CN108287317A - Power of battery prediction model generation method and system, power forecasting method and system - Google Patents
Power of battery prediction model generation method and system, power forecasting method and system Download PDFInfo
- Publication number
- CN108287317A CN108287317A CN201810068746.4A CN201810068746A CN108287317A CN 108287317 A CN108287317 A CN 108287317A CN 201810068746 A CN201810068746 A CN 201810068746A CN 108287317 A CN108287317 A CN 108287317A
- Authority
- CN
- China
- Prior art keywords
- power
- battery
- data
- prediction
- module
- 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
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]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
Abstract
The invention discloses a kind of power of battery prediction model generation method and system, power forecasting method and system, generation method includes the following steps:S1, obtain training sample and concentrate the power characterization data of each sample battery, power characterization data includes the discharge power data of each charge and discharge of the sample battery;S2, power characterization data is pre-processed to generate corresponding power data matrix;S3, using sample battery described in the training sample set power data matrix to convolutional neural networks algorithm carry out model training, to generate the prediction model of the power of battery.The present invention carries out the power data matrix of input using convolutional neural networks algorithm dimensionality reduction and the feature extraction of multilayer, can be derived that the prediction model of the power of battery.Power prediction can be carried out to the especially retired power battery of power battery to be predicted, can be realized to the Life cycle prediction of battery to be predicted and real time monitoring according to the result of power prediction using the prediction model.
Description
Technical field
The present invention relates to power battery field, more particularly to a kind of power of battery prediction model generation method and system, work(
Rate prediction technique and system.
Background technology
With the fast development and popularization of new-energy automobile, the demand of the power battery used in new-energy automobile also increasingly increases
It is long.It is limited to the technical merit of current power battery, when the loss of power battery reaches a certain level, power battery power supply is special
Property is unable to reach electric vehicle supply standard, must just be eliminated into retired power battery.Retired power battery has because of it
Standby storing up electricity and charging and discharging capabilities, it is not very high field, such as energy storage relatively to battery behavior requirement to be often used in some
Power station, charging pile etc. realize that the secondary use to power battery, the secondary use of this retired battery are called battery step profit
With.
Retired automobile power cell is certain compared to having with new battery in terms of battery behavior due to being secondary use
Loss, and reliability and safety are also difference with new battery.Therefore, it is real when actually using retired battery
Existing retired battery will also ensure the safety and reliability of battery while maximally utilizing, at this time just it is necessary to battery
Bulk properties data are monitored in real time and are shown, for user or the real-time multidate information of electrolytic cell of maintenance personal.
In addition to this, the service life of retired power battery is also often limited, and the aging of retired battery is also to have
Very strong uncertainty.The primary accidental emergency case of retired battery, for example battery is scrapped or the situations such as deep battery discharge,
The inconvenience used will be brought to user, safeguarded to producer and increased cost.Therefore, special according to inside battery characteristic information and outside
Property information predicted for the service life of battery and safety, realize relatively reliable to battery, safer management.It can not only
Retired power battery multidate information guidance is provided the user with, and is provided periodically to retired power battery producer maintenance personnel
Maintenance reference information, this comfort that user uses battery, maintenance personnel directiveness be undoubtedly of great significance.
Power battery itself has individual difference, even with model with the brand new cells of batch, its consistency there is also
Height problem.Along with the use of various complex working conditions, life prediction and health control to battery bring parameter, Wu Fajian
Single obtains the rule of its decay.In power battery, retired battery itself is a kind of service life and the electricity that safety is lost
Therefore pond is shown and is monitored for information such as the retired power of battery, service life and health datas and be more necessary.Battery
Service life is often for users invisible or unpredictalbe, paroxysmal battery failures or is scrapped to user or factory
It is undoubtedly stubborn problem for family therefore to be predicted and estimated for the service life and safety of battery, to user or producer
For be all that there is very strong directive significance.
Invention content
The technical problem to be solved by the present invention is in order to overcome in the prior art to the especially retired power electric of power battery
Without the defect of corresponding effective prediction scheme, provide a kind of can carry out the power characteristic of power battery the Decay Law in pond
Prediction and the higher power of battery prediction model generation method and system of accuracy, power forecasting method and system.
The present invention is to solve above-mentioned technical problem by following technical proposals:
The present invention provides a kind of power of battery prediction model generation methods, and feature is, includes the following steps:
S1, obtain training sample and concentrate the power characterization data of each sample battery, the power characterization data includes institute
State the discharge power data of each charge and discharge of sample battery;
S2, the power characterization data is pre-processed to generate corresponding power data matrix;
S3, using sample battery described in the training sample set power data matrix to convolutional neural networks algorithm into
Row model training, to generate the prediction model of the power of battery.
In the present solution, the discharge power data are the data for including sequential variable, power data matrix is also when including
The data matrix of sequence variable.
