CN110058162B - Parameter identification method based on linear time-invariant battery model structure - Google Patents
Parameter identification method based on linear time-invariant battery model structure Download PDFInfo
- Publication number
- CN110058162B CN110058162B CN201910438270.3A CN201910438270A CN110058162B CN 110058162 B CN110058162 B CN 110058162B CN 201910438270 A CN201910438270 A CN 201910438270A CN 110058162 B CN110058162 B CN 110058162B
- Authority
- CN
- China
- Prior art keywords
- battery
- model
- battery model
- identification
- matrix
- 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
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
- Tests Of Electric Status Of Batteries (AREA)
Abstract
A parameter identification method based on a linear time invariant battery model structure comprises the steps of firstly obtaining a design model parameter identification test through a target and priori knowledge, then determining a battery model structure, identifying battery model parameters, and finally verifying a battery model.
Description
Technical Field
The invention relates to a parameter identification method based on a linear time-invariant battery model structure.
Background
The battery system is a very complex system, the electrochemical reaction process of the battery system presents high nonlinear characteristics, the battery model is the basis of battery-related research and is one of the research difficulties in the research field, and how to identify the model parameters is the next question to be considered. Battery model identification may be defined as establishing parameter values for a battery model from given input and output data by minimizing a certain error criterion function. In the aspect of battery model identification, Githin K. of Pennsylvania State university in America and the like propose a battery model identification method based on offline least squares and an online iterative gradient correction algorithm, so that online identification of model parameters can be realized. The battery circuit model is identified online by genetic algorithm through Chris Mi and Johannes Unger of the university of Michigan and Vienna technology university of Austria, and the method has high precision and can be used for estimating the health state of the battery. The study on the online identification of the lithium ion battery circuit model is carried out by adopting a forgetting factor least square algorithm in Hu and the like of Beijing university of science and engineering, and the study shows that the algorithm can accurately predict the dynamic voltage response of the battery. In addition, similar identification algorithms, such as iterative least square adaptive filtering, linear model-based subspace algorithm, iterative augmented least square algorithm, and numerical calculation method, are also applied to battery model identification.
First, the cell model is actually a very complex nonlinear system that contains a large number of unknown parameters. The battery model is identified by grasping a large amount of prior knowledge, such as model order, model structure, various unknown constants, etc., which are usually difficult to obtain. Secondly, during the actual use of the battery, as the remaining capacity SOC and the state of aging SOH of the battery change, the parameters of the battery model change, and the rule of the change with time also changes. Therefore, the battery model parameters obtained by the method cannot guarantee the accuracy requirement of the system under the actual use environment. In view of this, it is desirable to provide a model based on a Linear Time-Invariant (LTI) model structure, and design a corresponding system identification algorithm.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a battery model parameter identification algorithm. According to the invention, through a battery model identification experiment, most of dynamic characteristics of the battery are excited, the structure of the battery model, the order of the model and input and output data are preprocessed, and factors such as the real-time performance and the operation complexity of an algorithm are comprehensively considered.
The battery model identification experiment is to artificially apply a certain excitation signal to the battery in order to obtain the structural parameters of the battery model, and then obtain the model parameters of the battery by using the input and output data obtained by measurement.
The battery model parameter identification algorithm comprises the following steps:
(1) determining the target and obtaining the prior knowledge
The using environment, the precision requirement and the like of the battery model are determined, and enough prior knowledge, such as the electrochemical reaction mechanism, the dynamic response characteristic, the time constant and the like of the battery, is obtained on the basis. These a priori knowledge will be instructive in identifying the experimental design and the preliminary determination of the model structure for the battery model.
(2) Design battery model identification experiment
And (2) designing a reasonable battery model identification experiment method according to the prior knowledge obtained in the step (1) so as to excite most dynamic characteristics of the battery. The battery model identification experiment is very important for battery system identification, and forms the data basis for battery system identification. In the case where the identification algorithm and the battery model structure have been determined, the accuracy of the battery model parameter identification will depend directly on the input signal.
And analyzing the dynamic characteristics of the battery from the aspect of frequency domain analysis, and designing a battery model identification experiment according to the dynamic characteristics. The frequency spectrum of the excitation current in the frequency domain must be sufficient to cover the frequency spectrum of the battery system. In order to obtain the frequency domain characteristics of the battery, the single battery is tested by adopting step return-to-zero response. Considering that the battery works in two modes of charging and discharging, two test methods of positive step return-to-zero and negative step return-to-zero are respectively designed. The positive step return to zero test consists of two parts: an 8A constant-current charging process consuming 300s and a standing process consuming 1500 s; the negative step return to zero test consisted of an 8A constant current discharge process that took 300s and a rest process that took 1500 s.
