CN106501721A - A kind of lithium battery SOC estimation method based on biological evolution - Google Patents
A kind of lithium battery SOC estimation method based on biological evolution Download PDFInfo
- Publication number
- CN106501721A CN106501721A CN201610387551.7A CN201610387551A CN106501721A CN 106501721 A CN106501721 A CN 106501721A CN 201610387551 A CN201610387551 A CN 201610387551A CN 106501721 A CN106501721 A CN 106501721A
- Authority
- CN
- China
- Prior art keywords
- network
- lithium battery
- sample
- soc
- estimation method
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
- G01R31/3828—Arrangements for monitoring battery or accumulator variables, e.g. SoC using current integration
- G01R31/3832—Arrangements for monitoring battery or accumulator variables, e.g. SoC using current integration without measurement of battery voltage
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
Abstract
The present invention provides a kind of lithium battery SOC estimation method based on biological evolution, it is characterised in that：Including the network test after BP network neural models, BP network neural algorithms, the acquisition of network sample data, the pretreatment of sample data, BP neural network structure design, network estimation SOC tests, GA genetic algorithms, the frame design that optimizes, optimization, optimize weights and the threshold value of the BP neural network using the GA genetic algorithms, its step is：A, determine network topology structure, b, determine genetic algorithm parameter and coding, c, decoding obtain weights and threshold value, d, the output for calculating neutral net simultaneously obtain adaptive value, e, determine whether to meet end condition, terminate if meeting end condition, if being unsatisfactory for end condition, just return again to step c through the new colony of genetic algorithm selection, variation, intersection generation to continue to train, the invention greatly reduces the error of lithium battery SOC estimations.
Description
Technical field
The present invention relates to lithium battery SOC estimation method, in particular it relates to a kind of lithium battery SOC based on biological evolution estimates
Calculation method.
Background technology
With science and technology and industrial technology development, energy crisis and atmosphere polluting problem increasingly serious.According to Beijing
Environmental Protection Agency of the city survey data of 2013, vehicular emission account for the 31.1% of whole PM2.5 discharge capacitys.From economy, technology and ring
From the aspect of guarantor, the advantage of electric automobile is that zero-emission, low noise, high efficiency, Development of EV will become improvement air
One of pollution, important channel of solution energy crisis.Current electric automobile has that cruising time is short, and power performance deficiency etc. is asked
Topic, and the key technical problem of Development of Electric Vehicles is that the development of power battery technology, national governments are proposed accordingly
Technology strategy.Japan implemented electric power storage strategy in 2012, proposes lithium battery industry and will reach the 50% of world's share in the year two thousand twenty；
The U.S. it is also proposed the target that the year two thousand twenty has 14,000,000 electric automobiles；France and Germany are proposed corresponding electrokinetic cell skill
Art strategy；Power battery technology is also classified as the weight of Eleventh Five-Year Plan, " 12 " energy-conservation and new-energy automobile major project for China
Want research direction.
Lithium battery has the characteristic of stable security performance and height ratio capacity, therefore becomes the master of electric automobile power battery
Ingredient is wanted, used as the Important Parameters for embodying lithium battery interior state, accurately estimation SOC is base in lithium battery technology to SOC
Plinth and the key link.Accurately the SOC of estimation lithium battery, has important practical significance：SOC first directly reflects lithium battery
Ability to work, can effectively lift hybrid electric automobiles on energy pipe as the reference of design mixed power automobile control system
The performance of reason system, reasonable distribution electric automobile energy are used；Secondly, accurate SOC estimate can effectively pre- anti-overcharge put, most
Limits extend the life-span of battery, monitor the SOC of battery and carry out energy point according to the difference of each cell SOC in set of cells
Match somebody with somebody, it is ensured that the concordance of battery electric quantity, the use time that can extend after set of cells single charge.
SOC estimates that the inaccurate safety that can cause battery reduces, and shortens the service life of battery；Electricity can be affected simultaneously
The accurate calculating in motor-car cruising time, reduces the effective utilization of battery.Intrinsic characteristics of the SOC as lithium battery, it is difficult to pass through to pass
Sensor direct measurement, is affected by factors such as temperature, discharge current, self discharge, battery lifes in its running, so as to present
Complicated is non-linear, therefore seeks the important subject that a kind of accurate SOC estimation method is current power battery industry.
SOC is the important parameter for describing lithium battery running status, but SOC is with other battery parameters, such as temperature, internal resistance, from
The relation of the influence factors such as electric discharge, life-span assumes the non-linear of height so that the SOC of lithium battery estimates that difficulty is very big, at present electricity
The strategy of pond SOC estimations mainly has following several：Open circuit voltage method, Impedance Analysis, ampere-hour integration method, Kalman filtering method.
Open circuit voltage method is Japanese EV Project Department, and DENSO Corporation put forward, and work as electricity
After prolonged standing, the voltage at its two ends has relatively-stationary functional relationship with SOC in pond, therefore can be according to lithium electricity
Two pole tensions in pond are estimating SOC.Open circuit voltage method at battery operation initial stage and latter stage be effective, but but have apparent
Shortcoming, battery long-time must stand by the open-circuit voltage measurement of battery, the voltage that measures when voltage reaches to be stablized relatively
It is only effectively, and is subject to the factor impression such as temperature, battery life, therefore open circuit voltage method is relatively suitable for the feelings of battery standing
SOC is estimated under condition.
Impedance Analysis be lithium battery in use, the internal resistance of cell includes AC internal Resistance and DC internal resistance.Electric current is handed over
Flow impedance refers to the transmission function between cell voltage and electric current.For battery AC impedance, which is affected by temperature very big,
And also there is larger difference in the detection state of AC impedance；D.C. resistance refers to the variable quantity and curent change of cell voltage
The ratio of amount, the detection of D.C. resistance are affected by detection time, working condition, and detection time is too short, only ohmage
Can detect, detection time is long, internal resistance model will complicate；At the electric discharge initial stage, the relative change of D.C. resistance is big, to electric discharge
Later stage, internal resistance are just relatively stable.Therefore, larger using internal resistance method estimation SOC difficulty, it is generally not used for actual estimation.
