CN115470742B - Lithium ion battery modeling method, system, equipment and storage medium - Google Patents
Lithium ion battery modeling method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN115470742B CN115470742B CN202211342111.1A CN202211342111A CN115470742B CN 115470742 B CN115470742 B CN 115470742B CN 202211342111 A CN202211342111 A CN 202211342111A CN 115470742 B CN115470742 B CN 115470742B
- Authority
- CN
- China
- Prior art keywords
- network
- loss function
- representing
- prediction model
- dod
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
- G06F17/13—Differential equations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/10—Energy storage using batteries
Abstract
The invention discloses a lithium ion battery modeling method, a system, equipment and a storage medium, wherein the method obtains a partial differential equation of a lithium ion battery according to a charge conservation law and an ohm law, and defines boundary conditions and initial conditions according to the partial differential equation; constructing a network prediction model based on a physical information neural network; constructing a first loss function, and performing iterative update on parameters of the network prediction model by adopting a gradient descent algorithm to complete first training of the network prediction model; constructing a second loss function, and performing iterative updating on parameters of the network prediction model by adopting a gradient descent algorithm to complete second training of the network prediction model; constructing a third loss function, and performing iterative updating on parameters of the network prediction model by adopting a gradient descent algorithm to finish third training of the network prediction model; and predicting parameters of the lithium ion battery through the network prediction model after three times of training. The invention can improve the prediction precision of the network prediction model.
Description
Technical Field
The invention relates to the technical field of battery modeling, in particular to a lithium ion battery modeling method, a lithium ion battery modeling system, lithium ion battery modeling equipment and a storage medium.
Background
The lithium ion battery becomes one of the first choices of the energy storage element of the new energy automobile due to a series of outstanding characteristics of high energy density, high voltage, low self-discharge rate, good stability and the like. In order to better understand the characteristics of the battery and to make reasonable use strategies, modeling of lithium ion batteries has become an important research direction. The performance of the battery electrodes is affected by the aspect ratio, the placement of the current collector tabs, and the total amount of current flowing through the electrodes. If one electrode is not optimally designed, the potential and current density will be non-uniform, as will the utilization of the active material on the electrode. This effect becomes more pronounced as the size of the electrode becomes larger, which may lead to accelerated degradation of the electrode due to excessive local utilization of active material on the electrode. Therefore, the optimal design of the electrode is very meaningful for the production of large-scale lithium ion batteries, and plays an important role in the modeling research of the lithium ion batteries.
In the existing battery modeling technology, a modeling method based on traditional numerical calculation, such as a finite element method or a finite difference method, solves a partial differential equation, and a large amount of computing resources and computing time are consumed. In the machine learning modeling method based on data driving, the electrode reaction of the battery is regarded as a black box model, only the relation between the input and the output of the model is concerned, and the training of the model is completed by updating network parameters by using input and output data obtained by actual testing. Although this method no longer needs complete equation information and a large amount of calculation, its accuracy depends on the precision and scale of the sampled data, and the cost of obtaining enough high-precision test data is also very large.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art. Therefore, the invention provides a lithium ion battery modeling method, system, equipment and storage medium, which can improve the prediction precision of a network prediction model and reduce the calculation amount of the network prediction model.
In a first aspect, an embodiment of the present invention provides a lithium ion battery modeling method, where the lithium ion battery modeling method includes:
according to the law of conservation of charge and the ohm law, obtaining a partial differential equation of the lithium ion battery, and defining a boundary condition and an initial condition according to the partial differential equation;
constructing a network prediction model based on a physical information neural network, wherein the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network;
constructing a first loss function of the anode network, the DOD network, the flux network and the cathode network according to data driving, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration frequency is reached or the first loss function is stably converged, thereby completing first training of the network prediction model;
constructing second loss functions of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration time is reached or the second loss functions are stably converged, thereby completing second training of the network prediction model;
constructing a third loss function of the positive pole network, the DOD network, the flux network and the negative pole network according to the data drive, the partial differential equation, the boundary condition and the initial condition, and constructing a total loss function according to the third loss function of the positive pole network, the DOD network, the flux network and the negative pole network; iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number is reached or the total loss function is stably converged, and completing third training of the network prediction model;
and predicting the parameters of the lithium ion battery through the network prediction model after three times of training.
Compared with the prior art, the first aspect of the invention has the following beneficial effects:
in order to improve the prediction accuracy of the network prediction model and reduce the calculated amount of the network prediction model, the partial differential equation of the lithium ion battery is obtained according to the charge conservation law and the ohm law, and the boundary condition and the initial condition are defined according to the partial differential equation; constructing a network prediction model based on a physical information neural network, wherein the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network; constructing a first loss function of the anode network, the DOD network, the flux network and the cathode network according to data driving, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration number is reached or the first loss function is stably converged, thereby completing first training of the network prediction model; constructing second loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network according to the data drive, the boundary condition and the initial condition, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration number or second loss function is stably converged, thereby completing second training of the network prediction model; constructing a third loss function of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the partial differential equation, the boundary condition and the initial condition, and constructing a total loss function according to the third loss function of the anode network, the DOD network, the flux network and the cathode network; iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number or a total loss function is stably converged, and completing third training of the network prediction model; and predicting parameters of the lithium ion battery through the network prediction model after three times of training. The method comprises the steps that a network prediction model is trained for the first time, and first loss functions of a positive pole network, a DOD network, a flux network and a negative pole network are constructed through data driving; the network prediction model is trained for the second time to construct a second loss function of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition; the third training of the network prediction model is carried out according to data drive, partial differential equations, boundary conditions and initial conditions to construct a third loss function of the positive electrode network, the DOD network, the flux network and the negative electrode network, and a total loss function is constructed according to the third loss function of the positive electrode network DOD network, the flux network and the negative electrode network; the prediction accuracy of the network prediction model is improved through three times of training. According to the method, the network prediction model is constructed based on the physical information neural network, the physical information is used, the data use amount is greatly reduced, and the training of the network prediction model can be completed only by a small amount of data, so that the calculation amount of the network prediction model is reduced.