Based on a frequency domain analysis method, the single battery is tested every 10% of SOC. Recording the step return-to-zero response data in the range from 300s to 1800s in real time, wherein the sampling period is 1 s. And at different SOC positions, performing fast Fourier transform on the step return-to-zero voltage response data to obtain the frequency domain characteristics of the battery under the charging and discharging conditions. SOC is the state of charge of the battery.
(3) Designing a battery model structure
And (3) determining the structural form of the battery model according to the battery model identification experiment in the step (2), wherein the structural form comprises a mathematical expression structure and an order of the battery model. According to a simplified process of electrochemical reaction of the battery, the mathematical expression structure of the battery model can be described as: when the battery is charged and discharged, three polarization phenomena, namely ohmic polarization, electrochemical polarization and concentration polarization, are generated in the battery. Ohmic polarization can be expressed by the equivalent internal resistance of the cell, concentration polarization and electrochemical polarization together constitute a hysteresis characteristic in the dynamic characteristics of the cell, and can be generally expressed by the equivalent polarization capacitance of the cell. The order of the model is comprehensively considered according to the difficulty and the precision of parameter identification.
The battery can be described by the following state space model structure:
wherein u iskIs the input current of the battery model; y iskIs the output terminal voltage of the battery model; x is the number ofkThe state vector of the battery model at the last moment is obtained; x is the number ofk+1The state vector of the battery model at the next moment is obtained; A. b, C, D is the system matrix to be identified; ekTo the equilibrium potential of the cell, EkThe value of (c) may be replaced by an open circuit voltage after the battery has been sufficiently quiescent. In a physical sense, the a matrix and the B matrix together establish the dynamic characteristics inside the battery. The A matrix describes the last moment state vector xkFor the next time state vector xk+1And thus can be used to characterize the concentration polarization of the cell; b matrix describes the input current u at the previous momentkFor the next time state vector xk+1And thus can be used to characterize the electrochemical polarization of the cell; the D matrix describes the input current u at the sampling instant kkOpposite terminal voltage ykAnd thus can be used to characterize the ohmic polarization of the cell.
(4) Battery model parameter identification
And (3) on the basis of the battery model identification experimental data obtained in the step (2) and the battery model structure designed in the step (3), obtaining parameters of the battery model by adopting a system identification algorithm. From the above system analysis, the batch identification problem of the model described by equation (1) can be described as: at a given number N of sampling times k ═ 0, 1, …, N-1]In the method, a batch input sequence u is obtained according to samplingkAnd output sequence ykAnd estimating a global optimal system matrix A, B, C, D of the battery model in N sampling time by adopting a batch identification algorithm. In order to ensure the uniqueness of the recognition result, A,B. C, D matrix is the system matrix after similarity transformation. The steps of the battery model parameter identification algorithm can be summarized as follows:
1) algorithm initialization: firstly, input u to the battery modelkAnd output ykSampling is carried out; next, the weight matrix of the algorithm is initialized
Model order n, past window p, and future window f, and p ≧ f needs to be satisfied.
2) Estimating system Markov parameters:
first, the cell model is considered to be input into the autoregressive model one step ahead, i.e.:
wherein the content of the first and second substances,
is the battery model prediction output at time k; y isk-iIs the battery model output at the current time;
is a system markov parameter matrix to be predicted:
and then using the acquired battery model input ukAnd output ykAnd (3) carrying out least square estimation on the data in the formula (2) to obtain the Markov parameter of the system.
3) And (3) system matrix estimation: the A, B, C, D matrix is calculated by a least squares method.
(5) Battery model for verification of design
And (4) after the parameters of the battery model are identified and obtained in the step (4), designing a model verification test, and verifying the performance of the battery model.
The electric automobile is considered to be frequently operated in the states of acceleration and regenerative braking under the actual working condition of the electric automobile. Therefore, under the high-frequency pulse working condition, the designed current pulse amplitude under the high-frequency pulse working condition is +/-10A, and the minimum pulse width is 10 s. And taking the high-frequency pulse working condition current as an input signal of the battery model.
Further, it is considered that the battery may operate in a constant current discharge state for a long time under the actual operation condition of the electric vehicle. Under the working condition of long-period constant-current charging and discharging, the designed working condition of long-period constant-current charging and discharging is divided into 8 stages in total, the amplitude of each stage is-2A, -5A, -8A, -10A, -7A, -3A, +2A and +4A in sequence, and each stage lasts 480s respectively.