Ampere-hour integration method is the SOC integrated to calculate battery by calculating the discharge current of lithium battery to certain time,
Assume that initial SOC is SOC0, then the SOC at current time isMiddle CN is the specified of battery
Capacity, I (τ) is the discharge current at τ moment, and η is coulombic efficiency, and ampere-hour integration method never needs to set up complicated SOC models, and
It is the mobility status for recording the energy content of battery, but this also results in estimation result and is subject to battery temperature, charge-discharge magnification, battery old
The impact of status consideration such as change and error occur, while current measurement there is also temporal cumulative error, other initial time
Self-discharge phenomenon can also affect the estimation result of SOC, therefore in actual applications, typically ampere-hour integration method is estimated as SOC
The reference value of calculation, is used in combination to improve estimation precision with additive method.
Kalman filtering algorithm be based on Minimum Mean Square Error principle, using SOC as system quantity of state, with the state side of system
Journey describes state migration procedure, describes the state correlation function between each moment by the transfer characteristic of state equation, so as to
Realize that SOC estimations [7], the core of Kalman filtering algorithm are to set up battery equivalent model, common model has equivalent circuit
And electrochemical model, Kalman filtering algorithm be used for battery SOC estimation subject matter be：It is higher to battery model requirement,
Set up accurate battery model and could obtain accurate SOC, and the requirement of order of accuarcy is higher, the complexity of model can also increase
Greatly；Kalman, filtering algorithm need to do substantial amounts of matrix operationss, and amount of calculation is very big.
Content of the invention
Based on the deficiency of the SOC estimation method of existing lithium battery, the purpose of the invention is that providing one kind solves
The problem existed due to lithium battery self character, using genetic algorithm come the weight threshold of Optimized BP Neural Network, drops significantly
The error of low lithium battery SOC estimations.
The SOC estimation method of existing lithium battery：Open circuit voltage method is subject to the factor impression such as temperature, battery life, therefore
Open circuit voltage method estimates SOC in the case of being relatively suitable for battery standing；For battery AC impedance, which is subject to temperature to Impedance Analysis
The impact of degree is very big, and the detection state in AC impedance also has larger point, the detection of D.C. resistance be subject to detection time,
The impact of working condition, detection time are too short, and only ohmage can be detected, and detection time is long, and internal resistance model will become
Complicated；At the electric discharge initial stage, the relative change of D.C. resistance is big, and to the electric discharge later stage, internal resistance is just relatively stable, is therefore estimated using internal resistance method
Calculate SOC difficulty larger, be generally not used for actual estimation；Peace times integration method can cause estimation result to be subject to battery temperature, discharge and recharge
The impact of the status considerations such as multiplying power, cell degradation and there is error, while current measurement there is also temporal cumulative error, separately
The self-discharge phenomenon of outer initial time can also affect the estimation result of SOC；Kalman filtering algorithm is used for the master of battery SOC estimation
Want problem to be higher to battery model requirement, set up accurate battery model and could obtain accurate SOC, and order of accuarcy
Requirement is higher, and the complexity of model can also increase；Kalman, filtering algorithm need to do substantial amounts of matrix operationss, and amount of calculation is very big.
The purpose of the present invention is：Solve the problems, such as due to lithium battery self character and, using genetic algorithm come excellent
Change the weight threshold of BP neural network, greatly reduce the error of lithium battery SOC estimations.
To achieve these goals, the invention provides a kind of lithium battery SOC estimation method based on biological evolution, which is special
Levy and be:Including BP network neural models, BP network neural algorithms, the acquisition of network sample data, the pretreatment of sample data, BP
Neural Network Structure Design, network estimation SOC tests, GA genetic algorithms, optimize frame design, optimize after network test, sharp
Optimize weights and the threshold value of the BP neural network with the GA genetic algorithms, its step is：A, determine network topology structure,
B, determine genetic algorithm parameter and coding, c, decoding obtain weights and threshold value, d, the output for calculating neutral net are simultaneously adapted to
Value, e, determines whether to meet end condition, terminates if meeting end condition, if being unsatisfactory for end condition, just through losing
Propagation algorithm is selected, is made a variation, intersection produces new colony and returns again to the continuation training of step c.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The multilayer neural network that the BP neural network of battery SOC evaluation method is made up of input layer, hidden layer, output layer, its
The most basic composition unit of each layer is exactly neuron, builds the BP neural network model according to the neuron, draws phase
The output layer transforming function transformation function that answers is linear function f (x)=x, and the hidden layer transforming function transformation function is unipolarity Sigmoid functionOr bipolarity Sigmoid function
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The BP neural network algorithm flow of battery SOC evaluation method is：(1) initialize, (2) input sample pair, calculate each layer and miss
Difference, (3) calculating network output errorIt is more to adopt root-mean-square error in practical application(4) each floor error number is calculated
(5) adjustment weights and threshold values(6) judge whether thus sample be all trained to, if sample
Count and be less than network training number of times, then sample counting and network training number of times all increase by 1, and return to step (2) otherwise goes to step
(7), (7) judge whether to meet error precision requirement and end condition, if meeting root-mean-square error less than minimum error or sample
This counting is counted less than maximum sample, and training is completed, and otherwise error sets to 0, and sample counting puts 1, is returned (2) and is continued training.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The network sample data of battery SOC evaluation method is obtained includes that the collection of network sample data, sample SOC are calculated.The net
The collection of network sample data can obtain voltage-SOC relations, calculated according to the sample SOC and obtained using the estimation of ampere-hour integration method
SOC be feasible, computing formula as standard SOC：The sample SOC is calculated temperature
Impact with discharge current multiplying power is considered into.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The pretreatment of the sample data of battery SOC evaluation method includes the random alignment of the normalized of data, data, described
Data normalization to interval [- 1,1], is used the mapminmax that MATLAB workboxes are carried by the normalized of data
Function, the random alignment of the data are training sample data to be carried out randomly ordered, and similar sample can be avoided excessively to concentrate.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The BP neural network structure design of battery SOC evaluation method includes the design of input and output layer, hidden layer design, and design is described
The number of plies of BP network neural structures, show that implicit number of layers, the hidden layer are designed by adopting heuristic, formula：J=
log2n,According to above-mentioned formula, hidden layer node number scope is determined, adopted using MATLAB workboxes
Transforming function transformation function of the tansig functions as hidden layer node, while carried out as Learning Algorithms building network using LM algorithms
Sound out, the average MSE for taking multiple training is criterion, can obtain the relation of network performance and hidden layer node, sets hidden
Number containing node layer.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The number of plies of the BP network neurals structure according to above-mentioned determination of battery SOC evaluation method, the implicit number of layers, described hidden
The network estimation SOC tests are carried out containing several sections of points.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The genetic operator of battery SOC evaluation method includes selection opertor, crossover operator, mutation operator, the GA genetic algorithms root
According to the screening process that individual fitness value comes under simulating natural environment, by selecting, intersecting, the operation such as variation produce new
Body, this process will cause whole population to develop towards the direction for being conducive to best fit approximation solution, and the chromosome coding is adopted
It is floating-point encoding.The fitness function should meet monodrome, non-negative, the condition such as continuous, and the fitness function should also be use up
May be simple, reduce amount of calculation.