In the present solution, the dimensionality reduction and feature of multilayer are carried out to the power data matrix of input using convolutional neural networks algorithm
Extraction, can be derived that the Mechanism Model i.e. prediction model of the power of battery.Electricity is carried out using convolutional neural networks algorithm in this programme
The training of the prediction model of pond power, it can be understood as found for single battery and show most similar individual packets, and root with it
According to the historical data of close individual, the decay situation of this individual, while the catastrophe in studying history data are inferred, form machine
Simulation efficiently and accurately predicts the Decay Law of power battery to realize.
Preferably, step S2Include the following steps:
S21, to the power characterization data carry out data extraction to obtain periodic component and trend component;
S22, the trend component is pre-processed to generate the corresponding power data matrix.
In the present solution, step S3Middle generation be trend component in the power of battery prediction model.
In the present solution, the discharge power data of battery are divided into trend component and periodic component.Periodic component is for electric discharge
It is a translational component for embodying its intrinsic speciality for this variable of power data.That is this component will not both decay or not
Meeting deformation, but in itself nor one can use the function of time that monomial is expressed.Conventionally, can with one group it is sinusoidal and
Or it integrates to indicate periodic component.
In the present solution, in battery status prediction, discharge power data are resolved into periodic component and trend component, will only be become
Mode input of the gesture component as convolutional neural networks algorithm improves convolutional neural networks to realize advance dimensionality reduction
Operation efficiency.
Preferably, the discharge power data include discharge voltage data and discharge current data, the power characteristic number
According to further including at least one of access times, frequency of use, charge and discharge operating mode.
In the present solution, the access times of power battery, frequency of use, charge and discharge operating mode and whether having unusual service condition such as mistake
It filled in the iterative process put when waiting variables that can participate in convolutional neural networks algorithm progress model training as coefficient.
In the present solution, at least two-dimensional data that discharge voltage data and discharge current data form is used for convolutional Neural net
Network algorithm carries out model training, enables to the accuracy higher of the prediction model generated.
Preferably, step S2Described in pretreatment include data invalid processing, data normalization handle and data matrixing
Processing.
In the present solution, the power characterization data by the sample battery that acquisition obtains can not be directly as convolutional Neural
The input information of network algorithm needs to handle by data invalid, after data normalization processing and data matrixing input processing
It reuses.When input data amount is very big, algorithm will face the promotion of calculation amount, for convenience of calculating by by power characterization data
Matrixing can improve the computational efficiency of algorithm.
Preferably, the periodic component is corresponding to discharge power curve of the corresponding sample battery under completely new state
Data.
In the present solution, for one piece of specific battery, the discharge power under its completely new state can be determined by empirical value
Curve and by this group of Data Identification be periodic component.
Preferably, step S1It is before further comprising the steps of:
S0, obtain total sample set, total sample set is sampled to obtain sample set using sample rate r, by institute
It states sample set and is set as the training sample set;
Step S3It is further comprising the steps of later:
S4, concentrate the power data matrix for the sample battery for having neither part nor lot in model training to described using the total sample
Prediction model is verified, if output error is more than default error, is adjusted the parameter of model training, is executed step S3。
In the present solution, the prediction model that the sample battery verification for having neither part nor lot in model training generates is concentrated with total sample, when defeated
When going out error in the reasonable scope, forecasting mechanism is completed;Otherwise, by adjusting model parameter duplication model training step, so that
The prediction model that must be generated is more accurate.
The present invention also provides a kind of power of battery prediction models to generate system, and feature is, including data acquisition mould
Block, preprocessing module and model generation module;
The data acquisition module concentrates the power characterization data of each sample battery for obtaining training sample, described
Power characterization data includes the discharge power data of each charge and discharge of the sample battery;
The preprocessing module, for being pre-processed to the power characterization data to generate corresponding power data square
Battle array;
The model generation module, for the power data matrix pair using sample battery described in the training sample set
Convolutional neural networks algorithm carries out model training, to generate the prediction model of the power of battery.
Preferably, the preprocessing module includes data extraction module and component preprocessing module;
The data extraction module, for carrying out data extraction to the power characterization data to obtain periodic component and become
Gesture component;
The component preprocessing module, for being pre-processed to the trend component to generate the corresponding power number
According to matrix.
Preferably, the discharge power data include discharge voltage data and discharge current data, the power characteristic number
According to further including at least one of access times, frequency of use, charge and discharge operating mode.
Preferably, pretreatment described in the preprocessing module includes data invalid processing, data normalization processing sum number
It is handled according to matrixing.
Preferably, the periodic component is corresponding to discharge power curve of the corresponding sample battery under completely new state
Data.
Preferably, it further includes sampling module and authentication module that the power of battery prediction model, which generates system,;
The sampling module uses r pairs of sample rate for obtaining total sample set before data acquisition module execution
Total sample set is sampled to obtain sample set, sets the sample set to the training sample set;
The authentication module, for having neither part nor lot in mould using total sample concentration after model generation module execution
The power data matrix of the sample battery of type training verifies the prediction model, if output error is more than default miss
Difference then adjusts the parameter of model training, calls the model generation module.