Drawings
FIG. 1 is a flow chart of battery model parameter identification;
FIG. 2 illustrates a positive step return to zero test method for a battery;
FIG. 3 illustrates a negative step return to zero test method for a battery;
FIG. 4 shows the frequency domain characteristics of the battery during charging;
FIG. 5 frequency domain characteristics of the cell in the discharged condition;
FIG. 6 is a test method for battery identification experiments in the form of an inverted M sequence;
FIG. 7 shows battery LTI model parameters obtained by identification at different SOCs;
in the case of different model orders in fig. 8, the accuracy of the obtained battery LTI model is identified.
Detailed Description
The present invention will be further described with reference to the following detailed description.
The parameter identification method based on the linear time-invariant battery model structure comprises the following steps:
(1) determining the target and obtaining the prior knowledge
The using environment, the precision requirement and the like of the battery model are determined, and enough prior knowledge, such as the electrochemical reaction mechanism, the dynamic response characteristic, the time constant and the like of the battery, is obtained on the basis. These a priori knowledge will be instructive in identifying the experimental design and the preliminary determination of the model structure for the battery model.
(2) Design battery model identification experiment
And (2) designing a reasonable battery model identification experiment method according to the prior knowledge obtained in the step (1) so as to excite most dynamic characteristics of the battery.
And analyzing the dynamic characteristics of the battery from the aspect of frequency domain analysis, and designing a battery model identification experiment according to the dynamic characteristics. The frequency spectrum of the excitation current in the frequency domain must be sufficient to cover the frequency spectrum of the battery system. In order to obtain the frequency domain characteristics of the battery, the single battery is tested by adopting step return-to-zero response. Considering that the battery can work in two modes of charging and discharging, two test methods of positive step return-to-zero and negative step return-to-zero are designed respectively. The positive step return to zero test consists of two parts: an 8A constant-current charging process consuming 300s and a standing process consuming 1500 s; the negative step return to zero test consisted of an 8A constant current discharge process that took 300s and a rest process that took 1500 s.
Based on a frequency domain analysis method, the single battery is tested every 10% of SOC. Recording the step return-to-zero response data in the range from 300s to 1800s in real time, wherein the sampling period is 1 s. And at different SOC positions, performing fast Fourier transform on the step return-to-zero voltage response data to obtain the frequency domain characteristics of the battery under the charging and discharging conditions.
(3) Designing a battery model structure
And (3) determining the structural form of the battery model according to the battery model identification experiment designed in the step (2), wherein the structural form comprises a mathematical expression structure and an order of the battery model. According to a simplified process of electrochemical reaction of the battery, the mathematical expression structure of the battery model can be described as: when the battery is charged and discharged, three polarization phenomena, namely ohmic polarization, electrochemical polarization and concentration polarization, are generated in the battery. Ohmic polarization can be expressed by the equivalent internal resistance of the cell, concentration polarization and electrochemical polarization together constitute a hysteresis characteristic in the dynamic characteristics of the cell, and can be generally expressed by the equivalent polarization capacitance of the cell. The order of the model is comprehensively considered according to the difficulty and the precision of parameter identification.
The battery can be described by the following state space model structure:
wherein u iskIs the input current of the battery model; y iskIs the output terminal voltage of the battery model; x is the number ofkIs the state vector of the battery model; A. b, C, D is the system matrix to be identified; ekThe value of the equalizing potential of the battery can be replaced by the open-circuit voltage after the battery is fully static.
In a physical sense, the a matrix and the B matrix together establish the dynamic characteristics inside the battery. The A matrix describes the last moment state vector xkFor the next time state vector xk+1And thus can be used to characterize the concentration polarization of the cell; b matrix describes the input current u at the previous momentkFor the next time state vector xk+1And thus can be used to characterize the electrochemical polarization of the cell; d matrix describes input current u at time kkTerminal voltage y at time kkAnd thus can be used to characterize the ohmic polarization of the cell.
(4) Battery model parameter identification
And (3) on the basis of the battery model identification experimental data obtained in the step (2) and the battery model structure designed in the step (3), obtaining parameters of the battery model by adopting a system identification algorithm. From the above system analysis, the batch identification problem of the model described by equation (1) can be described as: at a given number N of sampling times k ═ 0, 1, …, N-1]In the method, a batch input sequence u is obtained according to samplingkAnd output sequence ykThe global optimal system matrix A, B, C, D for the model over N sample times is estimated using a batch identification algorithm. In order to ensure the uniqueness of the identification result, the A, B, C, D matrix is a system matrix after similarity transformation. The steps of the battery model parameter identification algorithm can be summarized as follows:
1) algorithm initialization: firstly, input u to the battery modelkAnd output ykSampling is carried out; next, the weight matrix of the algorithm is initialized
Model order n, past window p, and future window f, and p ≧ f needs to be satisfied.