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The selection opertor of battery SOC evaluation method is fitness ratio method, and the selection opertor also includes uniform ranking method, most
Excellent conversation strategy, random league matches are selected, exclusion is selected etc..The crossover operator is the arithmetic crossover of the floating-point encoding, institute
Stating floating-point encoding also includes that discrete crossover etc., the crossover operator apply also for binary-coded single-point intersection, multiple spot and hand over
Fork, uniform crossover etc., the mutation operator are non-uniform mutation modes, the conventional mutation operator also include uniform variation,
Boundary mutation, Gaussian mutation etc..
Improve as lithium battery SOC estimation method of the present invention based on biological evolution, lithium of the present invention based on biological evolution
The optimization frame design of battery SOC evaluation method be BP neural network algorithm described in the genetic algorithm optimization be will be described
Genetic algorithm ability of searching optimum is strong and the advantage of fast convergence rate and combining for Neural Network Based Nonlinear capability of fitting, with
Reach the purpose for overcoming the slow box of neutral net convergence rate to be easily trapped into the minimum shortcoming of local error.The end of the genetic algorithm
Only condition precision or reaches maximum evolution number of times for needed for reaching.Finally compare the estimation effect and mistake after optimizing and before optimization
Difference compares.
Compared with prior art, the invention has the advantages that：Solve due to lithium battery self character and deposit
Problem, using genetic algorithm come the weight threshold of Optimized BP Neural Network, greatly reduce the mistake of lithium battery SOC estimations
Difference.
Description of the drawings
Fig. 1 is single neuron model of the present invention based on the lithium battery SOC estimation method preferred implementation of biological evolution
Fig. 2 is three-layer neural network of the present invention based on the lithium battery SOC estimation method preferred implementation of biological evolution
Structural representation
Fig. 3 is propagation algorithm flow chart of the present invention based on the lithium battery SOC estimation method preferred implementation of biological evolution
Fig. 4 is genetic algorithm optimization of the present invention based on the lithium battery SOC estimation method preferred implementation of biological evolution
BP neural network flow chart
Specific embodiment
1-4, is carried out based on the lithium battery SOC estimation method of biological evolution to the present invention as described below below in conjunction with the accompanying drawings.
Following related description is carried out to BP network neurals algorithm and GA genetic algorithms first.
BP network neural algorithms are also called error backpropagation algorithm, are neutral nets most widely used up to now, BP
Neural network structure is simple, autgmentability is strong, is widely used in the fields such as function approximation, pattern recognition, classification, data compression.
The multilayer neural network that BP neural network is made up of input layer, hidden layer, output layer, the most basic composition unit of its each layer
It is exactly neuron, basic neuron models Fig. 1, wherein X=(x1,x2...) be neuron input, y is the defeated of neuron
Go out, W=(w1,w2...) it is adjustable input weights, B=b is the threshold value of neuron, and f (net) is the excitation function of neuron.Defeated
Enter signal and neuron is entered by being input into weights connection (weighted sum), obtain exporting y by excitation function.
GA genetic algorithms be the gene genetic principle of the theory of evolution based on Darvin and Mendel develop a kind of excellent
Change algorithm.By problem representation to be solved into " chromosome ", initial population is individual in the range of problem disaggregation to be solved to genetic algorithm
Body, population are made up of the encoded individuality of certain amount, according to survival of the fittest and the rule of the survival of the fittest, are developed by generation
The approximate solution that becomes better and better.Genetic algorithm according to individual fitness value come the screening process under simulating natural environment, by choosing
Select, intersect, the operation such as variation produces new individuality, this process will cause whole population towards the side for being conducive to best fit approximation solution
To development.
The invention provides a kind of lithium battery SOC estimation method based on biological evolution, it is characterised in that：Which includes BP nets
Network neural model, BP network neural algorithms, the acquisition of network sample data, the pretreatment of sample data, BP neural network structure set
Meter, network estimation SOC tests, GA genetic algorithms, optimize frame design, optimize after network test, using the GA heredity calculation
Optimizing weights and the threshold value of the BP neural network, its step is method：A, determine network topology structure, b, determine genetic algorithm
Parameter and coding, c, decoding obtain weights and threshold value, and d, the output for calculating neutral net simultaneously obtain adaptive value, e, determine whether full
Sufficient end condition, terminates if meeting end condition, if being unsatisfactory for end condition, just through genetic algorithm select, variation,
Intersect the new colony of generation and return again to the continuation training of step c.
In the present embodiment, the present invention based on the BP neural network of the lithium battery SOC estimation method of biological evolution is
The multilayer neural network being made up of input layer, hidden layer, output layer, the most basic composition unit of its each layer is exactly neuron,
The BP neural network model is built according to the neuron, show that the output layer transforming function transformation function is linear function f accordingly
X ()=x, the hidden layer transforming function transformation function are unipolarity Sigmoid functionsOr bipolarity Sigmoid function.
In the present embodiment, the present invention is calculated based on the BP neural network of the lithium battery SOC estimation method of biological evolution
Method flow process is：(1) initialize, (2) input sample pair, calculate each layer error, (3) calculating network output errorIt is more to adopt root-mean-square error in practical application(4) calculate
Each layer error signal (5) adjustment weights and threshold values(6) judge whether thus sample be all trained to, if sample counting is less than network training time
Number, then sample counting and network training number of times all increase by 1, and return to step (2) otherwise goes to step (7), (7) judge whether full
Sufficient error precision is required and end condition, if meeting root-mean-square error is less than maximum sample less than minimum error or sample counting
Count, training is completed, and otherwise error sets to 0, sample counting puts 1, return (2) and continue training.
In the present embodiment, the network sample data of the present invention based on the lithium battery SOC estimation method of biological evolution
Obtaining includes that the collection of network sample data, sample SOC are calculated.The network sample data collection can obtain voltage-SOC passes
System, it is feasible as standard SOC to calculate the SOC obtained using the estimation of ampere-hour integration method according to the sample SOC, calculates public
Formula：The sample SOC is calculated and is considered into the impact of temperature and discharge current multiplying power.