According to some embodiments of the invention, the obtaining partial differential equations for lithium ion batteries according to the law of conservation of charge and ohm's law comprises:
according to the charge conservation law and the ohm law, the partial differential equation of the positive electrode potential and the negative electrode potential of the lithium ion battery is obtained by adopting a Poisson equation:
wherein the content of the first and second substances,the expression of the laplacian operator is shown,represents the potential of the positive electrode of the lithium ion battery,represents the negative electrode potential of the lithium ion battery,represents the resistance of the positive electrode of the lithium ion battery,represents the resistance of the negative electrode of the lithium ion battery,represents the spatial domain of the positive electrode of the lithium ion battery,represents the spatial domain of the negative electrode of the lithium ion battery,represents the current density between the positive electrode and the negative electrode of the lithium ion batteryIs expressed asY and U represent fitting parameters;
by integral termAnd converting the auxiliary depth of discharge to obtain a partial differential equation of the depth of discharge, wherein the expression for converting is as follows:
where D represents depth of discharge, t represents time,representing the theoretical capacity of the electrode per unit area.
According to some embodiments of the invention, the defining boundary conditions and initial conditions according to the partial differential equation comprises:
according to the partial differential equation of the anode potential, defining the boundary condition of the anode potential:
wherein the content of the first and second substances,representing the gradient in the normal direction outside the boundary,showing the regions of the positive electrode other than the tabs,a tab of the positive electrode is shown,which is representative of the current flowing through it,represents a length;
defining initial conditions for the anode potential:
wherein, the first and the second end of the pipe are connected with each other,which represents the initial value of the potential of the positive electrode,represents the aboveAn initial value of (1);
defining the boundary condition of the cathode potential according to the partial differential equation of the cathode potential:
wherein the content of the first and second substances,showing the negative electrode region other than the tab,a tab representing a negative electrode;
defining initial conditions for the cathode potential:
According to some embodiments of the invention, a loss function of the network prediction model is constructed, the loss function comprising a data-driven part and a physical information part, wherein:
Wherein the content of the first and second substances,representing the results of the prediction by the network prediction model,representing spatial position, i representing the current of the lithium ion battery, t representing time,which is indicative of the measured data and,indicating the amount of said measured data,representing a mean square error between a result of the network prediction model prediction and the measured data;
constructing a loss function of the physical information part, the loss function of the physical information part comprising a loss function of a partial differential equation, a loss function of an initial condition, and a loss function of a boundary condition, wherein:
Wherein the content of the first and second substances,the equation of control is expressed in terms of,which represents the number of sample points,to representSubstituting the result predicted by the network prediction model into the mean square error of the residual error generated in the control equation at each sampling point;
Wherein the content of the first and second substances,representing the result of the initial moment predicted by the network prediction model,a value representing a known initial time instant,representing the number of initial sampling points;
Wherein, the first and the second end of the pipe are connected with each other,expressing the result predicted by the network prediction model is brought into a boundary condition equation and calculatedThe sum of the residuals of the boundary condition equations at the boundary sample points;
constructing a loss function of the network prediction model according to the loss function of the data driving part and the loss function of the physical information part:
wherein the content of the first and second substances,representing a set of weights and bias parameters in the network,representing the total loss function value of the currently trained network prediction model,a weight of a loss function representing the data portion,weights representing loss functions of the partial differential equations,a weight of a loss function representing the initial condition,a weight of a loss function representing the boundary condition.
According to some embodiments of the invention, the constructing a first loss function of the positive network, the DOD network, the flux network, and the negative network comprises:
Wherein the content of the first and second substances,representing a loss function of a data driven portion of the positive network,represents the result of the prediction of the positive network,represents measured data corresponding to the positive network,representing a mean square error between a result of the positive network prediction and measured data corresponding to the positive network;
Wherein the content of the first and second substances,a loss function representing a data-driven portion of the DOD network,representing the result of the DOD network prediction,representing measured data corresponding to the DOD network,representing a mean square error between a result of the DOD network prediction and measured data corresponding to the DOD network;
Wherein the content of the first and second substances,a loss function representing a data driven portion of the flux network,represents the result of the flux network prediction,representing measured data corresponding to the flux network,representing a mean square error between a result of the flux network prediction and measured data corresponding to the flux network;
Wherein the content of the first and second substances,a loss function representing a data driven portion of the negative network,representing the result of the negative network prediction,representing measured data corresponding to the negative pole network,and the mean square error between the result of the negative pole network prediction and the measured data corresponding to the negative pole network is represented.
According to some embodiments of the invention, the constructing a second loss function for the positive network, the DOD network, the flux network, and the negative network comprises:
Wherein the content of the first and second substances,representing a loss function of a data driven portion of the positive network,weights representing loss functions of a data driven portion of the positive network,a loss function representing an initial condition of the positive network,a weight of a loss function representing an initial condition of the positive network,a loss function representing a boundary condition of the positive network,a weight of a loss function representing a boundary condition of the positive network;
Wherein the content of the first and second substances,a loss function representing a data-driven portion of the DOD network,weights representing a loss function of a data-driven portion of the DOD network,a loss function representing an initial condition of the DOD network,a weight of a loss function representing an initial condition of the DOD network,a loss function representing a boundary condition of the DOD network,a weight of a loss function representing a boundary condition of the DOD network;
Wherein the content of the first and second substances,a loss function representing a data driven portion of the flux network,weights representing loss functions of data-driven portions of the flux networkThe weight of the steel is heavy,a loss function representing an initial condition of the flux network,a weight of a loss function representing an initial condition of the flux network,a loss function representing a boundary condition of the flux network,a weight of a loss function representing a boundary condition of the flux network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a data driven portion of the negative network,weights representing loss functions of a data driven portion of the negative network,a loss function representing an initial condition of the negative network,a weight of a loss function representing an initial condition of the negative network,a loss function representing a boundary condition of the negative network,a weight of a loss function representing a boundary condition of the negative network.
According to some embodiments of the invention, the constructing a third loss function for the positive network, the DOD network, the flux network, and the negative network comprises:
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the positive network,weights representing loss functions of partial differential equations of the positive network;
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the DOD network,weights representing loss functions of partial differential equations of the DOD network;
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the flux network,weights representing loss functions of partial differential equations of the flux network;
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the negative network,weights representing loss functions of partial differential equations of the negative network.
