CN-122017582-A - Lithium battery cross-working condition multi-state joint estimation method only requiring single identification and no offline OCV data

CN122017582ACN 122017582 ACN122017582 ACN 122017582ACN-122017582-A

Abstract

The invention discloses a lithium battery cross-working condition multi-state joint estimation method which only needs single identification and does not need offline OCV data, and belongs to the technical field of lithium ion batteries. The method comprises the steps of collecting experimental data of a lithium battery in a reference state, constructing a fractional order physical model to be identified, then carrying out synchronous identification operation of parameters and OCVs based on quasi-static assumption, establishing an analytic mapping relation between model parameters and the SOCs in the reference state, constructing an adaptive fractional order model integrating a linear migration mechanism and internal resistance anchoring characteristics based on the mapping relation, and constructing a lithium battery SOCs and capacity joint estimation framework based on a unified augmentation state space and a sparse information injection mechanism through the adaptive fractional order model to complete joint estimation. The invention effectively overcomes the high dependence of the prior art on repeated parameter identification and offline OCV experimental data, solves the problem of error diffusion caused by multi-state coupling, and remarkably improves the engineering applicability of the model.

Inventors

CHEN ZHENG
LIU WEIQIANG
SHEN JIANGWEI
XIA XUELEI
Wei fuxing
SHEN DONGXU
SHAO QI
WAN ENZE

Assignees

昆明理工大学

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (10)