The present invention also provides a kind of power of battery prediction techniques, and feature is, includes the following steps:
T, using power of battery prediction model generation method above-mentioned generate the prediction model to battery to be predicted into
Row power prediction, to obtain the result of the power prediction of the battery to be predicted.
In the present solution, can be realized to the prediction of the Life cycle of battery to be predicted according to the result of power prediction and in real time
Monitoring.
Preferably, step T includes the following steps:
T1, obtain the power characterization data of the battery to be predicted;
T2, the power characterization data is pre-processed to generate corresponding power data matrix;
T3, by the power data Input matrix to the prediction model carry out power prediction, it is described to be predicted to obtain
The result of the power prediction of battery.
The present invention also provides a kind of power of battery forecasting system, feature is, including prediction module and electricity above-mentioned
Pond power prediction model generates system;
The prediction module, it is described to obtain for carrying out power prediction to battery to be predicted using the prediction model
The result of the power prediction of battery to be predicted.
Preferably, the prediction module include battery data acquisition module to be predicted, battery preprocessing module to be predicted and
Predict execution module;
The battery data acquisition module to be predicted, the power characterization data for obtaining the battery to be predicted;
The battery preprocessing module to be predicted, it is corresponding to generate for being pre-processed to the power characterization data
Power data matrix;
The prediction execution module, it is pre- for the power data Input matrix to the prediction model to be carried out power
It surveys, to obtain the result of the power prediction of the battery to be predicted.
The positive effect of the present invention is that:Power of battery prediction model generation method and system provided by the invention,
Power forecasting method and system carry out the power data matrix of input using convolutional neural networks algorithm dimensionality reduction and the spy of multilayer
Sign extraction, can be derived that the prediction model of the power of battery.It can be to power battery to be predicted especially using the prediction model
Retired power battery carries out power prediction, and the Life cycle to battery to be predicted can be realized according to the result of power prediction
Prediction and real time monitoring.Further, discharge power data are resolved into periodic component and trend component in the present invention, will only become
Mode input of the gesture component as convolutional neural networks algorithm improves convolutional neural networks to realize advance dimensionality reduction
Operation efficiency.
Description of the drawings
Fig. 1 is the flow chart of the power of battery prediction model generation method of the embodiment of the present invention 1.
Fig. 2 is that the power of battery prediction model of the embodiment of the present invention 2 generates the module diagram of system.
Fig. 3 is the flow chart of the power of battery prediction technique of the embodiment of the present invention 3.
Fig. 4 is the module diagram of the power of battery forecasting system of the embodiment of the present invention 4.
Fig. 5 is the flow chart that the present invention is applied to cell health state prediction.
Fig. 6 is the flow chart using battery predictive method when the present invention.
Specific implementation mode
It is further illustrated the present invention below by the mode of embodiment, but does not therefore limit the present invention to the reality
It applies among a range.
Embodiment 1
As shown in Figure 1, present embodiments providing a kind of power of battery prediction model generation method, include the following steps:
Step S0, total sample set is obtained, total sample set is sampled to obtain sample set using sample rate r,
Set the sample set to training sample set;
Step S1, the power characterization data that the training sample concentrates each sample battery, the power characteristic number are obtained
According to the access times including the sample battery, frequency of use, charge and discharge operating mode, the discharge power data of each charge and discharge, institute
It includes discharge voltage data and discharge current data to state discharge power data;
Step S2, data extraction is carried out to obtain periodic component and trend component, the week to the power characterization data
Phase component is the data corresponding to discharge power curve of the corresponding sample battery under completely new state;
Step S3, the trend component is pre-processed to generate the corresponding power data matrix, the pre- place
Reason includes data invalid processing, data normalization processing and the processing of data matrixing;
Step S4, convolutional neural networks are calculated using the power data matrix of sample battery described in the training sample set
Method carries out model training, to generate the prediction model of the power of battery;
Step S5, the power data matrix pair for the sample battery for having neither part nor lot in model training is concentrated using total sample
The prediction model is verified, and judges whether output error is more than default error, if so then execute step S6, if otherwise flow
Terminate;
Step S6, the parameter of model training is adjusted, step S4 is executed.
In the present embodiment, the pretreatment of step S3 can also extract exchange sequence with step S2 data, i.e., first to data into
Row pretreatment, then again carries out pretreated data the extraction of periodic component and trend component.
In the present embodiment, discharge power data, power data matrix, discharge voltage data and discharge current data are packet
Include the data of sequential variable.At least two-dimensional data that discharge voltage data and discharge current data form is used for convolutional Neural net
Network algorithm carries out model training, enables to accuracy of the prediction model generated than directly using one-dimensional discharge power data
It is high.