In a second aspect, an embodiment of the present invention further provides a lithium ion battery modeling system, where the lithium ion battery modeling system includes:
the equation acquisition unit is used for acquiring a partial differential equation of the lithium ion battery according to the charge conservation law and the ohm law, and defining boundary conditions and initial conditions according to the partial differential equation;
the model building unit is used for building a network prediction model based on the physical information neural network, and the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network;
the first training unit is used for constructing first loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network according to data driving, iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration time is reached or the first loss functions are stably converged, and completing first training of the network prediction model;
the second training unit is used for constructing second loss functions of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition, iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration time is reached or the second loss functions are stably converged, and completing second training of the network prediction model;
a third training unit, configured to construct a third loss function of the positive electrode network, the DOD network, the flux network, and the negative electrode network according to the data drive, the partial differential equation, the boundary condition, and the initial condition, and construct a total loss function according to the third loss function of the positive electrode network, the DOD network, the flux network, and the negative electrode network; iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number is reached or the total loss function is stably converged, and completing third training of the network prediction model;
and the parameter prediction unit is used for predicting the parameters of the lithium ion battery through the network prediction model which completes three times of training.
In a third aspect, an embodiment of the present invention further provides a lithium ion battery modeling apparatus, including at least one control processor and a memory, where the memory is used for being connected to the at least one control processor in a communication manner; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform a lithium ion battery modeling method as described above.
In a fourth aspect, the embodiment of the present invention further provides a computer-readable storage medium, where computer-executable instructions are stored, and the computer-executable instructions are configured to cause a computer to execute a lithium ion battery modeling method as described above.
It is to be understood that the advantageous effects of the second aspect to the fourth aspect in comparison with the related art are the same as the advantageous effects of the first aspect in comparison with the related art, and reference may be made to the related description in the first aspect, and details are not repeated here.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a flow chart of a lithium ion battery modeling method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the flow of current in parallel plate electrodes of a lithium ion battery in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of the positive electrode space domain of a lithium ion battery according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a negative electrode space domain of a lithium ion battery according to an embodiment of the invention;
FIG. 5 is a schematic diagram of a network prediction model according to an embodiment of the present invention;
fig. 6 is a structural diagram of a lithium ion battery modeling system according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
In the description of the present invention, if there are first, second, etc. described, it is only for the purpose of distinguishing technical features, and it is not understood that relative importance is indicated or implied or that the number of indicated technical features is implicitly indicated or that the precedence of the indicated technical features is implicitly indicated.
In the description of the present invention, it should be understood that the orientation or positional relationship referred to, for example, the upper, lower, etc., is indicated based on the orientation or positional relationship shown in the drawings, and is only for convenience of description and simplification of description, but does not indicate or imply that the device or element referred to must have a specific orientation, be constructed in a specific orientation, and be operated, and thus should not be construed as limiting the present invention.
In the description of the present invention, it should be noted that unless otherwise explicitly defined, terms such as arrangement, installation, connection and the like should be broadly understood, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific contents of the technical solutions.
First, several terms referred to in the present application are resolved:
lithium ion battery two-dimensional model: the current flows between the parallel plate electrodes of the battery and among the parallel plate electrodes in the process of charging and discharging the battery. The one-dimensional model assumes that the gradient of the variable employed in modeling is negligible in both directions parallel to the current collector, which assumption is valid for small-scale batteries. However, for large scale batteries, a two-dimensional model is used because there may be a significant potential drop along the current collector due to ohmic drop, which affects the current distribution.
Physical information: refers to the prior knowledge of the physics governing the physical process in the actual process, including the principle laws of physics, and the partial differential equations describing various physical processes. Physical information is valuable a priori knowledge in the modeling process.
Deep neural network: the neural network is an operational model and is formed by connecting a large number of nodes, or neurons. A particular output function, called the stimulus function, is represented by each node. Every connection between two nodes represents a weighted value, called weight, for the signal passing through the connection, which is equivalent to the memory of an artificial neural network. The output of the network is different according to the connection mode of the network, the weight value and the excitation function. The network itself is usually an approximation to some algorithm or function in nature, and may be an expression of a logic strategy.
Latin Hypercube Sampling (LHS): the method is hierarchical random sampling, and can perform efficient sampling from the distribution interval of variables.
In the existing battery modeling technology, a modeling method based on traditional numerical calculation, such as a finite element method or a finite difference method, solves a partial differential equation, and a large amount of computing resources and computing time are consumed. And the machine learning modeling method based on data driving regards the electrode reaction of the battery as a black box model, only concerns the relation between the input and the output of the model, and the training of the model is completed by updating network parameters by using input and output data obtained by actual testing. Although this method no longer requires complete equation information and a large amount of calculation, the accuracy depends on the accuracy and scale of the sampled data, and the cost of obtaining enough high-precision test data is also very large.
In order to solve the problems, in order to improve the prediction accuracy of a network prediction model and reduce the calculated amount of the network prediction model, a partial differential equation of a lithium ion battery is obtained according to a charge conservation law and an ohm law, and boundary conditions and initial conditions are defined according to the partial differential equation; constructing a network prediction model based on a physical information neural network, wherein the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network; according to data driving, constructing a first loss function of a positive electrode network, a DOD network, a flux network and a negative electrode network, and iteratively updating parameters of a network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration time is reached or the first loss function is stably converged, thereby completing first training of the network prediction model; constructing second loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network according to the data drive, the boundary condition and the initial condition, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration number or second loss function is stably converged, thereby completing second training of the network prediction model; constructing a third loss function of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the partial differential equation, the boundary condition and the initial condition, and constructing a total loss function according to the third loss function of the anode network, the DOD network, the flux network and the cathode network; iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number or a total loss function is stably converged, and completing third training of the network prediction model; and predicting parameters of the lithium ion battery through the network prediction model after three times of training. The network prediction model is trained for the first time, and a first loss function of a positive electrode network, a DOD network, a flux network and a negative electrode network is constructed through data driving; the network prediction model is trained for the second time to construct a second loss function of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition; the third training of the network prediction model constructs a third loss function of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the partial differential equation, the boundary condition and the initial condition, and constructs a total loss function according to the third loss function of the DOD network, the flux network and the cathode network of the anode network; the prediction accuracy of the network prediction model is improved through three times of training. According to the invention, the network prediction model is constructed based on the physical information neural network, the physical information is used, the data usage amount is greatly reduced, and the training of the network prediction model can be completed only by a small amount of data, so that the calculation amount of the network prediction model is reduced.
Referring to fig. 1, an embodiment of the present invention provides a lithium ion battery modeling method, where the lithium ion battery modeling method includes:
and S100, obtaining a partial differential equation of the lithium ion battery according to the charge conservation law and the ohm law, and defining boundary conditions and initial conditions according to the partial differential equation.