2) Estimating system Markov parameters:
first, the cell model is considered to be input into the autoregressive model one step ahead, i.e.:
wherein the content of the first and second substances,
is the battery model prediction output at time k; y isk-iIs the battery model output at the current time;
is a system markov parameter matrix to be predicted:
and then using the acquired battery model input ukAnd output ykAnd (3) carrying out least square estimation on the data in the formula (2) to obtain the Markov parameter of the system.
3) And (3) system matrix estimation: the A, B, C, D matrix is calculated by a least squares method.
(5) Battery model for verification of design
And (4) after the parameters of the battery model are identified and obtained in the step (4), designing a model verification test, and verifying the performance of the battery model.
The electric automobile is considered to be frequently operated in the states of acceleration and regenerative braking under the actual working condition of the electric automobile. Therefore, under the high-frequency pulse working condition, the designed current pulse amplitude under the high-frequency pulse working condition is +/-10A, and the minimum pulse width is 10 s. And taking the high-frequency pulse working condition current as an input signal of the battery model.
Further, it is considered that the battery may operate in a constant current discharge state for a long time under the actual operation condition of the electric vehicle. Under the working condition of long-period constant-current charging and discharging, the designed working condition of long-period constant-current charging and discharging is divided into 8 stages in total, the amplitude of each stage is-2A, -5A, -8A, -10A, -7A, -3A, +2A and +4A in sequence, and each stage lasts 480s respectively.
The specific application embodiment is as follows:
(1) determining a target and obtaining a priori knowledge
In general, the operation data of the battery system of the electric automobile in daily driving and charging processes cannot directly reflect the essential characteristics of the system. When the method is directly used for identification, the expected effect cannot be achieved, and the input signal of the identification experiment needs to be specially designed, so that the purpose of ensuring that the input and output data obtained by the identification experiment contain the key dynamic characteristics of the system is achieved.
The spectrum of the excitation current from the frequency domain must be sufficient to cover the spectrum of the battery system, and the parameters of the battery model can be accurately obtained based on such input-output data. Therefore, the dynamic characteristics of the battery are analyzed from the aspect of frequency domain analysis, and the battery model identification experiment is designed according to the dynamic characteristics. In order to obtain the frequency domain characteristics of the battery, the single battery is tested by adopting step return-to-zero response. Aiming at the fact that a battery works in a charging mode and a discharging mode, two testing methods of positive step return-to-zero and negative step return-to-zero are designed respectively. The positive step return to zero test consists of two parts: an 8A constant current charging process which takes 300s and a standing process which takes 1500s, wherein the 8A constant current charging process of 300s is used for stabilizing the polarization process of the battery, and the standing process of 1500s is used for gradually restoring the polarization state of the battery to a stable state, as shown in figure 2. Similarly, the negative step return to zero test consisted of an 8A constant current discharge process that took 300s and a rest process that took 1500s, as shown in fig. 3.
Based on the test methods shown in fig. 2 and 3, the unit cells were tested every 10% SOC with 90% SOC as a starting point. SOC is the battery state of charge. In the experimental process, the step return to zero response data in the range from the 300 th to the 1800 th intervals are recorded in real time, and the sampling period is 1 s. For ease of description, at 90% SOC, the positive step return-to-zero response curve and the negative step return-to-zero response curve of the battery are plotted. Next, at different SOCs, fast fourier transform is performed on the step return to zero voltage response data, and the frequency domain characteristics of the battery under the charging and discharging conditions are respectively shown in fig. 4 and fig. 5.
In order to facilitate description of the frequency domain characteristic of the battery with respect to the SOC, fig. 4 and 5 are described using both a 3-dimensional diagram and a 2-dimensional plan view. From fig. 4 and 5, the following a priori knowledge can be derived:
a. in the frequency range of 0Hz to 0.014Hz, the spectral amplitude of the single battery 1 is attenuated from 10 to 0.1, which is reduced by 100 times. Therefore, the main band interval of the unit cell 1 is considered to be 0Hz to 0.014 Hz. When the identification experiment is performed on the unit cell 1, the frequency spectrum of the excitation current signal must cover a frequency bandwidth of 0Hz to 0.014 Hz.