It is inevitable trend that power battery is gradually decayed with access times, meanwhile, the discharge power of battery has its week
Phase property.In the present embodiment, the discharge power data of battery are divided into trend component and periodic component.Periodic component is for the work(that discharges
It is a translational component for embodying its intrinsic speciality for this variable of rate data.I.e. will not both decay will not for this component
Deformation, but in itself nor one can use the function of time that monomial is expressed.Conventionally, can with one group it is sinusoidal and/or
It integrates to indicate periodic component.In battery status prediction, only using trend component as the mode input of convolutional neural networks algorithm,
To realize advance dimensionality reduction, the operation efficiency of convolutional neural networks is improved.
In the present embodiment, dimensionality reduction and the spy of multilayer are carried out to the power data matrix of input using convolutional neural networks algorithm
Sign extraction, can be derived that the Mechanism Model i.e. prediction model of the power of battery.The power of battery is carried out using convolutional neural networks algorithm
Prediction model training, it can be understood as found for single battery and with it show most similar individual packets, and according to close
The historical data of individual, infers the decay situation of this individual, while the catastrophe in studying history data, Forming Mechanism mould
Type efficiently and accurately predicts the Decay Law of power battery to realize.
Embodiment 2
As shown in Fig. 2, present embodiments providing a kind of power of battery prediction model generation system, including sampling module 0, number
According to acquisition module 1, preprocessing module 2, model generation module 3 and authentication module 4.
The sampling module 0 samples to obtain total sample set using sample rate r for obtaining total sample set
To sample set, it sets the sample set to training sample set.
The data acquisition module 1 concentrates the power characterization data of each sample battery for obtaining the training sample,
The power characterization data includes access times, frequency of use, charge and discharge operating mode and each charge and discharge of the sample battery
Discharge power data, the discharge power data include discharge voltage data and discharge current data.
The preprocessing module 2, for being pre-processed to the power characterization data to generate corresponding power data
Matrix.
The model generation module 3, for the power data matrix using sample battery described in the training sample set
Model training is carried out to convolutional neural networks algorithm, to generate the prediction model of the power of battery.
The authentication module 4, for being had neither part nor lot in using total sample concentration after the model generation module 3 execution
The power data matrix of the sample battery of model training verifies the prediction model, is preset if output error is more than
Error then adjusts the parameter of model training, calls the model generation module 3.
In the present embodiment, the preprocessing module 2 includes data extraction module 201 and component preprocessing module 202.It is described
Data extraction module 201 is extracted for carrying out data to the power characterization data to obtain periodic component and trend component, institute
It is the data corresponding to discharge power curve of the corresponding sample battery under completely new state to state periodic component.The component is located in advance
Module 202 is managed, for being pre-processed to the trend component to generate the corresponding power data matrix, the pretreatment
Including data invalid processing, data normalization processing and the processing of data matrixing.
In the present embodiment, the discharge power data of battery are divided into trend component and periodic component.In battery status prediction,
Only convolutional Neural is improved to realize advance dimensionality reduction using trend component as the mode input of convolutional neural networks algorithm
The operation efficiency of network.In the present embodiment, multilayer is carried out to the power data matrix of input using convolutional neural networks algorithm
Dimensionality reduction and feature extraction can be derived that the Mechanism Model i.e. prediction model of the power of battery.It can be treated using the prediction model pre-
The especially retired power battery of the power battery of survey carries out power prediction, can be realized to be predicted according to the result of power prediction
Battery Life cycle prediction and real time monitoring.
Embodiment 3
As shown in figure 3, present embodiments providing a kind of power of battery prediction technique, include the following steps:
T1, the power characterization data for obtaining battery to be predicted;
T2, the power characterization data is pre-processed to generate corresponding power data matrix;
T3, by the power data Input matrix to embodiment 1 power of battery prediction model generation method generate
The prediction model carries out power prediction, to obtain the result of the power prediction of the battery to be predicted.
In the present embodiment, the Decay Law of battery to be predicted can be obtained according to the result of power prediction, it is pre- by power
The result of survey realizes the life prediction of battery, further to carry out health control.
Power of battery prediction technique provided in this embodiment can realize to the prediction of the Life cycle of battery to be predicted and
Real time monitoring.
Embodiment 4
As shown in figure 4, a kind of power of battery forecasting system is present embodiments provided, including in prediction module 5 and embodiment 2
Power of battery prediction model generate system 6;
The prediction module 5, it is described to obtain for carrying out power prediction to battery to be predicted using the prediction model
The result of the power prediction of battery to be predicted.The prediction module 5 includes battery data acquisition module 501 to be predicted, to be predicted
Battery preprocessing module 502 and prediction execution module 503;
The battery data acquisition module 501 to be predicted, the power characterization data for obtaining the battery to be predicted;
The battery preprocessing module 502 to be predicted, for being pre-processed the power characterization data with generation pair
The power data matrix answered;
The prediction execution module 503, for the power data Input matrix to the prediction model to be carried out power
Prediction, to obtain the result of the power prediction of the battery to be predicted.