In the present embodiment, the present invention is based on the pre- of the sample data of the lithium battery SOC estimation method of biological evolution
Processing includes normalized, the random alignment of data of data, and the normalized of the data is by data normalization to area
Between [- 1,1], use the mapminmax functions that MATLAB workboxes are carried, the random alignment of the data is to training sample
Notebook data carries out randomly ordered, and similar sample can be avoided excessively to concentrate.
In the present embodiment, the BP neural network knot of the present invention based on the lithium battery SOC estimation method of biological evolution
Structure design includes the design of input and output layer, hidden layer design, designs the number of plies of the BP network neurals structure, draws the implicit number of plies
Mesh, the hidden layer are designed by adopting heuristic, formula：J=log2n,According to above-mentioned formula, determine
Hidden layer node number scope, adopts tansig functions as the transforming function transformation function of hidden layer node by the use of MATLAB workboxes, same
Shi Caiyong LM algorithms are soundd out as Learning Algorithms building network, and the average MSE for taking multiple training is criterion, can
To obtain the relation of network performance and hidden layer node, node in hidden layer is set.
In the present embodiment, institute according to above-mentioned determination of the present invention based on the lithium battery SOC estimation method of biological evolution
State the number of plies of BP network neural structures, the implicit number of layers, the implicant nodes and carry out the network estimation SOC and survey
Examination.
A kind of genetic operator of the lithium battery SOC estimation method based on biological evolution of the present invention include selection opertor,
Crossover operator, mutation operator.The GA genetic algorithms according to individual fitness value come the screening process under simulating natural environment,
By selecting, intersecting, the operation such as variation produce new individuality, this process will cause whole population towards being conducive to best fit approximation
Develop in the direction of solution.The chromosome coding uses floating-point encoding.The fitness function should meet monodrome, non-negative,
The condition such as continuous, the fitness function should also be as simple as possible, reduce amount of calculation.
In the present embodiment, the selection opertor of the present invention based on the lithium battery SOC estimation method of biological evolution is suitable
Response ratio method.The selection opertor also includes that uniform ranking method, optimum maintaining strategy, random league matches are selected, exclusion is selected
Deng.The crossover operator is the arithmetic crossover of the floating-point encoding, and the floating-point encoding also includes discrete crossover etc., described
Crossover operator applies also for binary-coded single-point intersection, multiple-spot detection, uniform crossover etc..The mutation operator is non-homogeneous
Variation mode, the conventional mutation operator also include uniform variation, boundary mutation, Gaussian mutation etc..
In the present embodiment, the optimization frame design of the present invention based on the lithium battery SOC estimation method of biological evolution
Be BP neural network algorithm described in the genetic algorithm optimization be by strong for the genetic algorithm ability of searching optimum and convergence rate
Combining for fast advantage and Neural Network Based Nonlinear capability of fitting, overcomes the slow box of neutral net convergence rate easy to reach
It is absorbed in the purpose of the minimum shortcoming of local error.The end condition of the genetic algorithm precision or reaches maximum for needed for reaching
Evolution number of times.Finally compare the estimation effect after optimizing and before optimization and application condition.
In the present embodiment, the present invention is calculated based on the BP network neurals of the lithium battery SOC estimation method of biological evolution
Method builds the BP neural network model first, three layers of BP neural network structure chart Fig. 2, and three layers of BP neural network includes described defeated
Enter layer, the hidden layer, the output layer, X=(x1,x2,…xn) be network input matrix, xnFor input feature value, W
=(w0,w1,…wn) it is connection weight matrix between input layer and hidden layer, wnIt is weight vector, wherein w0It is threshold vector,
V=(v0,v1,…vn) it is connection weight between hidden layer and output layer, vnIt is weight vector, wherein v0It is threshold vector, y is
Output vector, for the output layer, has following relation：
yk=f (netk) k=1,2,3 ... l (3-1)
For the hidden layer, there is following relation:
yj=f (netj) j=1,2,3 ... m (3-3)
In above-mentioned relation, the output layer transforming function transformation function is linear function:
F (x)=x (3-5)
The hidden layer transforming function transformation function is unipolarity Sigmoid function:
Or bipolarity Sigmoid function:
In the present embodiment, the present invention is calculated based on the BP network neurals of the lithium battery SOC estimation method of biological evolution
Method flow process is as follows：
(1) initialize
Initial value is assigned to weight vector W and V, sample counting p and network training number of times q are initialized as 1, error E sets to 0, study
Rate η is the decimal between 0～1, and network expects precision EminIt is set to positive decimal.
(2) input sample pair, calculates each layer error
It is input into current sample Xp,dp, and output valve Y and the O of each layer is calculated according to formula (3-5) and formula (3-7)；
(3) calculating network output error
For P to training sample, network is for the error of sample pCan be by the defeated of whole samples
Go out error and seek geometrical mean, as total error, in practical application be more using root-mean-square error you.
(4) each layer error signal is calculated
Calculation errorWith
Adjustment weights and threshold value
According to regulating error weights W next time, V:
(5) judge whether that all of sample is all trained to
If p<P, p, q increase by 1, and return to step (2) otherwise goes to step (7).
(6) judge whether to meet error precision requirement and end condition
If meeting ERME<EminOr p<Pmax, training completes, and otherwise E sets to 0, and p puts 1, and return to step (2) continues training.
In the present embodiment, the network sample of the present invention based on the lithium battery SOC estimation method of biological evolution
Data acquisition：
1.1.1 network sample data is gathered
Training sample needs preferably to cover whole parameter area.
1.1.2 sample SOC is calculated
Previously mentioned, ampere-hour integration method is simple, reliable, and being suitable for all of battery carries out the estimation of SOC, generally as the mark of SOC
Standard, is used in combination with additive method.But there is problems with actual applications in ampere-hour integration method：
(1) easily affected by battery temperature, current fluctuation；
(2) error that battery measurement is present can build up；
(3) great amount of samples data are needed；
The testing scheme that the present invention is adopted is constant-current discharge under constant temperature, therefore the fluctuation very little of temperature and electric current, can be with
Ignore；The current measurement precision of battery test system is very high, cumulative error very little；The interval sampling time of battery is 30s, always
Data volume reach more than 6500, meet the requirement of sample data quantity, therefore will be obtained using the estimation of ampere-hour integration method
SOC is feasible, computing formula as standard SOC：
Wherein I is current discharge current, η efficiency for charge-discharges, CI,TIt is current flow total electricity that battery can be released with a temperature of
Amount, due to lithium battery in the case of temperature, discharge current multiplying power difference the whole electricity that can release be different, therefore this
In the impact of temperature and discharge current multiplying power is considered into.