Specifically, referring to fig. 2, according to the law of conservation of charge and the ohm law, the poisson equation is used to obtain the partial differential equation of the positive electrode potential and the negative electrode potential of the lithium ion battery:
wherein, the first and the second end of the pipe are connected with each other,the expression of the laplacian operator is shown,represents the potential of the positive electrode of the lithium ion battery,represents the negative electrode potential of the lithium ion battery,represents the resistance of the positive electrode of the lithium ion battery,represents the resistance of the negative electrode of the lithium ion battery,represents the spatial domain of the positive electrode of the lithium ion battery,representing the spatial domain of the negative electrode of a lithium ion battery,represents the current density between the positive and negative electrodes of a lithium ion battery as a function of the potential difference between the positive and negative electrodes,the expression of (a) is:
by passingThe partial differential equation of the current density is obtained. Where Y and U represent fitting parameters that are functions of depth of discharge, the expressions for Y and U are as follows:
wherein, the first and the second end of the pipe are connected with each other,toIs a constant determined by experiment, and can obtain the current density on the electrodeIs a function of position and time on the electrode. An expression for depth of discharge can thus be obtained:
where t is the time of the discharge and,is the theoretical capacity of the electrode per unit area.
By integral termAnd converting the auxiliary depth of discharge to obtain a partial differential equation of the depth of discharge, wherein the expression for converting is as follows:
where D represents depth of discharge, t represents time,representing the theoretical capacity per unit area of the electrode.
According to the partial differential equation of the anode potential, the boundary condition of the anode potential is defined:
spatial domain、Andreferring to fig. 3, wherein a in fig. 3 represents a schematic diagram of a positive electrode of a lithium ion battery, and b in fig. 3 represents a spatial domainA diagram c in fig. 3 shows a schematic diagram of a positive electrode region except for a tab, and a diagram d in fig. 3 shows a schematic diagram of a tab of a positive electrode; wherein, the first and the second end of the pipe are connected with each other,representing the gradient in the normal direction outside the boundary,showing the region of the positive electrode other than the tabs,a tab of the positive electrode is shown,which is representative of the current flowing through it,indicating the length. Boundary conditionMeaning the electrode area other than the tabNo current flows; boundary conditionMeans a pass length ofEar tabHas a linear current density of。
According to the partial differential equation of the anode potential, the initial condition of the anode potential is defined as follows:
wherein the content of the first and second substances,which represents the initial value of the potential of the positive electrode,representThe initial value of (1);
defining the boundary condition of the cathode potential according to the partial differential equation of the cathode potential:
spatial domain、Andreferring to fig. 4, wherein a in fig. 4 represents a schematic diagram of a lithium ion battery cathode, and b in fig. 4 represents a spatial domainA schematic diagram, wherein a diagram c in fig. 4 shows a schematic diagram of a negative electrode region except a tab, and a diagram d in fig. 4 shows a tab schematic diagram of a negative electrode; wherein, the first and the second end of the pipe are connected with each other,showing the negative electrode area other than the tab,a tab of the negative electrode is shown. Boundary conditionMeaning the area of the electrode other than the tabNo current flows; boundary conditionMeaning the tab at the negative electrodeThe potential at (b) is fixed to zero as a reference potential.
According to the partial differential equation of the cathode potential, the initial condition of the cathode potential is defined:
In the above-mentioned equation, the first and second equations,、、andare known battery related parameters.
Step S200, constructing a network prediction model based on the physical information neural network, wherein the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network.
Specifically, referring to fig. 5, a network prediction model is constructed based on the physical information neural network, and the network prediction model includes a positive electrode network for predicting the positive electrode potential of the battery, a DOD network for predicting the depth of discharge on the battery electrodes, a flux network for predicting the current density between the battery electrodes, and a negative electrode network for predicting the negative electrode potential of the battery. Constructing a loss function of the network prediction model, wherein the loss function comprises a data driving part and a physical information part, and the loss function comprises the following steps:
Wherein the content of the first and second substances,representing the results of the prediction by the network prediction model,representing a spatial position, wherein the spatial position is a coordinate point randomly selected in space and time by adopting a Latin Hypercube Sampling (LHS) method, i represents the current of the lithium ion battery, t represents time,as an input to the network prediction model,the measured data is represented by a representation of,which is indicative of the amount of the measured data,representing the mean square error between the result predicted by the network prediction model and the measured data;
constructing a loss function of the physical information part, the loss function of the physical information part comprising a loss function of a partial differential equation, a loss function of an initial condition and a loss function of a boundary condition, wherein:
Wherein, the first and the second end of the pipe are connected with each other,the equation of control is expressed in terms of,which represents the number of sample points,to representSubstituting the result predicted by the network prediction model into the mean square error of residual errors generated in a control equation at each sampling point;
Wherein the content of the first and second substances,representing the results of the initial moments predicted by the network prediction model,a value representing a known initial time instant,representing the number of initial sampling points;
Wherein the content of the first and second substances,expressing the result predicted by the network prediction model is brought into the boundary condition equation and calculatedThe sum of the residuals of the boundary condition equations at the boundary sample points;
the goal of the training of the network prediction model is to find suitable network parameters that minimize the loss function, and therefore, the present embodiment considers introducing physical information into the loss function so that the training of the network prediction model can better converge on the true solution of the partial differential equation. Therefore, the present embodiment constructs a loss function of the network prediction model according to the loss function of the data driving part and the loss function of the physical information part:
wherein the content of the first and second substances,representing a set of weights and bias parameters in the network,representing the total loss function value of the currently trained network prediction model,the weight of the loss function representing the data portion,the weight of the loss function representing the partial differential equation,the weight of the loss function representing the initial conditions,the weight of the loss function representing the boundary condition.
Note that, in the DOD network of the present embodiment, the integral term thereofConversion by auxiliary output variables, i.e.. The calculation of the partial derivative term is realized by automatic differentiation of the neural network.
In this embodiment, the network structure used by the physical information neural network is a fully connected network structure, and the fully connected network has the advantages of high network throughput, high reliability and low delay. The embodiment utilizes the knowledge of the auxiliary physical information neural network, and the model converts the integral term by defining the auxiliary output variable, thereby facilitating the calculation of the network.
Step S300, a first loss function of the anode network, the DOD network, the flux network and the cathode network is constructed according to data driving, iterative updating is carried out on parameters of the network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration frequency is reached or the first loss function is stably converged, and first training of the network prediction model is completed.