In the present embodiment, the especially retired power battery of power battery to be predicted can be carried out using prediction module 5
Power prediction can be realized according to the result of power prediction to the Life cycle prediction of battery to be predicted and real time monitoring.
Continue with the technical solution by specific example, further illustrated the present invention and technique effect.
It is predicted applied to the Life cycle of power battery when the present invention is embodied, specific technical solution is as follows:
Entire prediction process is divided into two broad aspect of Forming Mechanism and application system.Wherein Forming Mechanism is divided into adopts for data
Collection verifies four steps with processing, modelling, model training, model, refers to Fig. 5.Power prediction and the method for health control
It is embodied in modelling, model training, refers to Fig. 6.
I Forming Mechanisms
I.1 data acquisition and procession
I.1.1 measurement data
According to industry experience, judge that the common counter of battery performance has:Charging voltage, charging current, discharge voltage, electric discharge
Electric current, the internal resistance of cell, SOC (state of charge, the state-of-charge of battery) etc..General measure means are i.e. by retired electricity
Pond carries out one or several complete charge and discharge process, and records corresponding empirical value charging voltage Ucharge(t), it fills
Electric current Icharge(t), discharge voltage Udischarge(t), discharge current Idischarge(t), internal resistance Rinner(t), SOC.
For terminal user, the variable of the use feeling of visual influence battery is discharge power, can be by discharge voltage
Udischarge(t), discharge current Idischarge(t) it is calculated, when the discharge power of battery decay to critical value or drastically decays,
Show that battery life expires, can not be continuing with.Therefore after calculating discharge power using discharge voltage, discharge current, as
Mode input is most straightforward approach.But it is not at the same time, to eliminate remaining measurement data not having battery life and health
Relevant property.
I.1.2 data are recorded
Even same battery, electric discharge shows and remaining life uses before with it number, frequency and charge and discharge
Operating mode, intentionally whether often operating mode (super-charge super-discharge) is related.Therefore these variables can be participated in as coefficient in iteration.
I.1.3 data processing
The data of acquisition can not need invalid data delete processing, data to return directly as the input information of algorithm
One changes processing and the processing of data Input matrix.
◆ data invalid processing
For input data xi, wherein xi∈{x1,x2…N, ifThen to xiIt is updated:
◆ data normalization processing
To prevent the disappearance or diverging of gradient in algorithm, needed by normalized, x before the data input of acquisitioni∈
{x1,2…N, the data after normalizationFor
◆ input data matrixization processing
When input data amount is very big, algorithm, which will face calculation amount, to be promoted, for convenience of calculating here by that will input number
Computational efficiency is improved according to matrixing.
I.2 modelling
It is inevitable trend that power battery is gradually decayed with access times, meanwhile, the discharge power of battery has its week
Phase property.Therefore the discharge power of battery is divided into trend component and periodic component, i.e. P (t)=P in the present inventiona(t)+Pb(t), wherein
PaFor trend component, PbFor periodic component.
I.2.1 the extraction of periodic component
Periodic component is a translational component for embodying its intrinsic speciality for variable (i.e. discharge power).I.e. this
A component will not both decay will not deformation, but itself nor one can use monomial express the function of time.Conventionally,
Periodic component can be indicated with one group of sinusoidal and/or integral.It, can be with but in the method, for one piece of specific battery
The discharge power curve P under its completely new state is determined by empirical value0(t) and by this group of Data Identification it is periodic component.
I.2.2 the extraction of trend component
Pa(t)=P (t)-P0(t)。
I.3 model training
Convolutional neural networks algorithm is by after the dimensionality reduction of the input matrix progress multilayer to n*n and feature extraction, obtaining
Mechanism Model.Wherein, the matrix of n*n can be the characteristic extracted from picture, can also be the matrix (packet of pure values
Include sequential variable).The extensive use in image recognition of convolutional neural networks algorithm, from result, image recognition is also prison
Superintend and direct one kind of formula study.The method of the present invention shows most similar individual point it can be appreciated that being found for single battery with it
Group, and according to the historical data of close individual, infer the decay situation of this individual, while the mutation feelings in studying history data
Condition, Forming Mechanism.
I.3.1 mode input matrix
The discharge power array of single battery m is P in sampleAm, e(t)∈{Pam,1(t), Pam,2(t), Pam,3(t) ...,
Pam,Em(t) }, wherein footmark e={ 1,2,3 ..., EmIt is expressed as charge and discharge number, EmFor the full life charge and discharge number of battery m.And
And single charge-discharge cycle PAm, e(t) be one group of continuous data or high frequency sampled data trend component.
For total sample number battery M, power of battery matrix is:
I.3.2 dimensionality reduction and feature extraction
In an outstanding convolutional Neural algorithm network, dimensionality reduction layer can replace with feature extraction layer, be used for multiple times and reached
To the effect of deep learning.And learning outcome is normalized in last layer, ensure that measurement standard is consistent.