In the present embodiment, the present invention is based on the pre- of the sample data of the lithium battery SOC estimation method of biological evolution
Process includes two steps：
(1) normalized of data
Sample data has different dimensions, generally before network training is normalized training sample, generally adopts
Mode be to interval [- 1,1] by data normalization, use the mapminmax functions that MATLAB workboxes are carried.
(2) data is randomly ordered
Due to sample data, constant-current discharge is measured at a certain temperature, and it is change that therefore can there is voltage, but electric current
All it is constant situation with temperature, if the order according to sample data script carries out network training, electric current and temperature excessively collect
In, network training learning process can be caused vibration occur so that convergence time is elongated or situation about not restraining occurs.Therefore need
Training sample data are carried out randomly ordered, it is to avoid similar sample is excessively concentrated.
In the present embodiment, the BP neural network knot of the present invention based on the lithium battery SOC estimation method of biological evolution
Structure design is as follows：
Input and output layer is designed
|input paramete is chosen as voltage, discharge current, temperature, and therefore input layer has 3 nodes；Output parameter is SOC, therefore
Output layer has 1 node, and the transforming function transformation function of output node is purelin (purely linear function).
Hidden layer is designed
The BP neural network structure of design is 3 layers, and it is 1 therefore to imply number of layers.Node in hidden layer purpose is determined due to net
The research of network structural theory is not perfect enough, can instruct without specific theory, therefore adopt heuristic according to practical situation.?
Before exploration, first node in hidden layer purpose approximate range is obtained according to some empirical equations：
J=log2n(3-13)
N is input layer number.
M is output layer nodes, and n is input layer number, and α is the integer between 0～10, and the present invention adopts formula (3-14), really
Determining hidden layer node number scope is, adopts tansig functions as the change exchange the letters of hidden layer node by the use of MATLAB workboxes
Number, while being soundd out as Learning Algorithms building network using Leverberg-Marquardt (LM) algorithm, is taken repeatedly
The average MSE (Minimum Mean Square Error) of training is criterion, can obtain the relation of network performance and hidden layer node.
In the present embodiment, the present invention is surveyed based on the network estimation SOC of the lithium battery SOC estimation method of biological evolution
Examination：
According to above-mentioned analysis, corresponding data separate MATLAB programmed environments test network estimation effect is obtained.BP neural network
Estimation effect preferably, actual value and estimated value are basically identical.For more directly perceived and accurate analytical error, by actual value and
Estimated value does absolute error.
In the present embodiment, the present invention mainly elaborates BP based on the above of lithium battery SOC estimation method of biological evolution
The correlation theory of neutral net, and the model analysiss of BP neural network have been carried out for SOC estimations, briefly describe whole BP god
Through the process that network is set up, the effect of SOC is analyzed to be estimated to BP neural network finally.
In the present embodiment, the GA genetic algorithm of the present invention based on the lithium battery SOC estimation method of biological evolution is first
Genetic algorithm flow chart is first built, and Fig. 3, the present invention produce initial population 2 using floating-point encoding 1, calculate individual adaptation degree 3,
Whether meet termination condition 4, terminate 9 if meeting condition, carry out selecting 5 if being unsatisfactory for condition, intersect 6, variation 7,
Population of future generation 8 is obtained, individual adaptation degree is returned again to and is calculated 3 continuation calculating.
In the present embodiment, the chromosome coding choosing of the present invention based on the lithium battery SOC estimation method of biological evolution
Floating-point encoding 1 is selected, and floating-point encoding 1 is that individual gene is represented with the floating number in particular range, individual volume
Code length is equal to the number of variable.Adopted herein is the method for floating-point encoding 1.
In the present embodiment, the present invention based on the fitness function of the lithium battery SOC estimation method of biological evolution is
Be used for simulating adaptability of the population at individual to whole natural environment, corresponding in genetic algorithm individual in disaggregation scope most
It is close to the degree of optimal solution.Fitness function is the criterion for evaluating individual adaptability in the environment, and fitness function should be
Non-negative, sometimes asked is maximum, and sometimes asked is minima.Design fitness function should meet monodrome, and non-negative connects
The condition such as continuous；Also need to as far as possible simply, to reach the purpose for reducing amount of calculation；Complexity will be according to particular problem
Fixed.
In the present embodiment, the present invention is included based on the genetic operator of the lithium battery SOC estimation method of biological evolution：
(1) 5 are selected
It is in colony, select adaptable individuality to participate in new Swarm Evolution process to select 5, and this process is also referred to as again
System.It is the basis for intersecting and making a variation to select 5 operations, and its main purpose is lost in order to avoid adaptable gene, improves complete
Office's convergence and computational efficiency.Method used by the present embodiment is fitness ratio method, in other embodiments also using suitable
The methods such as response ratio method (roulette method), uniform ranking method, optimum maintaining strategy, random league matches are selected, exclusion selection,.
For assuming certain individual i, its fitness is f to the operating process of fitness ratio methodi, Population Size is N, then the individual quilt
The probability that chooses is
(2) 6 are intersected
The operation of intersection 6 in genetic algorithm is that two chromosomes form new dye by restructuring during mimic biology natural evolution
The process of colour solid, intersects 6 operations and can constantly produce new individuality, increase population diversity, expand Search Range, in heredity calculation
Method extension solution room is played an important role.The present embodiment uses floating-point encoding 1, thus the present embodiment using be calculate
Art intersects 6, and in other embodiments also using binary-coded single-point intersection, multiple-spot detection, uniform crossover etc., floating number is compiled
Discrete crossover of code 1 etc..
Arithmetic crossover 6 is to select individual gene to carry out linear combination to produce the process of new chromosome, it is assumed that have selected two
Individual XA、XB, by the process that two individualities carry out intersecting 6 computings it is:
X′A=α XB+(1-α)XA(4-2)
X′B=α XA+(1-a)XB(4-3)
If α is exactly uniform arithmetic crossover 6 for constant, if α is variable, referred to as nonuniform arithmetical crossover.