Specifically, according to data driving, a first loss function of the anode network, the DOD network, the flux network and the cathode network is constructed, iterative updating is carried out on parameters of the network prediction model through a gradient descent algorithm, the first loss function is descended until a preset first maximum iteration number is reached or the first loss function is stably converged, and first training of the network prediction model is completed. Wherein:
Wherein, the first and the second end of the pipe are connected with each other,representing the loss function of the data-driven part of the positive network,the results of the positive network prediction are shown,represents the measured data corresponding to the positive network,the mean square error between the result of the anode network prediction and the measured data corresponding to the anode network is represented;
Wherein the content of the first and second substances,representing the loss function of the data-driven part of the DOD network,representing the result of the DOD network prediction,represents the measured data corresponding to the DOD network,representing the mean square error between the result of the DOD network prediction and the measured data corresponding to the DOD network;
Wherein the content of the first and second substances,representing the loss function of the data driven part of the flux network,the results of the flux network prediction are shown,represents the measured data corresponding to the flux network,representing a mean square error between a result of the flux network prediction and measured data corresponding to the flux network;
Wherein the content of the first and second substances,representing the loss function of the data driven part of the negative network,the result of the negative network prediction is shown,indicating the measured data corresponding to the negative network,and the mean square error between the result of the negative electrode network prediction and the measured data corresponding to the negative electrode network is shown.
After first loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network are constructed, parameters of four networks in the network prediction model are respectively updated in an iterative mode through a gradient descent algorithm, the first loss function corresponding to each network is descended until a preset first maximum iteration time is reached or the first loss function is stably converged, and first training of the network prediction model is completed.
It should be noted that the actual measurement data in this embodiment is low-frequency data, and the low-frequency data is obtained by sampling, and fine sampling is not performed, so that the information amount of the data is relatively small.
And S400, constructing second loss functions of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration time is reached or the second loss function is stably converged, so as to complete second training of the network prediction model.
Specifically, second loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network are constructed according to the data drive, the boundary conditions and the initial conditions, and parameters of the network prediction model are iteratively updated by adopting a gradient descent algorithm, so that the second loss functions are descended until a preset second maximum iteration number is reached or the second loss functions are stably converged, and second training of the network prediction model is completed. Wherein:
Wherein the content of the first and second substances,representing the loss function of the data-driven part of the positive network,weights representing loss functions of the data-driven part of the positive network,a loss function representing the initial conditions of the positive network,the weight of the loss function representing the initial conditions of the positive network,a loss function representing the boundary conditions of the positive network,weights of loss functions representing boundary conditions of the positive network;
Wherein the content of the first and second substances,representing the loss function of the data-driven part of the DOD network,weights representing loss functions of the data-driven portion of the DOD network,a loss function representing initial conditions of the DOD network,weights of the loss functions representing initial conditions of the DOD network,a loss function representing the boundary conditions of the DOD network,weights of loss functions representing boundary conditions of the DOD network;
Wherein the content of the first and second substances,representing the loss function of the data driven part of the flux network,weights representing loss functions of data-driven portions of the flux network,a loss function representing the initial conditions of the flux network,the weight of the loss function representing the initial conditions of the flux network,a loss function representing the boundary conditions of the flux network,weights of loss functions representing boundary conditions of the flux network;
Wherein the content of the first and second substances,representing the loss function of the data driven part of the negative network,weights representing loss functions of the data-driven part of the negative network,a loss function representing the initial conditions of the negative network,the weight of the loss function representing the initial conditions of the negative network,a loss function representing the boundary conditions of the negative network,a weight of a loss function representing a boundary condition of the negative network.
After second loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network are constructed, iterative updating is carried out on parameters of each network in the network prediction model by adopting a gradient descent algorithm, so that the second loss function corresponding to each network is descended until a preset second maximum iteration number is reached or the second loss function is stably converged, and second training of the network prediction model is completed.
It should be noted that, in the loss function of the data driving portion of this embodiment, the measured data is used, and the used measured data is high frequency data, and the high frequency data is obtained by sampling and performing fine sampling, so that the information amount of the data is relatively large.
Step S500, constructing third loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network according to the data drive, the partial differential equation, the boundary condition and the initial condition, and constructing a total loss function according to the third loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network; and (3) iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number or a total loss function is stably converged, and finishing third training of the network prediction model.
Specifically, a third loss function of the positive electrode network, the DOD network, the flux network and the negative electrode network is constructed according to the data drive, the partial differential equation, the boundary condition and the initial condition, and a total loss function is constructed according to the third loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network; and (3) iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm, so that the total loss function descends until a preset third maximum iteration number is reached or the total loss function is stably converged, and finishing the third training of the network prediction model. Wherein:
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a partial differential equation of the positive network,weights representing loss functions of partial differential equations of the positive network;
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the DOD network,weights representing loss functions of partial differential equations of the DOD network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a partial differential equation of the flux network,weights representing loss functions of partial differential equations of the flux network;
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the negative network,representing the weight of the loss function of the partial differential equation of the negative network.
After the third loss functions of the anode network, the DOD network, the flux network and the cathode network are constructed, the total loss function of the network prediction model is constructed according to the third loss functions of the anode network, the DOD network, the flux network and the cathode networkComprises the following steps:
and after a total loss function of the network prediction model is constructed, simultaneously training a positive electrode network, a DOD network, a flux network and a negative electrode network, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration time or the total loss function is stably converged, thereby completing the third training of the network prediction model.
It should be noted that, in the loss function of the data driving portion of this embodiment, the measured data is used, and the used measured data is high frequency data, and the high frequency data is obtained by sampling and performing fine sampling, so that the information amount of the data is relatively large.
And S600, predicting parameters of the lithium ion battery through the network prediction model after three times of training.
Specifically, parameters of the lithium ion battery are predicted through a network prediction model which completes three times of training, and the parameters of the lithium ion battery comprise the positive electrode potential of the lithium ion battery, the negative electrode potential of the lithium ion battery, the current density distribution between electrodes of the lithium ion battery and the discharge depth on the electrodes of the lithium ion battery.
It should be noted that, in the present embodiment, the network prediction models trained for the first time and the second time are obtained by respectively training four networks in the network prediction models to complete the training of the network prediction models, and the network prediction model trained for the third time is obtained by simultaneously training four networks in the network prediction models to complete the training of the network prediction models.