Dimensionality reduction is to increase the speed of calculating in order to reduce the redundancy of data while ensureing reliability.The main means of dimensionality reduction
There are low-frequency sampling, all boundary values to take average etc..
Feature extraction is to be filtered out unwanted data by filtering tool.Dimensionality reduction and the multiple of feature extraction are used alternatingly
It can ensure that the reservation of characteristic.
I.3.3 model exports
Model output is the forecasting mechanism of discharge power matrix.
I.4 model is verified
Not used data in sample S verify above-mentioned model.
II application systems
For the new samples in practical application, carry out real time monitoring and Life cycle prediction, and compare real time data and
Sample is included in total sample in the case of significant difference and updates mechanism again by passing predicted value.
Although specific embodiments of the present invention have been described above, it will be appreciated by those of skill in the art that this is only
For example, protection scope of the present invention is to be defined by the appended claims.Those skilled in the art without departing substantially from
Under the premise of the principle and substance of the present invention, many changes and modifications may be made, but these change and
Modification each falls within protection scope of the present invention.
Claims (16)
1. a kind of power of battery prediction model generation method, which is characterized in that include the following steps:
S1, obtain training sample and concentrate the power characterization data of each sample battery, the power characterization data includes the sample
The discharge power data of each charge and discharge of battery;
S2, the power characterization data is pre-processed to generate corresponding power data matrix;
S3, using sample battery described in the training sample set power data matrix to convolutional neural networks algorithm carry out mould
Type training, to generate the prediction model of the power of battery.
2. power of battery prediction model generation method as described in claim 1, which is characterized in that
Step S2Include the following steps:
S21, to the power characterization data carry out data extraction to obtain periodic component and trend component;
S22, the trend component is pre-processed to generate the corresponding power data matrix.
3. power of battery prediction model generation method as described in claim 1, which is characterized in that the discharge power data packet
Discharge voltage data and discharge current data are included, the power characterization data further includes access times, frequency of use, charge and discharge electrician
At least one of condition.
4. power of battery prediction model generation method as described in claim 1, which is characterized in that step S2Described in pre-process
Including data invalid processing, data normalization processing and the processing of data matrixing.
5. power of battery prediction model generation method as claimed in claim 2, which is characterized in that the periodic component is to correspond to
Discharge power curve of the sample battery under completely new state corresponding to data.
6. power of battery prediction model generation method as described in claim 1, which is characterized in that step S1Further include before with
Lower step:
S0, obtain total sample set, total sample set is sampled to obtain sample set using sample rate r, by the sample
Subset is set as the training sample set;
Step S3It is further comprising the steps of later:
S4, concentrate the power data matrix for the sample battery for having neither part nor lot in model training to the prediction using the total sample
Model is verified, if output error is more than default error, is adjusted the parameter of model training, is executed step S3。
7. a kind of power of battery prediction model generates system, which is characterized in that including data acquisition module, preprocessing module and mould
Type generation module;
The data acquisition module concentrates the power characterization data of each sample battery, the power for obtaining training sample
Performance data includes the discharge power data of each charge and discharge of the sample battery;
The preprocessing module, for being pre-processed to the power characterization data to generate corresponding power data matrix;
The model generation module, for the power data matrix using sample battery described in the training sample set to convolution
Neural network algorithm carries out model training, to generate the prediction model of the power of battery.
8. power of battery prediction model as claimed in claim 7 generates system, which is characterized in that
The preprocessing module includes data extraction module and component preprocessing module;
The data extraction module, for carrying out data extraction to the power characterization data to obtain periodic component and trend point
Amount;
The component preprocessing module, for being pre-processed to the trend component to generate the corresponding power data square
Battle array.
9. power of battery prediction model as claimed in claim 7 generates system, which is characterized in that the discharge power data packet
Discharge voltage data and discharge current data are included, the power characterization data further includes access times, frequency of use, charge and discharge electrician
At least one of condition.
10. power of battery prediction model as claimed in claim 7 generates system, which is characterized in that in the preprocessing module
The pretreatment includes data invalid processing, data normalization processing and the processing of data matrixing.
11. power of battery prediction model as claimed in claim 8 generates system, which is characterized in that the periodic component is pair
Data corresponding to discharge power curve of the sample battery answered under completely new state.
12. power of battery prediction model as claimed in claim 7 generates system, which is characterized in that the power of battery prediction
It further includes sampling module and authentication module that model, which generates system,;
The sampling module, for obtaining total sample set before data acquisition module execution, using sample rate r to described
Total sample set is sampled to obtain sample set, sets the sample set to the training sample set;
The authentication module is instructed for having neither part nor lot in model using total sample concentration after model generation module execution
The power data matrix of experienced sample battery verifies the prediction model, if output error is more than default error,
The parameter for then adjusting model training calls the model generation module.