The bandwidth from b.0Hz to 0.014Hz corresponds to the time domain width of [75s, ∞ ], and the minimum time constant of the battery is tau 1/0.014 approximately to 75 s. The interpretation in the time domain is: when testing the battery, the excitation current signal should last at least 75s, so that there is enough time for the battery to exhibit its main dynamic characteristics.
c. In terms of identifying experimental sampling period choices, the value is preferably too small relative to the minimum time constant, otherwise there may be no difference in the output sample values at adjacent times. This can cause the product matrix containing the input and output data to be nearly singular, which in turn can cause the system of ill-conditioned equations to make the solution of the identification unreliable. Meanwhile, the sampling period value cannot be too large, otherwise, the useful information loss of the system is too large, and the identification precision is directly influenced. In combination with the above considerations, the sampling period value of the selected identification experiment is 10 s.
(2) Lithium ion battery parameter identification experiment design
And designing a test method of the battery identification experiment based on the prior knowledge of the battery characteristics. In order to ensure the accuracy of identification and the general applicability of the model, it is desirable that the input signal is capable of exciting an essential feature or mode of the battery system, such input signal being referred to as a continuous excitation signal. An M-sequence having an approximately white noise property is employed as an excitation signal. The M sequence has the property of approximate white noise, and good identification precision can be obtained. The M-sequence contains a dc component, which causes a "net perturbation" to the battery system, i.e., changes the state of charge of the battery. The inverse M-sequence will overcome this disadvantage and is a more ideal pseudo-random code sequence than the M-sequence. As shown in fig. 6, the battery identification test method in the form of an inverse M-sequence consists of a set of positive and negative pulses with different durations, and the minimum pulse width is 75 s. In the aspect of selecting the amplitude of the excitation current, the amplitude of the excitation current is not too small, and if the amplitude of the excitation current is too small, noise in a system is dominant to submerge useful signals; the amplitude of the exciting current is not too large, otherwise the battery system enters a nonlinear region. Under the use environment of the electric automobile, the average charge-discharge multiplying power of the power battery is usually lower than 1C, so 0.75C, namely 8A is selected as the amplitude of the charge-discharge current.
(3) Designing a battery model structure
When the battery is charged and discharged, three polarization phenomena, namely ohmic polarization, electrochemical polarization and concentration polarization, are generated in the battery. The battery can be described by the following state space model structure:
wherein u iskIs the input current of the battery model; y iskIs the output terminal voltage of the battery model; x is the number ofkIs the state vector of the battery model; A. b, C, D is the system matrix to be identified; ekThe value of the equalizing potential of the battery can be replaced by the open-circuit voltage after the battery is fully static.
In a physical sense, the a matrix and the B matrix together establish the dynamic characteristics inside the battery. The A matrix describes the last moment state vector xkFor the next time state vector xk+1And thus can be used to characterize the concentration polarization of the cell; b matrix describes the input current u at the previous momentkFor the next time state vector xk+1Due to the influence ofBut can be used to characterize the electrochemical polarization of the cell; d matrix describes input current u at time kkTerminal voltage y at time kkAnd thus can be used to characterize the ohmic polarization of the cell.
(4) Battery model parameter identification
And identifying the model parameters by using a subspace-based model identification method. The subspace identification firstly adopts methods such as singular value decomposition, model structure optimization and the like to decompose a data space formed by input/output data into two mutually orthogonal subspaces of signals and noise, so as to obtain a Kalman state sequence of the system, and a system matrix is estimated on the basis.
From the above system analysis, the batch identification problem of the model can be described as: at a given number N of sampling times k ═ 0, 1, …, N-1]In the method, a batch input sequence u is obtained according to samplingkAnd output sequence ykThe global optimal system matrix A, B, C, D for the model over N sample times is estimated using a batch identification algorithm. To ensure the uniqueness of the recognition result, AT、BT、CT、DTThe matrix is a system matrix after similarity transformation.
At different SOCs, the identified battery model parameters are shown in FIG. 7. FIG. 7 depicts A at different SOCs, respectivelyTElement a in the matrix0、a1、a2、a3Value of (A), BTElement b in the matrix0、b1、b2、b3Value of (D)TThe value of the element d in the matrix, and the value of the equalizing potential E at different SOCs. It can be seen that A is identifiedTMatrix sum BTThe matrix element values exhibit bilateral symmetry with 50% SOC as the center. This is because lithium ions are inserted into and extracted from the positive and negative electrodes during the reaction, and when the SOC is 50%, the lithium ion concentrations in the positive and negative electrodes of the battery are equal, and therefore the dynamic characteristics of the battery are bilaterally symmetric around the 50% SOC.
