estimation method according to claim 6, it is characterised in that:
The current integration modification method is using U-Ih slope of a curve as the foundation for judging whether battery pack is full of or is vented.
8. SOC estimation method according to claim 7, it is characterised in that:
It is started to charge from battery pack emptying state, and current value is integrated, according to specific sampling rule in charging process
Middle acquisition data sample, until battery pack reaches full state, after obtaining data sample, therefrom selected part sample carries out BP
Neural network relearns, and updates connection weight, completes the amendment to BP neural network.
9. SOC estimation method according to claim 8, it is characterised in that:
Sample is selected in such a way that subregion is chosen.
10. SOC estimation method according to claim 9, it is characterised in that:
Respectively in charging stage and discharge regime using SOC value as partitioning standards, two regions are all averagely divided into 24 cells
Domain, totally 48 zonules, only allow to select a sample in each zonule.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810953759.XA CN109031147B (en) | 2018-08-21 | 2018-08-21 | SOC estimation method of lithium iron phosphate battery pack |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810953759.XA CN109031147B (en) | 2018-08-21 | 2018-08-21 | SOC estimation method of lithium iron phosphate battery pack |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109031147A true CN109031147A (en) | 2018-12-18 |
| CN109031147B CN109031147B (en) | 2020-12-01 |
Family
ID=64626972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810953759.XA Active CN109031147B (en) | 2018-08-21 | 2018-08-21 | SOC estimation method of lithium iron phosphate battery pack |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109031147B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109738807A (en) * | 2019-01-03 | 2019-05-10 | 温州大学 | The method for estimating SOC based on the BP neural network after ant group algorithm optimization |
| CN110007239A (en) * | 2019-04-24 | 2019-07-12 | 中富通集团股份有限公司 | A kind of battery group prediction technique and system based on Neural Network Data mining algorithm |
| CN112379272A (en) * | 2020-11-16 | 2021-02-19 | 北京理工大学 | Lithium ion battery system SOC estimation method based on artificial intelligence |
| CN112731155A (en) * | 2020-12-08 | 2021-04-30 | 浙江南都电源动力股份有限公司 | Method for judging charge-discharge state of lithium iron phosphate battery |
| CN113391225A (en) * | 2021-05-19 | 2021-09-14 | 北京航空航天大学 | Lithium battery state-of-charge estimation method considering capacity degradation |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1890574A (en) * | 2003-12-18 | 2007-01-03 | 株式会社Lg化学 | Apparatus and method for estimating state of charge of battery using neural network |
| CN101964018A (en) * | 2010-08-26 | 2011-02-02 | 湘潭大学 | SOC estimation method of battery of electric vehicle |
| CN102385660A (en) * | 2010-09-03 | 2012-03-21 | 湘潭大学 | Evaluation method for system on chip (SOC) of charging station battery |
| CN102967831A (en) * | 2012-09-17 | 2013-03-13 | 常州大学 | On-line detection system and detection method of lead-acid storage battery performance |
| CN104360285A (en) * | 2014-11-28 | 2015-02-18 | 山东鲁能智能技术有限公司 | Battery capacity correction method based on improved ampere-hour integral method |
| CN104360286A (en) * | 2014-12-01 | 2015-02-18 | 重庆长安汽车股份有限公司 | Lithium ion battery charge state estimation modification method |
| KR101595956B1 (en) * | 2014-11-12 | 2016-02-22 | 충북대학교 산학협력단 | Apparatus and method for measuring state of charge(soc) for lithium ion battery |
| CN106443480A (en) * | 2016-11-04 | 2017-02-22 | 天津市捷威动力工业有限公司 | Lithium ion battery system SOC estimation method |
| CN106443453A (en) * | 2016-07-04 | 2017-02-22 | 陈逸涵 | Lithium battery SOC estimation method based on BP neural network |
| CN106716158A (en) * | 2014-06-11 | 2017-05-24 | 北京交通大学 | Method and device for estimating state of charge of battery |
| CN107991623A (en) * | 2017-11-27 | 2018-05-04 | 山东大学 | It is a kind of to consider temperature and the battery ampere-hour integration SOC methods of estimation of degree of aging |
-
2018
- 2018-08-21 CN CN201810953759.XA patent/CN109031147B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1890574A (en) * | 2003-12-18 | 2007-01-03 | 株式会社Lg化学 | Apparatus and method for estimating state of charge of battery using neural network |
| CN101964018A (en) * | 2010-08-26 | 2011-02-02 | 湘潭大学 | SOC estimation method of battery of electric vehicle |
| CN102385660A (en) * | 2010-09-03 | 2012-03-21 | 湘潭大学 | Evaluation method for system on chip (SOC) of charging station battery |
| CN102967831A (en) * | 2012-09-17 | 2013-03-13 | 常州大学 | On-line detection system and detection method of lead-acid storage battery performance |
| CN106716158A (en) * | 2014-06-11 | 2017-05-24 | 北京交通大学 | Method and device for estimating state of charge of battery |
| KR101595956B1 (en) * | 2014-11-12 | 2016-02-22 | 충북대학교 산학협력단 | Apparatus and method for measuring state of charge(soc) for lithium ion battery |
| CN104360285A (en) * | 2014-11-28 | 2015-02-18 | 山东鲁能智能技术有限公司 | Battery capacity correction method based on improved ampere-hour integral method |
| CN104360286A (en) * | 2014-12-01 | 2015-02-18 | 重庆长安汽车股份有限公司 | Lithium ion battery charge state estimation modification method |
| CN106443453A (en) * | 2016-07-04 | 2017-02-22 | 陈逸涵 | Lithium battery SOC estimation method based on BP neural network |