13. a kind of power of battery prediction technique, which is characterized in that include the following steps:
T, the prediction model generated using claim 1 to 6 any one of them power of battery prediction model generation method
Power prediction is carried out to battery to be predicted, to obtain the result of the power prediction of the battery to be predicted.
14. power of battery prediction technique as claimed in claim 13, which is characterized in that step T includes the following steps:
T1, obtain the power characterization data of the battery to be predicted;
T2, the power characterization data is pre-processed to generate corresponding power data matrix;
T3, by the power data Input matrix to the prediction model carry out power prediction, to obtain the battery to be predicted
The result of power prediction.
15. a kind of power of battery forecasting system, which is characterized in that described in prediction module and any one of claim 7 to 12
Power of battery prediction model generate system;
The prediction module, for carrying out power prediction to battery to be predicted using the prediction model, with obtain it is described wait for it is pre-
Survey the result of the power prediction of battery.
16. power of battery forecasting system as claimed in claim 15, which is characterized in that the prediction module includes electricity to be predicted
Pond data acquisition module, battery preprocessing module to be predicted and prediction execution module;
The battery data acquisition module to be predicted, the power characterization data for obtaining the battery to be predicted;
The battery preprocessing module to be predicted, for being pre-processed to the power characterization data to generate corresponding power
Data matrix;
The prediction execution module, for the power data Input matrix to the prediction model to be carried out power prediction, with
Obtain the result of the power prediction of the battery to be predicted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810068746.4A CN108287317B (en) | 2018-01-24 | 2018-01-24 | Battery power prediction model generation method and system and power prediction method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810068746.4A CN108287317B (en) | 2018-01-24 | 2018-01-24 | Battery power prediction model generation method and system and power prediction method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108287317A true CN108287317A (en) | 2018-07-17 |
CN108287317B CN108287317B (en) | 2020-08-14 |
Family
ID=62835677
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810068746.4A Active CN108287317B (en) | 2018-01-24 | 2018-01-24 | Battery power prediction model generation method and system and power prediction method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108287317B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109738809A (en) * | 2019-01-10 | 2019-05-10 | 上海动力储能电池系统工程技术有限公司 | The estimation method of power and energy-storage battery power characteristic |
CN110059377A (en) * | 2019-04-02 | 2019-07-26 | 西南交通大学 | A kind of fuel battery service life prediction technique based on depth convolutional neural networks |
CN110069810A (en) * | 2019-03-11 | 2019-07-30 | 北京百度网讯科技有限公司 | Battery failures prediction technique, device, equipment and readable storage medium storing program for executing |
CN111025041A (en) * | 2019-11-07 | 2020-04-17 | 深圳供电局有限公司 | Electric vehicle charging pile monitoring method and system, computer equipment and medium |
CN112240952A (en) * | 2019-10-24 | 2021-01-19 | 北京新能源汽车技术创新中心有限公司 | Power testing method, system, computer device and storage medium |
CN112798963A (en) * | 2021-04-14 | 2021-05-14 | 杭州宇谷科技有限公司 | Method, apparatus and medium for detecting battery charging characteristic abnormality based on time series |
CN113219341A (en) * | 2021-03-23 | 2021-08-06 | 陈九廷 | Model generation and battery degradation estimation device, method, medium, and apparatus |
CN113805064A (en) * | 2021-09-18 | 2021-12-17 | 北京航空航天大学 | Lithium ion battery pack health state prediction method based on deep learning |
CN115276177A (en) * | 2022-08-18 | 2022-11-01 | 上海采日能源科技有限公司 | Energy storage battery charging and discharging power control method and device and battery control system |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103698710A (en) * | 2013-12-12 | 2014-04-02 | 中南大学 | Prediction method for life cycle of battery |
CN103954913A (en) * | 2014-05-05 | 2014-07-30 | 哈尔滨工业大学深圳研究生院 | Predication method of electric vehicle power battery service life |
CN104537268A (en) * | 2015-01-19 | 2015-04-22 | 重庆长安汽车股份有限公司 | Estimation method and device for maximum discharge power of battery |
US20160259014A1 (en) * | 2015-03-03 | 2016-09-08 | Samsung Electronics Co., Ltd. | Method and apparatus for automatically estimating remaining useful life (rul) of battery in real time |
CN107329088A (en) * | 2016-04-29 | 2017-11-07 | 株式会社日立制作所 | The health status diagnostic device and method of battery |
-
2018
- 2018-01-24 CN CN201810068746.4A patent/CN108287317B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103698710A (en) * | 2013-12-12 | 2014-04-02 | 中南大学 | Prediction method for life cycle of battery |
CN103954913A (en) * | 2014-05-05 | 2014-07-30 | 哈尔滨工业大学深圳研究生院 | Predication method of electric vehicle power battery service life |