(5) Verifying battery models
And evaluating the performance of the model based on the battery model obtained by the identification, and adopting a prediction Variance ratio (VAF) as a standard For judging the quality of the prediction result of the algorithm. VAF is defined as follows:
wherein the content of the first and second substances,
for the predicted value of the dynamic voltage response of the battery, ykThe measured value of the dynamic voltage response of the battery is N, and the number of the sampling points is N. The larger the predicted variance ratio VAF value is, the better the performance of the model is, and generally, the application requirement of the model can be met when the predicted variance ratio VAF value is more than 95%.
The predicted variance ratio VAF values obtained by identification at different model orders are shown in FIG. 8. It can be seen that the identified model prediction variance ratio VAF value has reached 97% when the model order n is 4, and the prediction variance ratio VAF value has not significantly improved when the model order n > 4. Therefore, the LTI model structure of 4 orders is selected to be optimal.
Claims (1)
1. A parameter identification method based on a linear time invariant battery model structure is characterized in that: firstly, acquiring a design model parameter identification test through a target and priori knowledge, then determining a battery model structure, identifying battery model parameters, and finally verifying a battery model; the specific identification process of the battery model parameters comprises the following steps:
(1) determining the target and obtaining the prior knowledge
Determining the use environment and the precision requirement of the battery model, and obtaining enough prior knowledge on the basis, wherein the prior knowledge comprises an electrochemical reaction mechanism, a dynamic response characteristic and a time constant of the battery; the priori knowledge plays a guiding role in battery model identification experimental design and preliminary determination of a model structure;
(2) design battery model identification experiment
Designing a reasonable battery model identification experiment method according to the prior knowledge obtained in the step (1) so as to excite most dynamic characteristics of the battery; the battery model identification experiment forms a data basis for battery system identification, and under the condition that an identification algorithm and a battery model structure are determined, the accuracy of battery model parameter identification directly depends on an input signal;
analyzing the dynamic characteristics of the battery from the angle of frequency domain analysis, and designing a battery model identification experiment according to the dynamic characteristics; the frequency spectrum of the excitation current in the frequency domain must be sufficient to cover the frequency spectrum of the battery system; in order to obtain the frequency domain characteristics of the battery, testing the single battery by adopting step return-to-zero response; considering that the battery works in two modes of charging and discharging, respectively designing two test methods of positive step return-to-zero and negative step return-to-zero; the positive step return to zero test consists of two parts: an 8A constant-current charging process consuming 300s and a standing process consuming 1500 s; the negative step return-to-zero test consists of an 8A constant current discharge process consuming 300s and a standing process consuming 1500 s;
testing the single battery every 10% of SOC based on a frequency domain analysis method; recording step return-to-zero response data in the range from 300s to 1800s in real time, wherein the sampling period is 1 s; performing fast Fourier transform on the step return-to-zero voltage response data at different SOC positions to obtain the frequency domain characteristics of the battery under the charging and discharging conditions; SOC is the state of charge of the battery;
(3) designing a battery model structure
Determining the structural form of the battery model according to the battery model identification experiment in the step (2), wherein the structural form comprises a mathematical expression structure and an order of the battery model; according to the simplified process of the electrochemical reaction of the battery, the mathematical expression structure of the battery model is described as follows: when the battery is charged and discharged, three polarization phenomena, namely ohmic polarization, electrochemical polarization and concentration polarization, can be generated in the battery; ohmic polarization is expressed by the equivalent internal resistance of the battery, concentration polarization and electrochemical polarization jointly form the hysteresis characteristic in the dynamic characteristic of the battery, and are usually expressed by the equivalent polarization capacitance of the battery; comprehensively considering the order of the model according to the difficulty and the precision of parameter identification;
the battery is described by the following state space model structure:
wherein u iskIs the input current of the battery model; y iskIs the output terminal voltage of the battery model; x is the number ofkA state vector of the battery model at the last moment is obtained; x is the number ofk+1The state vector of the battery model at the next moment is A, B, C, D a system matrix to be identified; ekThe value of the balanced potential of the battery can be replaced by the open-circuit voltage after the battery is fully static;
in a physical sense, the matrix A and the matrix B jointly establish the dynamic characteristics in the battery; the A matrix describes the last moment state vector xkFor the next time state vector xk+1To characterize the concentration polarization of the cell; b matrix describes the input current u at the previous momentkFor the next time state vector xk+1To characterize the electrochemical polarization of the cell; d matrix describes input current u at time kkTerminal voltage y at time kkTo characterize ohmic polarization of the cell;
(4) battery model parameter identification
On the basis of the battery model identification experimental data obtained in the step (2) and the battery model structure designed in the step (3), acquiring parameters of the battery model by adopting a system identification algorithm;
from the above system analysis, the batch identification problem of the model of equation (1) is described as: at a given number N of sampling times k ═ 0, 1, …, N-1]In the method, a batch input sequence u is obtained according to samplingkAnd output sequence ykEstimating a global optimal system matrix A, B, C, D of the model in N sampling times by adopting a batch identification algorithm;
in order to ensure the uniqueness of the identification result, the A, B, C, D matrix is a system matrix after similarity transformation, and the steps of the battery model parameter identification algorithm are summarized as follows:
1) algorithm initialization: firstly, input u to the battery modelkAnd output ykSampling is carried out; then theWeight matrix of initialization algorithm
Model order n, past window p, and future window f, and p ≧ f needs to be satisfied;
2) estimating system Markov parameters:
first, the cell model is considered to be input into the autoregressive model one step ahead, i.e.:
wherein the content of the first and second substances,
is the battery model prediction output at time k; y isk-iIs the battery model output at the current time;
is a system markov parameter matrix to be predicted:
and then using the acquired battery model input ukAnd output ykCarrying out least square estimation on the data in the formula (2) to obtain Markov parameters of the system;
3) and (3) system matrix estimation: calculating A, B, C, D a matrix by a least squares method;
(5) battery model for verification of design
After the parameters of the battery model are identified and obtained in the step (4), designing a model verification test, and verifying the performance of the battery model;
considering that the electric automobile can often work in frequent acceleration and regenerative braking states under the actual working condition; therefore, under the working condition of high-frequency pulse, the designed current pulse amplitude under the working condition of high-frequency pulse is +/-10A, and the minimum pulse width is 10 s; taking the high-frequency pulse working condition current as an input signal of a battery model;
further, the battery can work in a constant current discharge state for a long time under the actual working condition of the electric automobile; under the working condition of long-period constant-current charging and discharging, the designed working condition of long-period constant-current charging and discharging is divided into 8 stages in total, the amplitude of each stage is-2A, -5A, -8A, -10A, -7A, -3A, +2A and +4A in sequence, and each stage lasts 480s respectively.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910438270.3A CN110058162B (en) | 2019-05-24 | 2019-05-24 | Parameter identification method based on linear time-invariant battery model structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910438270.3A CN110058162B (en) | 2019-05-24 | 2019-05-24 | Parameter identification method based on linear time-invariant battery model structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110058162A CN110058162A (en) | 2019-07-26 |
| CN110058162B true CN110058162B (en) | 2020-12-01 |
Family
ID=67324311
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910438270.3A Active CN110058162B (en) | 2019-05-24 | 2019-05-24 | Parameter identification method based on linear time-invariant battery model structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110058162B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110348062B (en) * | 2019-06-14 | 2023-05-26 | 湖北锂诺新能源科技有限公司 | Construction method of equivalent circuit model of lithium ion battery |
| CN112130081B (en) * | 2020-09-24 | 2024-01-09 | 闽江学院 | Method for rapidly realizing SOC estimation of lead-acid battery |
| CN113489039B (en) * | 2021-07-29 | 2023-07-21 | 广东电网有限责任公司 | Charging scheduling method, device, equipment and storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102232257A (en) * | 2008-12-01 | 2011-11-02 | 康奈可关精株式会社 | Battery model identification method |