| CN106443480A (en) * | 2016-11-04 | 2017-02-22 | 天津市捷威动力工业有限公司 | Lithium ion battery system SOC estimation method |
| CN107991623A (en) * | 2017-11-27 | 2018-05-04 | 山东大学 | It is a kind of to consider temperature and the battery ampere-hour integration SOC methods of estimation of degree of aging |
Non-Patent Citations (2)
| Title |
|---|
| XUANJU DANG: "Open-Circuit Voltage-Based State of Charge Estimation of Lithium-ion Battery Using Dual Neural Network Fusion Battery Model", 《ELECTROCHIMICA ACTA》 * |
| 肖慧明: "基于改进BP神经网络的光伏组件状态监控研究与实现", 《信息系统工程》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109738807A (en) * | 2019-01-03 | 2019-05-10 | 温州大学 | The method for estimating SOC based on the BP neural network after ant group algorithm optimization |
| CN110007239A (en) * | 2019-04-24 | 2019-07-12 | 中富通集团股份有限公司 | A kind of battery group prediction technique and system based on Neural Network Data mining algorithm |
| CN110007239B (en) * | 2019-04-24 | 2021-01-19 | 中富通集团股份有限公司 | Storage battery pack prediction method and system based on neural network data mining algorithm |
| CN112379272A (en) * | 2020-11-16 | 2021-02-19 | 北京理工大学 | Lithium ion battery system SOC estimation method based on artificial intelligence |
| WO2022100229A1 (en) * | 2020-11-16 | 2022-05-19 | 北京理工大学 | Artificial-intelligence-based soc estimation method for lithium ion battery system |
| US11796597B2 (en) | 2020-11-16 | 2023-10-24 | Beijing Institute Of Technology | Method for estimating state of charge (SOC) of lithium-ion battery system based on artificial intelligence (AI) |
| CN112731155A (en) * | 2020-12-08 | 2021-04-30 | 浙江南都电源动力股份有限公司 | Method for judging charge-discharge state of lithium iron phosphate battery |
| CN112731155B (en) * | 2020-12-08 | 2022-02-22 | 浙江南都电源动力股份有限公司 | Method for judging charge-discharge state of lithium iron phosphate battery |
| CN113391225A (en) * | 2021-05-19 | 2021-09-14 | 北京航空航天大学 | Lithium battery state-of-charge estimation method considering capacity degradation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109031147B (en) | 2020-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109031147A (en) | A kind of SOC estimation method of ferric phosphate lithium cell group | |
| Zheng et al. | A novel capacity estimation method for lithium-ion batteries using fusion estimation of charging curve sections and discrete Arrhenius aging model | |
| CN107991623B (en) | Battery ampere-hour integral SOC estimation method considering temperature and aging degree | |
| CN110568361A (en) | Method for predicting health state of power battery | |
| CN103020445B (en) | A kind of SOC and SOH Forecasting Methodology of electric-vehicle-mounted ferric phosphate lithium cell | |
| Shrivastava et al. | Review on technological advancement of lithium-ion battery states estimation methods for electric vehicle applications | |
| CN108107372A (en) | Accumulator health status quantization method and system based on the estimation of SOC subregions | |
| CN105207243B (en) | A kind of battery energy management method for the forecast amendment of wind power plant realtime power | |
| CN110118937A (en) | The storage battery charge state edge calculations optimizing detection method of adaptive prediction model | |
| CN110474400A (en) | A kind of battery pack equilibrium method and device | |
| CN113255205B (en) | Life cycle cost and battery temperature optimization method based on electric automobile battery | |
| CN109752660B (en) | Battery state of charge estimation method without current sensor | |
| CN107340476A (en) | The electrical state monitoring system and electrical state monitoring method of battery | |
| CN113030761A (en) | Method and system for evaluating health state of battery of super-large-scale energy storage power station | |
| CN110362897A (en) | A kind of series-connected cell group multiple-objection optimization equalization methods | |
| CN115616425A (en) | Battery performance analysis method, electronic equipment and energy storage system | |
| CN116148670A (en) | Method and device for estimating service life of battery of electrochemical energy storage power station | |
| CN113917336A (en) | Lithium ion battery health state prediction method based on segment charging time and GRU | |
| CN113156316A (en) | Estimation algorithm for SOC of brine battery | |
| CN114636948A (en) | Energy storage system service life assessment method and device, electronic equipment and storage medium | |
| CN115327416A (en) | Lithium ion battery SOC estimation method based on group intelligent optimization and particle filtering | |
| CN116879822A (en) | SOC calibration method and related device | |
| CN117595439A (en) | Balanced control method and system for power grid energy storage battery management | |
| Geng et al. | SOC Prediction of power lithium battery using BP neural network theory based on keras | |
| CN113420444A (en) | Lithium ion battery SOC estimation method based on parameter online identification |
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 | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 411100 1 Xingye Avenue, Jiuhua Economic District, Xiangtan, Hunan Patentee after: Hunan anhuayuan Power Technology Co.,Ltd. Address before: 411100 1 Xingye Avenue, Jiuhua Economic District, Xiangtan, Hunan Patentee before: HUNAN XINGYE GREEN POWER TECHNOLOGY Co.,Ltd. |
|
| CP01 | Change in the name or title of a patent holder |