1. The lithium battery cross-working condition multi-state joint estimation method which only needs single identification and does not need offline OCV data is characterized by comprising the following steps of: S1, collecting experimental data of a lithium battery in a reference state; S2, constructing a fractional order physical model to be identified based on the acquired reference data; s3, based on the fractional order physical model to be identified and the reference data, performing synchronous identification operation of parameters based on quasi-static assumption and OCV, and establishing an analytic mapping relation between model parameters and SOC in a reference state; S4, constructing a self-adaptive fractional order model integrating a linear migration mechanism and internal resistance anchoring characteristics based on the analytic mapping relation of the model parameters and the SOC in the reference state; and S5, constructing a lithium battery SOC and capacity joint estimation framework based on a unified augmentation state space and a sparse information injection mechanism through a self-adaptive fractional order model, and finally completing joint estimation of the lithium battery SOC and capacity under the influence of multi-stress coupling.
2. The method for cross-working-condition multi-state joint estimation of the lithium battery, which is only required to be identified once and offline OCV data is not required, is characterized in that in the step S1, the environmental condition is set to be that the environmental temperature is 25 ℃ and the battery is not extruded externally, and the change data of the current and the voltage of the battery along with the time are collected.
3. The method for cross-operating mode multi-state joint estimation of lithium battery with single identification and offline OCV data as claimed in claim 1, wherein in S2, the fractional physical model to be identified is a first-order RC fractional equivalent circuit model, and in order to ensure convergence of the initial identification process, the fractional order is defined Is a fixed value of 0.9, and the expression is as follows: In the formula, Is that The terminal voltage of the lithium battery at the moment, In the case of a fractional order differential operator, For an open circuit voltage related to the current state of charge SOC, For the polarization voltage to be applied, Is the ohmic internal resistance, For the internal resistance of polarization, In the event of a current flow, Is the generalized capacitive coefficient of the constant phase element CPE.
4. The method for cross-operating mode multi-state joint estimation of lithium battery with only single identification and no offline OCV data according to claim 1, wherein the step of identifying the parameters based on quasi-static assumption and the OCV in S3 comprises the following steps: S3.1 defining discretized fractional derivative with G-L and assuming within a preset sampling interval Reconstructing the nonlinear model into a linear regression equation: In the formula, As an observation of the terminal voltage, the voltage of the terminal, As a regression vector of the data, As the parameter vector to be identified, Is a prediction error; s3.2, identifying model parameters and OCV in a reference state on line by utilizing AFFRLS algorithm, and according to the prediction error The forgetting factor is adjusted in real time to balance the parameter tracking speed and the noise suppression capability; S3.3, constructing a reference analytic mapping relation between model parameters and the SOC by adopting 5-order polynomial fitting: In the formula, As a function of ohmic internal resistance with respect to SOC, As a function of the internal polarization resistance with respect to SOC, As a function of the generalized capacitive coefficient with respect to SOC, For an open circuit voltage related to the current state of charge SOC, 、、 And The fitting coefficients of the respective items are respectively represented and are obtained through fitting after AFFRLS identification.
5. The method for cross-operating mode multi-state joint estimation of lithium battery with single identification and offline OCV data as set forth in claim 4, wherein the forgetting factor in S3.2 is The updated formula of (c) is as follows: In the formula, For a preset minimum forgetting factor, In order to predict the error of the signal, As a reference error threshold value, Is a nonlinear activation function for mapping the normalized error to the adjustment range of the forgetting factor.
6. The method for cross-operating mode multi-state joint estimation of lithium battery with only single identification and no offline OCV data according to claim 1, wherein the linear migration mechanism in S4 is represented as: In the formula, In order to adapt to the model parameters of the current specific operating conditions, For the first and second impedance transitions, Is a model parameter in a reference state; is the SOC after the migration correction, For a coarse SOC under the influence of multiple stress couplings, Is the scaling and translation factor of the SOC, and the model parameters under the reference state Comprises ohmic internal resistance Internal resistance of polarization Generalized capacity coefficient of constant phase original Open circuit voltage And fractional order 。
7. The method for cross-operating mode multi-state joint estimation of lithium battery with only single identification and no offline OCV data according to claim 6, wherein in S4, a capacity observation equation is constructed based on internal resistance anchoring characteristics, and the method comprises the following steps: S4.1, constructing a reference based on the internal resistance physical characteristics : In the formula, Is a reference state of health estimate based on internal resistance; an internal resistance threshold at the end of battery life; Is ohmic internal resistance; nominal internal resistance of the battery when leaving the factory; s4.2, introducing an adaptive correction layer, and establishing a final SOH and capacity observation equation: In the formula, For the final corrected SOH estimate, And The scale correction factor and the offset correction factor are respectively corrected based on the terminal voltage residual error; in order to be able to use the capacity in practice, Is the standard rated capacity of the battery.
8. The method for cross-working condition multi-state joint estimation of the lithium battery, which is only required to be identified once and offline OCV data is not required, according to claim 7, wherein in S4, a specific parameter evolution equation set of the self-adaptive fractional order model which fuses the linear migration mechanism and the internal resistance anchoring characteristic is constructed as follows: In the formula, Representing coefficient terms, superscripts Representing the adaptive model parameters corrected by the migration mechanism, To describe the effect of multiple stress coupling conditions on the model parameters, Mapping functions for reference parameters; for the rough SOC value under the influence of multi-stress coupling, the method uses linear transformation Mapping to a definition domain of the reference model, so that cross-working condition parameter self-adaption without additional off-line calibration is realized; for the adaptive fractional order, As the order of the reference fraction, And Respectively an ohmic internal resistance and a polarization internal resistance after migration, For the generalized capacity coefficient after migration, For the open circuit voltage after the transition, In order to migrate the back-end voltage model output, For the polarization voltage after the migration, Is the current.
9. The method for cross-operating mode multi-state joint estimation of lithium battery with single identification and offline OCV data as claimed in claim 1, wherein the step of S5 comprises the following steps: S5.1, initial particle generation: In the formula, For the prior probability density, As the particles at the initial moment in time, As the total number of particles, Index for the total number of particles; is a dirac function; is the first The number of primary particles is chosen to be the same, In order to fast-change the sub-vector, In the form of a slow-varying subvector, In the form of a gaussian distribution, As an initial SOC-estimation value, a reference value, And Initial error covariance matrices of the fast-varying and slow-varying parameter subvectors, respectively; s5.2, initializing augmentation particles, namely constructing a unified augmentation state vector based on capturing the trans-scale coupling characteristics between different states of the lithium battery under a unified frame : In the formula, Is a discrete time The SOC at the time of the start of the process, Is a discrete time The fractional order polarization voltage at the time is, For the purpose of representing the fast time scale, For describing a slow time scale; s5.3, sparse information injection and state evolution, namely, based on a unified augmentation state space equation, introducing a sparse information injection mechanism, and constructing a state evolution equation comprising a time-varying regulation matrix: In the formula, Is the first The a priori prediction states of the individual particles, As a non-linear state transfer function, For the input to the system, To meet the process noise of the sparse injection mechanism, the obeying mean value is 0, and the covariance is Is a gaussian distribution of (c); The dynamic process noise covariance matrix is used for controlling the evolution bandwidths of different dimensional states, and is defined as follows: In the formula, And Process noise variance of SOC and polarization voltage, respectively; And Process noise variance for correction parameters; is a very small positive number, and the number of the positive numbers is very small, Is a macroscopic update period; s5.4, performing risk sensitivity weight updating, namely calculating the voltage of a particle prediction terminal, introducing an exponential cost function to calculate particle weight, and amplifying the difference of good and bad particles: In the formula, Is the first The non-normalized weight of the individual particles, As a risk-sensitive factor, For the purpose of actually measuring the terminal voltage, Is the first A terminal voltage prediction value of each particle; s5.5, weight normalization: In the formula, For the normalized weights to be used, Is a non-normalized weight; S5.6, performing particle resampling and state estimation, namely resampling the prior set according to particle weights to inhibit particle degradation, and calculating a posterior estimation value of the augmented state by using weighted summation: In the formula, Is the optimal estimated vector containing the SOC and capacity correction factors.
10. The method for cross-operating mode multi-state joint estimation of lithium battery with single identification and no offline OCV data according to claim 9, wherein the S5 joint estimation framework corrects the current actual available capacity through a correction factor outputted by filtering, and the correction factor is used as a feedback variable to be injected into an ampere-hour integral equation to carry out closed-loop correction on the rough SOC, and the mathematical expression is as follows: In the formula, In order to estimate the current available capacity, For a nominal capacity of the battery, As the mean value of the scale correction factors, For a reference SOH based on physical characteristics, Is the mean value of the offset correction factors; Is that The rough SOC at the time of the time, In order to achieve a coulombic efficiency, Is that The load current at the time of the start of the process, Is a sampling time interval; And performing adaptive scaling and translational correction on the rough SOC by using the linear migration parameters estimated by the particle filtering, and performing posterior weighting calculation based on the particle weight, thereby obtaining a final SOC estimated value, wherein the mathematical expression is as follows: In the formula, For a true SOC value based on capacity correction under the influence of multiple stress couplings, Is that The state of the 1 st particle at the time, Is that State of the 2 nd particle.