In the present embodiment, the purpose of modeling the lithium ion battery electrodes is to predict the positive electrode potential of the lithium ion battery, the negative electrode potential of the lithium ion battery, the current density distribution between the electrodes of the lithium ion battery, and the depth of discharge on the electrodes of the lithium ion battery. The embodiment simultaneously utilizes physical information and data drive, uses the physical information, greatly reduces the data use amount, and can complete the training of the network prediction model by only a small amount of data, so that the established model is more accurate and efficient. Therefore, parameters such as the positive electrode potential of the lithium ion battery, the negative electrode potential of the lithium ion battery, the current density distribution between the electrodes of the lithium ion battery and the discharge depth on the electrodes of the lithium ion battery can be accurately obtained through the constructed network prediction model, so that the characteristics of the battery can be better mastered and reasonable electrode design can be carried out. In the embodiment, low-frequency information is used for pre-training in model training, and high-frequency information is used for training in two stages to finish loss function convergence, so that the training precision and efficiency are enhanced. The present embodiment also introduces battery current as a control input to help the network predictive model better identify and predict.
Referring to fig. 6, an embodiment of the present invention further provides a lithium ion battery modeling system, where the lithium ion battery modeling system includes an equation obtaining unit 100, a model building unit 200, a first training unit 300, a second training unit 400, a third training unit 500, and a parameter prediction unit 600, where:
the equation obtaining unit 100 is configured to obtain a partial differential equation of the lithium ion battery according to the charge conservation law and the ohm law, and define a boundary condition and an initial condition according to the partial differential equation;
the model building unit 200 is used for building a network prediction model based on a physical information neural network, wherein the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network;
the first training unit 300 is configured to construct a first loss function of the positive electrode network, the DOD network, the flux network and the negative electrode network according to data driving, and iteratively update parameters of the network prediction model by using a gradient descent algorithm until a preset first maximum iteration number or a first loss function is stably converged, thereby completing first training of the network prediction model;
the second training unit 400 is configured to construct second loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network according to the data driving, the boundary condition and the initial condition, and iteratively update parameters of the network prediction model by using a gradient descent algorithm until a preset second maximum iteration number is reached or the second loss function is stably converged, so as to complete second training of the network prediction model;
a third training unit 500, configured to construct a third loss function of the positive electrode network, the DOD network, the flux network, and the negative electrode network according to the data driver, the partial differential equation, the boundary condition, and the initial condition, construct a total loss function according to the third loss function of the positive electrode network, the DOD network, the flux network, and the negative electrode network, iteratively update parameters of the network prediction model by using a gradient descent algorithm until a preset third maximum iteration number is reached or the total loss function is stably converged, and complete third training of the network prediction model;
and the parameter prediction unit 600 is configured to predict parameters of the lithium ion battery through the network prediction model after the three times of training.
It should be noted that, since the lithium ion battery modeling system in the embodiment is based on the same inventive concept as the lithium ion battery modeling method, the corresponding contents in the method embodiment are also applicable to the embodiment of the system, and detailed descriptions thereof are omitted here.
The embodiment of the invention also provides a lithium ion battery modeling device, which comprises: at least one control processor and a memory for communicative connection with the at least one control processor.
The memory, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. Further, the memory may include high speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory located remotely from the processor, and these remote memories may be connected to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The non-transitory software programs and instructions required to implement a lithium ion battery modeling method of the above-described embodiments are stored in a memory, and when executed by a processor, perform a lithium ion battery modeling method of the above-described embodiments, for example, perform the above-described method steps S100 to S600 in fig. 1.
The above described system embodiments are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may also be distributed over a plurality of network elements. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
Embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions, which, when executed by one or more control processors, may cause the one or more control processors to perform a lithium ion battery modeling method in the above method embodiments, for example, to perform the functions of the method steps S100 to S600 in fig. 1 described above.
It will be understood by those of ordinary skill in the art that all or some of the steps, systems, and methods disclosed above may be implemented as software, firmware, hardware, or suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.
Claims (10)
1. A lithium ion battery modeling method is characterized by comprising the following steps:
according to the law of conservation of charge and the ohm law, obtaining a partial differential equation of the lithium ion battery, and defining a boundary condition and an initial condition according to the partial differential equation;
constructing a network prediction model based on a physical information neural network, wherein the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network;
constructing a first loss function of the anode network, the DOD network, the flux network and the cathode network according to data driving, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration frequency is reached or the first loss function is stably converged, thereby completing first training of the network prediction model;
constructing second loss functions of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition, and iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration time is reached or the second loss functions are stably converged, thereby completing second training of the network prediction model;
constructing a third loss function of the positive pole network, the DOD network, the flux network and the negative pole network according to the data drive, the partial differential equation, the boundary condition and the initial condition, and constructing a total loss function according to the third loss function of the positive pole network, the DOD network, the flux network and the negative pole network; iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number is reached or the total loss function is stably converged, and completing third training of the network prediction model;
and predicting the parameters of the lithium ion battery through the network prediction model after three times of training.