(3) variation 7
Variation 7 operation be simulation natural evolution during gene undergo mutation, so as to produce the process of new chromosome.Variation 7
Operation is exactly to realize replacing some of individual chromosome coded strings gene under 7 probability of definitive variation, and it can be effectively
Expand the Local Search scope of genetic algorithm, prevent algorithm Premature Convergence.The present embodiment mainly uses non-uniform mutation side
Formula, in other embodiments also using uniform variation, boundary mutation, non-uniform mutation, Gaussian mutation etc.：
The process of inhomogeneous boundary layer is:
Wherein γ is random 0 or 1, vkIt is the gene according to certain probability selection, v 'kFor the gene for newly producing, function Δ
The concrete formula of (t, y) can be：
Wherein r is the random number of [0,1], and T is maximum algebraically, and t is current algebraically, and b is the systematic parameter of non-uniformity, typically takes
It is worth for 2～5.
In the present embodiment, the optimization frame design of the present invention based on the lithium battery SOC estimation method of biological evolution
As follows：
Genetic algorithm optimization BP neural network algorithm be by the advantage of strong for genetic algorithm ability of searching optimum and fast convergence rate with
Combining for Neural Network Based Nonlinear capability of fitting, is missed with reaching to overcome neutral net convergence rate slow and be easily trapped into local
The purpose of minimum shortcoming, such as Fig. 4 is differed from, flow process is：Determine network topology structure 10, determine genetic algorithm parameter and coding 11,
Decoding obtains weights and threshold value 12, calculates the output of neutral net and obtains fitness value 13, if meets condition 14, if full
Sufficient flow process terminates 16, and if being unsatisfactory for, genetic algorithm is selected, and intersects, variation produce new colony return decoding obtain weights and
Threshold value 12 continues flow process.
The absolute error maximum of the SOC value that fitness function is estimated with network using actual soc-value, i.e.,：
E=max (Y 'out-Yout) (4-6)
Wherein Y 'outBe network estimation SOC value, YoutActual SOC value, the end condition of algorithm for needed for reaching precision or
Person reaches maximum evolution number of times.
In the present embodiment, the present invention finally does the network after optimizing based on the lithium battery SOC estimation method of biological evolution
Test, compares the estimation effect after optimizing and before optimization and application condition.
Compared with prior art, the invention has the advantages that：Solve due to lithium battery self character and deposit
Problem, using genetic algorithm come the weight threshold of Optimized BP Neural Network, greatly reduce the error of SOC estimations.
Although the present invention is described by reference to specific embodiment, those skilled in the art are by reading
After stating description, obvious modification can be made to the present invention and is modified, and without prejudice to the intent of the present invention and essence, this
These modifications and modification are included within the scope of the claims by invention intentionally.
Claims (10)
1. a kind of lithium battery SOC estimation method based on biological evolution, it is characterised in that：Which includes BP network neural models, BP
Network neural algorithm, the acquisition of network sample data, the pretreatment of sample data, BP neural network structure design, network estimation SOC
Test, GA genetic algorithms, optimize frame design, optimize after network test, optimize the BP using the GA genetic algorithms
The weights of neutral net and threshold value, its step is：A, determine network topology structure, b, determine genetic algorithm parameter and coding, c,
Decoding obtains weights and threshold value, and d, the output for calculating neutral net simultaneously obtain adaptive value, e, determines whether to meet end condition, such as
Fruit meets end condition and just terminates, if being unsatisfactory for end condition, just produces through genetic algorithm selection, variation, intersection new
Colony returns again to step c and continues training.
2. lithium battery SOC estimation method according to claim 1 based on biological evolution, it is characterised in that：The BP nerves
The multilayer neural network that network is made up of input layer, hidden layer, output layer, the most basic composition unit of its each layer are exactly god
Through unit, the BP neural network model is built according to the neuron, show that the output layer transforming function transformation function is linear accordingly
Function f (x)=x, the hidden layer transforming function transformation function are unipolarity Sigmoid functionsOr bipolarity
Sigmoid functions
3. lithium battery SOC estimation method according to claim 2 based on biological evolution, it is characterised in that：The BP nerves
Network algorithm flow process is：(1) initialize, (2) input sample pair, calculate each layer error, (3) calculating network output errorIt is more to adopt root-mean-square error in practical application(4) calculate
Each layer error signal (5) adjustment weights and threshold values(6) judge whether thus sample be all trained to, if sample counting is less than network training time
Number, then sample counting and network training number of times all increase by 1, and return to step (2) otherwise goes to step (7), (7) judge whether full
Sufficient error precision is required and end condition, if meeting root-mean-square error is less than maximum sample less than minimum error or sample counting
Count, training is completed, and otherwise error sets to 0, sample counting puts 1, return (2) and continue training.
4. lithium battery SOC estimation method according to claim 3 based on biological evolution, it is characterised in that：The network sample
Notebook data is obtained includes that the collection of network sample data, sample SOC are calculated.Network sample data collection can obtain voltage-
SOC relations, it is feasible, meter as standard SOC to calculate the SOC obtained using the estimation of ampere-hour integration method according to the sample SOC
Calculate formula：The sample SOC is calculated and is considered the impact of temperature and discharge current multiplying power
Enter.
5. lithium battery SOC estimation method according to claim 4 based on biological evolution, it is characterised in that：The sample number
According to pretreatment include the random alignment of the normalized of data, data, the normalized of the data is by data normalizing
Change to interval [- 1,1], use the mapminmax functions that MATLAB workboxes are carried, the random alignment of the data is right
Training sample data carry out randomly ordered, and similar sample can be avoided excessively to concentrate.
6. lithium battery SOC estimation method according to claim 5 based on biological evolution, it is characterised in that：The BP nerves
Network structure design includes the design of input and output layer, hidden layer design, designs the number of plies of the BP network neurals structure, draws hidden
Contain number of layers, the hidden layer is designed by adopting heuristic, formula：J=log2n,According to above-mentioned public affairs
Formula, determines hidden layer node number scope, adopts tansig functions as the conversion of hidden layer node by the use of MATLAB workboxes
Function, while being soundd out as Learning Algorithms building network using LM algorithms, the average MSE for taking multiple training is judgement
Standard, can obtain the relation of network performance and hidden layer node, set node in hidden layer.
7. lithium battery SOC estimation method according to claim 6 based on biological evolution, it is characterised in that：According to above-mentioned true
The number of plies of the fixed BP network neurals structure, the implicit number of layers, the implicant nodes carry out the network estimation
SOC is tested.
8. lithium battery SOC estimation method according to claim 7 based on biological evolution, it is characterised in that：The heredity is calculated
Attached bag includes selection opertor, crossover operator, mutation operator.The GA genetic algorithms simulate nature ring according to individual fitness value
Screening process under border, produces new individuality by operations such as selection, intersection, variations, and this process will cause whole population court
And be conducive to the direction of best fit approximation solution to develop.The chromosome coding uses floating-point encoding.The fitness function
Monodrome, non-negative, the condition such as continuous should be met, the fitness function should also be as simple as possible, reduces amount of calculation.