Description

Lithium battery cross-working condition multi-state joint estimation method only requiring single identification and no offline OCV data Technical Field The invention relates to the technical field of lithium ion batteries, in particular to a cross-working-condition multi-state joint estimation method of a lithium battery, which only needs single identification and does not need offline OCV data. Background With the increase of global energy crisis and the increase of environmental protection consciousness, electric Vehicles (EV) and Hybrid Electric Vehicles (HEV) have become main development directions of the automobile industry. The lithium ion battery is a preferred power source of the electric automobile by virtue of the advantages of high energy density, long cycle life, low self-discharge rate and the like. One of the key tasks of the Battery management system (Battery MANAGEMENT SYSTEM, BMS) as a core component for ensuring safe and efficient operation of a Battery is to accurately monitor the State of Charge (SOC), state of Health (SOH) and capacity of the Battery in real time. These parameters are not only the basis for energy management, but also the key basis for prolonging the service life of the battery. However, the existing battery state estimation technology still faces a plurality of bottlenecks in practical application, namely firstly, a complex strong coupling relation exists between the SOC (fast changing state) and the capacity (slow changing parameter), the traditional method is difficult to effectively decouple and easily causes errors to rapidly accumulate through ampere-hour integration, secondly, high-precision modeling generally depends on complicated mixed pulse power characteristic (Hybrid Pulse Power Characterization, HPPC) testing and parameter identification, the development period is long, the calibration cost is high and flexibility is lacking, and furthermore, the existing fixed-order fractional order model is difficult to accurately capture the nonlinear dispersion characteristic of the battery under the dynamic working condition and the wide temperature range, and the generalization capability and the estimation precision of the model are difficult to be obtained. Therefore, it is important to develop a joint estimation method that can realize multi-state high-precision decoupling without repeating the parameter identification and the off-line open circuit voltage (Open Circuit Voltage, OCV) test. Disclosure of Invention In order to solve the technical problems, the invention provides a lithium battery cross-working condition multi-state joint estimation method which only needs single identification and does not need offline OCV data. In order to implement the above technique, the steps include: s1, acquiring experimental data of a lithium battery in a reference state; S2, constructing a fractional order physical model to be identified, namely establishing an equivalent circuit model containing fractional order impedance characteristics; S3, synchronous identification of parameters and OCVs based on quasi-static assumption, namely, introducing the quasi-static OCV assumption, reconstructing nonlinear voltage response into a linear regression form, and utilizing a self-adaptive forgetting factor recursive least square method (Adaptive Forgetting Factor Recursive Least Squares, AFFRLS) to identify impedance parameters and OCV curves on line so as to establish an analytic mapping relation between model parameters and SOC in a reference state; S4, constructing an adaptive fractional order model integrating a linear migration mechanism and internal resistance anchoring characteristics, namely, based on a reference analytic mapping relation, introducing a linear migration factor to perform first-order approximate compensation on nonlinear drift of model parameters under the multi-stress coupling effects of temperature, aging, pressure and the like, constructing a dynamic mapping mechanism from a reference parameter library to a current multi-stress working condition to obtain adaptive parameters implying multi-physical field coupling information; S5, constructing a unified augmentation state vector comprising SOC, polarization voltage and capacity correction factors aiming at strong rigid coupling characteristics of the lithium battery, designing a time-varying noise covariance matrix to implement sparse information injection and realize decoupling evolution of a fast-varying state and a slow-varying parameter, introducing a particle filtering algorithm of risk sensitive criteria on the basis, estimating the capacity state in a long time domain, correcting the SOC in real time through voltage residual errors in a short time domain and updated capacity in the long time domain, and finally completing joint estimation of the SOC and the capacity of the lithium battery under the influence of multi-stress coupling. In the preferred technical scheme of the invention, in the S1, the