2. The lithium ion battery modeling method of claim 1, wherein obtaining a partial differential equation for a lithium ion battery according to conservation of charge law and ohm law comprises:
according to the law of conservation of charge and the ohm law, the Poisson equation is adopted to obtain partial differential equations of the positive electrode potential and the negative electrode potential of the lithium ion battery:
wherein the content of the first and second substances,the expression of the laplacian operator is shown,represents the potential of the positive electrode of the lithium ion battery,represents the negative electrode potential of the lithium ion battery,represents the resistance of the positive electrode of the lithium ion battery,represents the lithium ion batteryThe resistance of the negative electrode is set to be,represents the spatial domain of the positive electrode of the lithium ion battery,represents the spatial domain of the negative electrode of the lithium ion battery,represents the current density between the positive electrode and the negative electrode of the lithium ion batteryIs expressed asY and U represent fitting parameters;
by integral termAnd converting the auxiliary depth of discharge to obtain a partial differential equation of the depth of discharge, wherein the expression for converting is as follows:
3. The lithium ion battery modeling method of claim 2, wherein said defining boundary conditions and initial conditions according to said partial differential equation comprises:
defining the boundary condition of the anode potential according to the partial differential equation of the anode potential:
wherein, the first and the second end of the pipe are connected with each other,representing the gradient in the normal direction outside the boundary,showing the region of the positive electrode other than the tabs,a tab of the positive electrode is shown,which is representative of the current flowing through it,represents a length;
defining initial conditions of the anode potential:
wherein, the first and the second end of the pipe are connected with each other,which represents the initial value of the potential of the positive electrode,represents the aboveThe initial value of (1);
defining the boundary condition of the cathode potential according to the partial differential equation of the cathode potential:
wherein the content of the first and second substances,showing the negative electrode region other than the tab,a tab representing a negative electrode;
defining initial conditions for the cathode potential:
4. The lithium ion battery modeling method of claim 1, wherein a loss function of the network prediction model is constructed, the loss function comprising a data-driven part and a physical information part, wherein:
Wherein, the first and the second end of the pipe are connected with each other,representing the results of the prediction by the network prediction model,representing spatial position, i representing the current of the lithium ion battery, t representing time,which is indicative of the measured data and,indicating the amount of said measured data,representing a mean square error between results of the network prediction model prediction and the measured data;
constructing a loss function of the physical information part, the loss function of the physical information part comprising a loss function of a partial differential equation, a loss function of an initial condition, and a loss function of a boundary condition, wherein:
Wherein, the first and the second end of the pipe are connected with each other,the equation of control is expressed in terms of,which represents the number of sample points,representSubstituting the result predicted by the network prediction model into the mean square error of the residual error generated in the control equation at each sampling point;
Wherein, the first and the second end of the pipe are connected with each other,representing the results of the initial moments predicted by the network prediction model,a value representing a known initial time instant,representing the number of initial sampling points;
Wherein, the first and the second end of the pipe are connected with each other,expressing the result predicted by the network prediction model is brought into a boundary condition equation and calculatedThe sum of the residuals of the boundary condition equations at the boundary sample points;
constructing a loss function of the network prediction model according to the loss function of the data driving part and the loss function of the physical information part:
wherein, the first and the second end of the pipe are connected with each other,representing a set of weights and bias parameters in the network,representing the total loss function value of the currently trained network prediction model,a weight of a loss function representing the data portion,weights representing loss functions of the partial differential equations,a weight of a loss function representing the initial condition,a weight of a loss function representing the boundary condition.
5. The lithium ion battery modeling method of claim 4, wherein the constructing a first loss function for the positive network, the DOD network, the flux network, and the negative network comprises:
Wherein, the first and the second end of the pipe are connected with each other,representing a loss function of a data driving portion of the positive network,represents the result of the positive network prediction,represents measured data corresponding to the positive electrode network,representing a mean square error between a result of the positive network prediction and measured data corresponding to the positive network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a data driven portion of the DOD network,representing the result of the DOD network prediction,representing measured data corresponding to the DOD network,representing a mean square error between a result of the DOD network prediction and measured data corresponding to the DOD network;
Wherein the content of the first and second substances,a loss function representing a data driven portion of the flux network,representing the result of the flux network prediction,representing measured data corresponding to the flux network,representing a mean square error between a result of the flux network prediction and measured data corresponding to the flux network;
Wherein the content of the first and second substances,a loss function representing a data driven portion of the negative network,representing the result of the negative network prediction,representing measured data corresponding to the negative pole network,and the mean square error between the result of the negative electrode network prediction and the measured data corresponding to the negative electrode network is represented.
6. The lithium ion battery modeling method of claim 4, wherein the constructing a second loss function for the positive network, the DOD network, the flux network, and the negative network comprises:
Wherein the content of the first and second substances,representing a loss function of a data driving portion of the positive network,weights representing loss functions of a data driven portion of the positive network,a loss function representing an initial condition of the positive network,a weight of a loss function representing an initial condition of the positive network,a loss function representing a boundary condition of the positive network,a weight of a loss function representing a boundary condition of the positive network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a data driven portion of the DOD network,weights representing loss functions of a data driven portion of the DOD network,a loss function representing an initial condition of the DOD network,a weight of a loss function representing an initial condition of the DOD network,a loss function representing a boundary condition of the DOD network,a weight of a loss function representing a boundary condition of the DOD network;
Wherein the content of the first and second substances,a loss function representing a data driven portion of the flux network,weights representing loss functions of a data driven portion of the flux network,a loss function representing an initial condition of the flux network,a weight of a loss function representing an initial condition of the flux network,a loss function representing a boundary condition of the flux network,a weight of a loss function representing a boundary condition of the flux network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a data driven portion of the negative network,weights representing loss functions of a data driven portion of the negative network,is shown inThe loss function of the initial condition of the negative network,a weight of a loss function representing an initial condition of the negative network,a loss function representing a boundary condition of the negative network,a weight of a loss function representing a boundary condition of the negative network.
7. The lithium ion battery modeling method of claim 6, wherein the constructing a third loss function for the positive network, the DOD network, the flux network, and the negative network comprises:
Wherein the content of the first and second substances,a loss function representing a partial differential equation of the positive network,weights representing loss functions of partial differential equations of the positive network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a partial differential equation of the DOD network,weights representing loss functions of partial differential equations of the DOD network;
Wherein, the first and the second end of the pipe are connected with each other,a loss function representing a partial differential equation of the flux network,weights representing loss functions of partial differential equations of the flux network;
8. A lithium ion battery modeling system, comprising:
the equation acquisition unit is used for acquiring a partial differential equation of the lithium ion battery according to the charge conservation law and the ohm law and defining a boundary condition and an initial condition according to the partial differential equation;
the model building unit is used for building a network prediction model based on a physical information neural network, and the network prediction model comprises a positive electrode network, a DOD network, a flux network and a negative electrode network;
the first training unit is used for constructing first loss functions of the positive electrode network, the DOD network, the flux network and the negative electrode network according to data driving, iteratively updating parameters of the network prediction model by adopting a gradient descent algorithm until a preset first maximum iteration time is reached or the first loss functions are stably converged, and completing first training of the network prediction model;
the second training unit is used for constructing second loss functions of the anode network, the DOD network, the flux network and the cathode network according to the data drive, the boundary condition and the initial condition, and performing iterative updating on parameters of the network prediction model by adopting a gradient descent algorithm until a preset second maximum iteration number is reached or the second loss function is stably converged, so as to complete second training of the network prediction model;
a third training unit, configured to construct a third loss function of the positive electrode network, the DOD network, the flux network, and the negative electrode network according to the data drive, the partial differential equation, the boundary condition, and the initial condition, and construct a total loss function according to the third loss function of the positive electrode network, the DOD network, the flux network, and the negative electrode network; iteratively updating the parameters of the network prediction model by adopting a gradient descent algorithm until a preset third maximum iteration number is reached or the total loss function is stably converged, and completing third training of the network prediction model;
and the parameter prediction unit is used for predicting the parameters of the lithium ion battery through the network prediction model which completes three times of training.