9. lithium battery SOC estimation method according to claim 8 based on biological evolution, it is characterised in that：The selection is calculated
Son is fitness ratio method.The selection opertor also includes that uniform ranking method, optimum maintaining strategy, random league matches are selected, arranged
Squeeze selection etc..The crossover operator is the arithmetic crossover of the floating-point encoding, and the floating-point encoding also includes discrete crossover
Deng the crossover operator applies also for binary-coded single-point intersection, multiple-spot detection, uniform crossover etc..The mutation operator
It is non-uniform mutation mode, the conventional mutation operator also includes uniform variation, boundary mutation, Gaussian mutation etc..
10. lithium battery SOC estimation method according to claim 9 based on biological evolution, it is characterised in that：The optimization structure
It is that BP neural network algorithm described in the genetic algorithm optimization is that the genetic algorithm ability of searching optimum is strong and receives to set up meter
Combining for fireballing advantage and Neural Network Based Nonlinear capability of fitting is held back, overcomes neutral net convergence rate slow to reach
Box is easily trapped into the purpose of the minimum shortcoming of local error, and the end condition of the genetic algorithm precision or is reached for needed for reaching
To maximum evolution number of times, finally compare the estimation effect after optimizing and before optimization and application condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610387551.7A CN106501721A (en) | 2016-06-03 | 2016-06-03 | A kind of lithium battery SOC estimation method based on biological evolution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610387551.7A CN106501721A (en) | 2016-06-03 | 2016-06-03 | A kind of lithium battery SOC estimation method based on biological evolution |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106501721A true CN106501721A (en) | 2017-03-15 |
Family
ID=58287248
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610387551.7A Pending CN106501721A (en) | 2016-06-03 | 2016-06-03 | A kind of lithium battery SOC estimation method based on biological evolution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106501721A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107367693A (en) * | 2017-07-07 | 2017-11-21 | 淮阴工学院 | SOC detection system for power battery of electric vehicle |
| CN107748936A (en) * | 2017-11-03 | 2018-03-02 | 国网江苏省电力公司信息通信分公司 | Based on the improved BP neural network life of storage battery prediction algorithm of genetic algorithm |
| CN108537337A (en) * | 2018-04-04 | 2018-09-14 | 中航锂电技术研究院有限公司 | Lithium ion battery SOC prediction techniques based on optimization depth belief network |
| CN108572324A (en) * | 2018-04-13 | 2018-09-25 | 芜湖职业技术学院 | Battery SOC estimation device based on immune algorithm Optimized BP Neural Network |
| CN109033288A (en) * | 2018-07-13 | 2018-12-18 | 电子科技大学 | A kind of intelligent terminal security level classification method based on BP neural network |
| CN109738808A (en) * | 2019-01-03 | 2019-05-10 | 温州大学 | The method of SOC is estimated based on the BP neural network after Simulated Anneal Algorithm Optimize |
| CN110082682A (en) * | 2019-03-12 | 2019-08-02 | 浙江大学 | A kind of lithium battery charge state estimation method |
| CN110208715A (en) * | 2019-07-02 | 2019-09-06 | 上海禾他汽车科技有限公司 | A kind of method and automobile batteries management system measuring automobile batteries packet state of charge |
| CN110232432A (en) * | 2018-03-05 | 2019-09-13 | 重庆邮电大学 | A kind of lithium battery group SOC prediction technique based on artificial life model |
| CN110298467A (en) * | 2018-03-23 | 2019-10-01 | 中国科学院微电子研究所 | A kind of estimation method and system of remaining capacity |
| CN110492185A (en) * | 2019-03-27 | 2019-11-22 | 华中科技大学 | A kind of lithium battery group equalization methods and system |
| CN112611972A (en) * | 2020-11-30 | 2021-04-06 | 上海理工大学 | Method for estimating SOC (state of charge) of lithium battery under condition of low-frequency sampling data |
| CN112858917A (en) * | 2021-01-15 | 2021-05-28 | 哈尔滨工业大学（威海） | Battery system multi-fault diagnosis method based on genetic algorithm optimization neural network |
| CN113219358A (en) * | 2021-04-29 | 2021-08-06 | 东软睿驰汽车技术（沈阳）有限公司 | Battery pack health state calculation method and system and electronic equipment |
| CN114799415A (en) * | 2022-03-11 | 2022-07-29 | 南京航空航天大学 | Arc additive remanufacturing welding parameter-welding bead size positive and negative neural network prediction model |
| CN116359762A (en) * | 2023-04-27 | 2023-06-30 | 北京玖行智研交通科技有限公司 | Battery state of charge estimation method based on deep learning and network compression |
| CN116774091A (en) * | 2023-08-24 | 2023-09-19 | 南京市计量监督检测院 | High-precision power battery pack SOH online measurement system and method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003197275A (en) * | 2001-12-27 | 2003-07-11 | Panasonic Ev Energy Co Ltd | Estimating method of polarized voltage of secondary battery, estimating method and device of residual capacity of secondary battery, as well as battery pack system |
| 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 |
| CN101198922A (en) * | 2005-06-13 | 2008-06-11 | Lg化学株式会社 | Apparatus and method for testing state of charge in battery |
| CN102324582A (en) * | 2011-08-12 | 2012-01-18 | 重庆东电通信技术有限公司 | Intelligent maintenance device of multifunctional lead-acid battery and capacity prediction method |
-
2016
- 2016-06-03 CN CN201610387551.7A patent/CN106501721A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003197275A (en) * | 2001-12-27 | 2003-07-11 | Panasonic Ev Energy Co Ltd | Estimating method of polarized voltage of secondary battery, estimating method and device of residual capacity of secondary battery, as well as battery pack system |
| 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 |
| CN101198922A (en) * | 2005-06-13 | 2008-06-11 | Lg化学株式会社 | Apparatus and method for testing state of charge in battery |
| CN102324582A (en) * | 2011-08-12 | 2012-01-18 | 重庆东电通信技术有限公司 | Intelligent maintenance device of multifunctional lead-acid battery and capacity prediction method |
Non-Patent Citations (3)
| Title |
|---|
| 万永凯: "混合动力电动汽车锂离子电池SOC估算方法研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑（月刊）》 * |