| CN102937704A (en) * | 2012-11-27 | 2013-02-20 | 山东省科学院自动化研究所 | Method for identifying RC (resistor-capacitor) equivalent model of power battery |
| WO2015104493A1 (en) * | 2014-01-08 | 2015-07-16 | Renault S.A.S | Method for communication with a circuit for controlling an accumulator battery |
| CN104965179A (en) * | 2015-07-06 | 2015-10-07 | 首都师范大学 | Lithium ion storage battery temperature combinational circuit model and parameter identification method thereof |
| CN105912799A (en) * | 2016-04-27 | 2016-08-31 | 华中科技大学 | Modeling method of liquid state or semi-liquid state metal battery |
| US9927492B2 (en) * | 2013-05-29 | 2018-03-27 | Nxp Usa, Inc. | Cell monitoring apparatus, battery monitoring apparatus, integrated circuit and method of monitoring a rechargeable cell |
| CN108072847A (en) * | 2018-01-29 | 2018-05-25 | 西南交通大学 | A kind of method of estimation of dynamic lithium battery identification of Model Parameters and remaining capacity |
-
2019
- 2019-05-24 CN CN201910438270.3A patent/CN110058162B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102232257A (en) * | 2008-12-01 | 2011-11-02 | 康奈可关精株式会社 | Battery model identification method |
| CN102937704A (en) * | 2012-11-27 | 2013-02-20 | 山东省科学院自动化研究所 | Method for identifying RC (resistor-capacitor) equivalent model of power battery |
| US9927492B2 (en) * | 2013-05-29 | 2018-03-27 | Nxp Usa, Inc. | Cell monitoring apparatus, battery monitoring apparatus, integrated circuit and method of monitoring a rechargeable cell |
| WO2015104493A1 (en) * | 2014-01-08 | 2015-07-16 | Renault S.A.S | Method for communication with a circuit for controlling an accumulator battery |
| CN104965179A (en) * | 2015-07-06 | 2015-10-07 | 首都师范大学 | Lithium ion storage battery temperature combinational circuit model and parameter identification method thereof |
| CN105912799A (en) * | 2016-04-27 | 2016-08-31 | 华中科技大学 | Modeling method of liquid state or semi-liquid state metal battery |
| CN108072847A (en) * | 2018-01-29 | 2018-05-25 | 西南交通大学 | A kind of method of estimation of dynamic lithium battery identification of Model Parameters and remaining capacity |
Non-Patent Citations (2)
| Title |
|---|
| Subspace-based modeling and parameter identification of lithium-ion batteries;Yong Li 等;《INTERNATIONAL JOURNAL OF ENERGY RESEARCH》;20141231;第1024-1038页 * |
| 锂离子动力电池参数辨识及其SOC估计;黄菊花 等;《南昌大学学报（工科版）》;20190331;第41卷(第1期);第62-69页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110058162A (en) | 2019-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Farmann et al. | Critical review of on-board capacity estimation techniques for lithium-ion batteries in electric and hybrid electric vehicles | |
| CN110058162B (en) | Parameter identification method based on linear time-invariant battery model structure | |
| CN111090047B (en) | Lithium battery health state estimation method based on multi-model fusion | |
| CN111581904B (en) | Lithium battery SOC and SOH collaborative estimation method considering cycle number influence | |
| US10222426B2 (en) | Double-pulse technique for on-line diagnostics of electrochemical systems | |
| WO2018161486A1 (en) | Method and system for estimating soc of power battery on the basis of dynamic parameters | |
| CN111679199B (en) | Lithium ion battery SOC estimation method and device | |
| WO2015031437A1 (en) | On-board state of health monitoring of batteries using incremental capacity analysis | |
| US9658291B1 (en) | Methods and apparatus for dynamic estimation of battery open-circuit voltage | |
| CN109507554B (en) | Electrical equipment insulation state evaluation method | |
| CN107885928B (en) | Stepping stress acceleration performance degradation reliability analysis method considering measurement error | |
| Zhang et al. | Battery state estimation using unscented kalman filter | |
| CN104267261A (en) | On-line secondary battery simplified impedance spectroscopy model parameter estimating method based on fractional order united Kalman filtering | |
| Waag et al. | Application-specific parameterization of reduced order equivalent circuit battery models for improved accuracy at dynamic load | |
| CN106126783B (en) | A kind of lithium ion battery change time scale model parameter estimation method | |
| CN110196395B (en) | Storage battery SOC estimation method | |
| CN109917299B (en) | Three-layer filtering estimation method for state of charge of lithium battery | |
| CN110596606A (en) | Lithium battery residual capacity estimation method, system and device | |
| CN109901072B (en) | Retired battery parameter detection method based on historical data and laboratory test data | |
| Sadabadi et al. | Model-based state of health estimation of a lead-acid battery using step-response and emulated in-situ vehicle data | |
| CN111142025A (en) | Battery SOC estimation method and device, storage medium and electric vehicle | |
| CN113075569A (en) | Battery state of charge estimation method and device based on noise adaptive particle filtering | |
| CN112433154A (en) | Lithium ion battery SOC estimation algorithm based on FFRLS and EKF | |
| CN114441984A (en) | Lithium battery health state estimation method | |
| CN113884935A (en) | SOH estimation system and method based on lithium battery online electrochemical impedance spectroscopy measurement |
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 |