9. A lithium ion battery modeling apparatus comprising at least one control processor and a memory for communicative connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform the lithium ion battery modeling method of any of claims 1-7.
10. A computer-readable storage medium having stored thereon computer-executable instructions for causing a computer to perform the lithium ion battery modeling method of any of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211342111.1A CN115470742B (en) | 2022-10-31 | 2022-10-31 | Lithium ion battery modeling method, system, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211342111.1A CN115470742B (en) | 2022-10-31 | 2022-10-31 | Lithium ion battery modeling method, system, equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115470742A CN115470742A (en) | 2022-12-13 |
CN115470742B true CN115470742B (en) | 2023-03-14 |
Family
ID=84336989
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211342111.1A Active CN115470742B (en) | 2022-10-31 | 2022-10-31 | Lithium ion battery modeling method, system, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115470742B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016151336A1 (en) * | 2015-03-26 | 2016-09-29 | Imperial Innovations Limited | Battery model comprising plurality of equivalent circuit networks and interface element coupling them |
CN113111579A (en) * | 2021-04-02 | 2021-07-13 | 华北电力大学(保定) | Lithium battery equivalent circuit model parameter identification method of adaptive longicorn whisker optimization neural network |
WO2021218693A1 (en) * | 2020-04-30 | 2021-11-04 | 华为技术有限公司 | Image processing method, network training method, and related device |
CN113777496A (en) * | 2021-09-06 | 2021-12-10 | 北京化工大学 | Lithium ion battery residual life prediction method based on time convolution neural network |
CN114236410A (en) * | 2021-11-24 | 2022-03-25 | 广东工业大学 | Lithium battery SOH estimation method based on novel generation type antagonistic neural network |
CN114966451A (en) * | 2022-06-06 | 2022-08-30 | 北京航空航天大学 | Power supply product residual service life online prediction method based on physical information neural network |
CN115221798A (en) * | 2022-09-19 | 2022-10-21 | 中南大学 | Battery thermal process space-time modeling prediction method, system, equipment and medium |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10354026B2 (en) * | 2016-02-16 | 2019-07-16 | Dassault Systemes Simulia Corp. | System and method for the generation and use of an electro-thermal battery model |
-
2022
- 2022-10-31 CN CN202211342111.1A patent/CN115470742B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016151336A1 (en) * | 2015-03-26 | 2016-09-29 | Imperial Innovations Limited | Battery model comprising plurality of equivalent circuit networks and interface element coupling them |
WO2021218693A1 (en) * | 2020-04-30 | 2021-11-04 | 华为技术有限公司 | Image processing method, network training method, and related device |
CN113111579A (en) * | 2021-04-02 | 2021-07-13 | 华北电力大学(保定) | Lithium battery equivalent circuit model parameter identification method of adaptive longicorn whisker optimization neural network |
CN113777496A (en) * | 2021-09-06 | 2021-12-10 | 北京化工大学 | Lithium ion battery residual life prediction method based on time convolution neural network |
CN114236410A (en) * | 2021-11-24 | 2022-03-25 | 广东工业大学 | Lithium battery SOH estimation method based on novel generation type antagonistic neural network |
CN114966451A (en) * | 2022-06-06 | 2022-08-30 | 北京航空航天大学 | Power supply product residual service life online prediction method based on physical information neural network |
CN115221798A (en) * | 2022-09-19 | 2022-10-21 | 中南大学 | Battery thermal process space-time modeling prediction method, system, equipment and medium |
Non-Patent Citations (3)
Title |
---|
Spatial Correlation-Based Incremental Learning for Spatiotemporal Modeling of Battery Thermal Process;Bing-Chuan Wang;《IEEE》;20200430;第67卷(第4期);第1-9页 * |
基于动态学习率深度神经网络的抗干扰信道编码算法;徐建业等;《计算机应用研究》;20201231(第07期);第257-260页 * |
耦合温度的锂离子电池机理建模及仿真试验;李旭昊等;《储能科学与技术》;20201030(第06期);第405-413页 * |
Also Published As
Publication number | Publication date |
---|---|
CN115470742A (en) | 2022-12-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107957562B (en) | Online prediction method for residual life of lithium ion battery | |
Li et al. | Prognostic health condition for lithium battery using the partial incremental capacity and Gaussian process regression | |
Xiong et al. | A data-driven based adaptive state of charge estimator of lithium-ion polymer battery used in electric vehicles | |
CN107632272B (en) | Battery discharge state of charge accurate estimation method based on battery core temperature prediction | |
Li et al. | A novel parameter and state-of-charge determining method of lithium-ion battery for electric vehicles | |
CN109633472B (en) | State of charge estimation algorithm of single lithium battery | |
CN115632179B (en) | Intelligent quick charging method and system for lithium ion battery | |
CN104182630A (en) | Residual battery capacity detection method based on simplified least square support vector machine | |
CN115221798B (en) | Battery thermal process space-time modeling prediction method, system, equipment and medium | |
CN110688808A (en) | Particle swarm and LM optimization hybrid iterative identification method of power battery model | |
CN105203963A (en) | Charge state estimation method based on open-circuit voltage hysteretic characteristics | |
CN112462282B (en) | Method for determining real-time state of charge of battery pack based on mechanism model | |
CN112269133B (en) | SOC estimation method based on pre-charging circuit model parameter identification | |
CN113868934A (en) | Parallel lithium ion battery electrochemical parameter identification method | |
CN110196395A (en) | Battery SOC estimation method | |
Yu et al. | Life-cycle parameter identification method of an electrochemical model for lithium-ion battery pack | |
CN113917336A (en) | Lithium ion battery health state prediction method based on segment charging time and GRU | |
CN116643196A (en) | Battery health state estimation method integrating mechanism and data driving model | |
CN110232432B (en) | Lithium battery pack SOC prediction method based on artificial life model | |
CN114839542A (en) | Sample data set generation method and SOC estimation method for power lithium battery | |
Liu et al. | Online state of charge estimation for lithium-ion battery by combining incremental autoregressive and moving average modeling with adaptive H-infinity filter | |
CN115470742B (en) | Lithium ion battery modeling method, system, equipment and storage medium | |
CN116449219A (en) | Lithium battery model parameter identification method for improving VFFRLS | |
CN113466710B (en) | SOC and SOH collaborative estimation method for energy storage battery in receiving-end power grid containing new energy | |
Sangwan et al. | Estimation of optimal li-ion battery parameters considering c-rate, SOC and temperature |
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 |