CN104537268A (en) * | 2015-01-19 | 2015-04-22 | 重庆长安汽车股份有限公司 | Estimation method and device for maximum discharge power of battery |
US20160259014A1 (en) * | 2015-03-03 | 2016-09-08 | Samsung Electronics Co., Ltd. | Method and apparatus for automatically estimating remaining useful life (rul) of battery in real time |
CN107329088A (en) * | 2016-04-29 | 2017-11-07 | 株式会社日立制作所 | The health status diagnostic device and method of battery |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109738809A (en) * | 2019-01-10 | 2019-05-10 | 上海动力储能电池系统工程技术有限公司 | The estimation method of power and energy-storage battery power characteristic |
CN110069810A (en) * | 2019-03-11 | 2019-07-30 | 北京百度网讯科技有限公司 | Battery failures prediction technique, device, equipment and readable storage medium storing program for executing |
CN110059377A (en) * | 2019-04-02 | 2019-07-26 | 西南交通大学 | A kind of fuel battery service life prediction technique based on depth convolutional neural networks |
CN110059377B (en) * | 2019-04-02 | 2022-07-05 | 西南交通大学 | Fuel cell life prediction method based on deep convolutional neural network |
CN112240952A (en) * | 2019-10-24 | 2021-01-19 | 北京新能源汽车技术创新中心有限公司 | Power testing method, system, computer device and storage medium |
CN111025041A (en) * | 2019-11-07 | 2020-04-17 | 深圳供电局有限公司 | Electric vehicle charging pile monitoring method and system, computer equipment and medium |
CN113219341A (en) * | 2021-03-23 | 2021-08-06 | 陈九廷 | Model generation and battery degradation estimation device, method, medium, and apparatus |
CN112798963A (en) * | 2021-04-14 | 2021-05-14 | 杭州宇谷科技有限公司 | Method, apparatus and medium for detecting battery charging characteristic abnormality based on time series |
CN113805064A (en) * | 2021-09-18 | 2021-12-17 | 北京航空航天大学 | Lithium ion battery pack health state prediction method based on deep learning |
CN113805064B (en) * | 2021-09-18 | 2022-09-20 | 北京航空航天大学 | Lithium ion battery pack health state prediction method based on deep learning |
CN115276177A (en) * | 2022-08-18 | 2022-11-01 | 上海采日能源科技有限公司 | Energy storage battery charging and discharging power control method and device and battery control system |
CN115276177B (en) * | 2022-08-18 | 2023-09-26 | 上海采日能源科技有限公司 | Method and device for controlling charge and discharge power of energy storage battery and battery control system |
Also Published As
Publication number | Publication date |
---|---|
CN108287317B (en) | 2020-08-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108287317A (en) | Power of battery prediction model generation method and system, power forecasting method and system | |
Park et al. | LSTM-based battery remaining useful life prediction with multi-channel charging profiles | |
Guo et al. | Online estimation of SOH for lithium-ion battery based on SSA-Elman neural network | |
CN107957562B (en) | Online prediction method for residual life of lithium ion battery | |
CN112666464B (en) | Battery health state prediction method, device, electronic equipment and readable storage medium | |
CN110059377A (en) | A kind of fuel battery service life prediction technique based on depth convolutional neural networks | |
CN110568359B (en) | Lithium battery residual life prediction method | |
KR20200140093A (en) | Prediction Method and Prediction System for predicting Capacity Change according to Charging / Discharging Cycle of Battery | |
JP2015090709A (en) | Strategic modeling for economic optimization of grid-tied energy contexts | |
JP2023501995A (en) | Battery performance prediction | |
CN106055775A (en) | Prediction method for life of secondary battery based on particle filter and mechanism model | |
Bockrath et al. | State of charge estimation using recurrent neural networks with long short-term memory for lithium-ion batteries | |
CN114035098A (en) | Lithium battery health state prediction method integrating future working condition information and historical state information | |
Zhao et al. | The li-ion battery state of charge prediction of electric vehicle using deep neural network | |
CN109800446B (en) | Method and device for estimating voltage inconsistency in discharging process of lithium ion battery | |
CN114740388A (en) | Lithium battery residual life state evaluation method based on improved TCN | |
CN113917336A (en) | Lithium ion battery health state prediction method based on segment charging time and GRU | |
CN108594776A (en) | A kind of GaAs quality conformance control method and system based on critical process | |
CN116859255A (en) | Method, device, equipment and medium for predicting state of health of energy storage battery | |
CN116804711A (en) | Method and apparatus for providing a predicted state of health of a device battery based on a predicted usage pattern | |
Eddahech et al. | Neural networks based model and voltage control for lithium polymer batteries | |
CN117407776A (en) | Method and system for estimating health state of energy storage battery under variable working condition | |
Basia et al. | Comparison of data driven algorithms for SoH estimation of Lithium-ion batteries | |
CN113884927B (en) | Battery life active control method and system based on cloud big data | |
CN116609686B (en) | Battery cell consistency assessment method based on cloud platform big data |
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 |