| 余胜威: "BP神经网络的分析流程", 《MATLAB优化算法案例分析与应用》 * |
| 冯楠: "基于ARM9的嵌入式系统在电动汽车电源管理系统中的应用", 《中国优秀硕士学位论文全文数据库 信息科技辑（月刊）》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107367693B (en) * | 2017-07-07 | 2018-05-29 | 淮阴工学院 | SOC detection system for power battery of electric vehicle |
| CN107367693A (en) * | 2017-07-07 | 2017-11-21 | 淮阴工学院 | SOC detection system for power battery of electric vehicle |
| CN107748936A (en) * | 2017-11-03 | 2018-03-02 | 国网江苏省电力公司信息通信分公司 | Based on the improved BP neural network life of storage battery prediction algorithm of genetic algorithm |
| CN110232432A (en) * | 2018-03-05 | 2019-09-13 | 重庆邮电大学 | A kind of lithium battery group SOC prediction technique based on artificial life model |
| CN110232432B (en) * | 2018-03-05 | 2022-09-20 | 重庆邮电大学 | Lithium battery pack SOC prediction method based on artificial life model |
| CN110298467A (en) * | 2018-03-23 | 2019-10-01 | 中国科学院微电子研究所 | A kind of estimation method and system of remaining capacity |
| CN108537337A (en) * | 2018-04-04 | 2018-09-14 | 中航锂电技术研究院有限公司 | Lithium ion battery SOC prediction techniques based on optimization depth belief network |
| CN108572324A (en) * | 2018-04-13 | 2018-09-25 | 芜湖职业技术学院 | Battery SOC estimation device based on immune algorithm Optimized BP Neural Network |
| CN109033288A (en) * | 2018-07-13 | 2018-12-18 | 电子科技大学 | A kind of intelligent terminal security level classification method based on BP neural network |
| CN109738808A (en) * | 2019-01-03 | 2019-05-10 | 温州大学 | The method of SOC is estimated based on the BP neural network after Simulated Anneal Algorithm Optimize |
| CN110082682B (en) * | 2019-03-12 | 2020-04-24 | 浙江大学 | Lithium battery state of charge estimation method |
| CN110082682A (en) * | 2019-03-12 | 2019-08-02 | 浙江大学 | A kind of lithium battery charge state estimation method |
| CN110492185A (en) * | 2019-03-27 | 2019-11-22 | 华中科技大学 | A kind of lithium battery group equalization methods and system |
| CN110492185B (en) * | 2019-03-27 | 2020-10-02 | 华中科技大学 | Lithium battery pack equalization method and system |
| CN110208715A (en) * | 2019-07-02 | 2019-09-06 | 上海禾他汽车科技有限公司 | A kind of method and automobile batteries management system measuring automobile batteries packet state of charge |
| CN112611972A (en) * | 2020-11-30 | 2021-04-06 | 上海理工大学 | Method for estimating SOC (state of charge) of lithium battery under condition of low-frequency sampling data |
| CN112858917A (en) * | 2021-01-15 | 2021-05-28 | 哈尔滨工业大学（威海） | Battery system multi-fault diagnosis method based on genetic algorithm optimization neural network |
| CN113219358A (en) * | 2021-04-29 | 2021-08-06 | 东软睿驰汽车技术（沈阳）有限公司 | Battery pack health state calculation method and system and electronic equipment |
| CN114799415A (en) * | 2022-03-11 | 2022-07-29 | 南京航空航天大学 | Arc additive remanufacturing welding parameter-welding bead size positive and negative neural network prediction model |
| CN116359762A (en) * | 2023-04-27 | 2023-06-30 | 北京玖行智研交通科技有限公司 | Battery state of charge estimation method based on deep learning and network compression |
| CN116359762B (en) * | 2023-04-27 | 2024-05-07 | 北京玖行智研交通科技有限公司 | Battery state of charge estimation method based on deep learning and network compression |
| CN116774091A (en) * | 2023-08-24 | 2023-09-19 | 南京市计量监督检测院 | High-precision power battery pack SOH online measurement system and method thereof |
| CN116774091B (en) * | 2023-08-24 | 2023-10-17 | 南京市计量监督检测院 | High-precision power battery pack SOH online measurement system and method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106501721A (en) | A kind of lithium battery SOC estimation method based on biological evolution | |
| CN106599520B (en) | A kind of air pollutant concentration forecasting procedure based on LSTM-RNN models | |
| CN102496069B (en) | Cable multimode safe operation evaluation method based on fuzzy analytic hierarchy process (FAHP) | |
| CN104849671B (en) | A kind of battery capacity detecting system based on combination neural net | |
| CN108491970A (en) | A kind of Predict Model of Air Pollutant Density based on RBF neural | |
| CN105717460B (en) | A kind of power battery SOC methods of estimation and system based on nonlinear observer | |
| CN112557907A (en) | SOC estimation method of electric vehicle lithium ion battery based on GRU-RNN | |
| CN109118020A (en) | A kind of subway station energy consumption short term prediction method and its forecasting system | |
| CN109324291A (en) | A kind of prediction technique for Proton Exchange Membrane Fuel Cells life prediction | |
| CN111366848A (en) | Battery health state prediction method based on PSO-ELM algorithm | |
| CN111709186A (en) | Integrated estimation method for health state of retired power lithium battery | |
| CN112986830A (en) | Lithium battery capacity estimation method based on convolution time memory neural network | |
| CN109871622A (en) | A kind of low-voltage platform area line loss calculation method and system based on deep learning | |
| CN106991504A (en) | Building energy consumption Forecasting Methodology, system and building based on metering separate time series | |
| CN108804800A (en) | Lithium ion battery SOC on-line prediction methods based on echo state network | |
| CN106327357A (en) | Load identification method based on improved probabilistic neural network | |
| CN109856545B (en) | The battery group residual capacity detection method and system of solar telephone | |
| CN112881914B (en) | Lithium battery health state prediction method | |
| CN108009585A (en) | Lead-acid battery health status Forecasting Methodology based on local message fusion | |
| CN109507598A (en) | The lithium battery SOC prediction technique of the LM-BP neural network of Bayesian regularization | |
| CN113917334A (en) | Battery health state estimation method based on evolution LSTM self-encoder | |
| CN110232432B (en) | Lithium battery pack SOC prediction method based on artificial life model | |
| CN111999651B (en) | Power performance test method of lithium ion battery for pure electric automobile | |
| Niu et al. | State of charge prediction study of vanadium redox-flow battery with BP neural network | |
| CN109738808B (en) | Method for estimating SOC (State of Charge) based on BP (Back propagation) neural network optimized by simulated annealing algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170315 |