Battery state estimation method and system and medium

Abstract

The invention provides a battery state estimation method and system and a medium, and relates to the technical field of a battery management system.The method comprises the steps that a second-order RC equivalent circuit model is established, and parameters such as ohm internal resistance, polarization internal resistance and polarization capacitance are identified online through an RLS with a forgetting factor; constructing a microscopic state SRCKF channel and a macroscopic parameter SRCKF channel, performing real-time correction on an SOC influence factor and a current influence factor based on a filtering performance index, adaptively adjusting process noise and measurement noise covariance according to sub-channel logic, and introducing cross-channel cooperative self-adaption; the state channel outputs SOC and terminal voltage estimated values, the parameter channel outputs dynamic internal resistance and maximum available capacity, and SOH is calculated and output. The method improves the estimation precision and robustness under the working conditions of low SOC and large current, has good numerical stability, and is suitable for monitoring the whole life cycle of the power battery.

Description

Technical Field

[0001] This invention belongs to the field of battery management system technology, and in particular relates to a battery state estimation method, system and medium. Background Technology

[0002] In the operation of electric vehicles and energy storage systems, the accurate estimation of key parameters such as remaining battery capacity and health status by the battery management system is a core element in ensuring safe battery operation, extending battery life, and optimizing system scheduling. However, the electrochemical characteristics of power batteries are complex, and their state and parameters exhibit strong nonlinearity and time-varying behavior with charge-discharge cycles, temperature changes, and aging processes, posing challenges to high-precision estimation.

[0003] In existing battery state estimation methods, the extended Kalman filter linearizes the nonlinear system, which easily introduces linearization errors, especially in the low SOC range where estimation accuracy drops significantly. While the unscented Kalman filter approximates the nonlinear distribution through volume points, the design of these volume points can easily lead to numerical instability in the covariance matrix. Furthermore, when a single filter channel simultaneously processes state and parameter estimation, it struggles to adapt to the differentiated characteristics of rapidly changing states and slowly time-varying parameters, easily generating coupling interference and reducing estimation robustness.

[0004] To improve estimation performance, previous studies have attempted to combine RLS with Kalman filtering. However, problems remain, such as the inability to accurately match noise covariance to different SOC ranges and current conditions, the fact that key influencing factors are mostly fixed presets and difficult to dynamically optimize based on real-time filter performance, and the difficulty in effectively decoupling state and parameters with a single channel. Therefore, how to decouple state and parameter estimation through dual filtering channels and optimize noise covariance adjustment by combining real-time performance feedback has become a technical bottleneck that urgently needs to be solved. Summary of the Invention

[0005] In view of this, the present invention aims to provide a battery state estimation method, system and medium to at least solve one of the problems in the background art.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0007] Firstly, this solution discloses a battery state estimation method, including:

[0008] A second-order RC equivalent circuit model is established to determine the battery state equation, measurement equation and parameter evolution equation. A state vector is defined to characterize the polarization voltage and remaining charge, and a parameter vector is defined to characterize the dynamic internal resistance and maximum usable capacity.

[0009] The polarization resistance and polarization capacitance of the second-order RC equivalent circuit model are identified online using the recursive least squares method with a forgetting factor to obtain model parameters for subsequent filtering.

[0010] A dual adaptive square root capacitive Kalman filter framework is constructed, which includes a state square root capacitive Kalman filter channel for battery state estimation and a parameter square root capacitive Kalman filter channel for battery parameter estimation.

[0011] Define adaptive factors related to SOC range and current conditions, and correct the adaptive factors in real time based on filter performance indicators;

[0012] Based on the corrected adaptive factor, the process noise covariance and measurement noise covariance of the state square root volume Kalman filter channel and the process noise covariance and measurement noise covariance of the parameter square root volume Kalman filter channel are respectively adaptively adjusted for each channel, and a collaborative adaptive mechanism of state and parameter is introduced.

[0013] The battery terminal voltage estimate and remaining power estimate are output through the state square root volume Kalman filter channel, and the dynamic internal resistance estimate and maximum available capacity estimate are output through the parameter square root volume Kalman filter channel.

[0014] The battery health status is calculated based on the estimated maximum available capacity, and the estimated results of terminal voltage, remaining charge, maximum available capacity, and health status are output.

[0015] Furthermore, in the second-order RC equivalent circuit model:

[0016] The state vectors are U1, U2 and SOC, where U1 and U2 are the polarization voltages of the two polarization branches, respectively.

[0017] The parameter vector is R0 and Qn, where R0 is the dynamic internal resistance and Qn is the maximum available capacity;

[0018] The parameter evolution equation is a slowly time-varying process of parameters.

[0019] Furthermore, the recursive least squares method includes:

[0020] Initialize the parameter vector to be identified and the covariance matrix, and set the forgetting factor;

[0021] A regression vector is constructed based on the measured terminal voltage values ​​and the model prediction values.

[0022] Calculate the recursive least squares gain and update the vector of parameters to be identified;

[0023] Update the covariance matrix;

[0024] Physical constraints are applied to the parameters to be identified, and smoothing is performed.

[0025] Furthermore, the relationship between open-circuit voltage and SOC in the measurement equation is obtained by fitting the experimental data with a polynomial.

[0026] Furthermore, the filtering performance index is constructed based on the mean, first-order autocorrelation coefficient, and variance of the voltage error, and is used to characterize the stability of the filter and the degree of error convergence.

[0027] Furthermore, the real-time correction of the adaptive factor includes:

[0028] Voltage error, SOC, and current data are stored using a sliding window.

[0029] Calculate the filtering performance in the low SOC range, medium SOC range, high SOC range, and small current range, medium current range, and large current range respectively;

[0030] When the performance in the low SOC range is lower than the preset ratio of the performance in the high SOC range, the adaptive factor used for SOC weight adjustment is increased.

[0031] When the performance in the high current range is lower than the preset ratio of the performance in the low current range, the adaptive factor used for current weight adjustment is increased.

[0032] An adaptive factor for SOC decay control is based on performance gradient adjustment within the SOC range.

[0033] Boundary constraints are then applied to the modified adaptive factor.

[0034] Furthermore, the process noise covariance of the state square root volume Kalman filter channel is adaptively adjusted according to the SOC and current, and the process noise weight is increased in the low SOC range and under high current conditions to compensate for the model uncertainty caused by nonlinearity and polarization effects.

[0035] Furthermore, the measurement noise covariance of the state square root volume Kalman filter channel is adaptively adjusted based on the sliding window variance of the terminal voltage residual to reduce the interference of abnormal measurements on state estimation.

[0036] Furthermore, the process noise covariance of the parameter square root volume Kalman filter channel is adaptively adjusted based on the statistical characteristics of the dynamic internal resistance change and the capacity change, and the measurement noise covariance of the parameter square root volume Kalman filter channel is adaptively adjusted based on the statistical characteristics of the voltage residual during the parameter update stage.

[0037] Furthermore, the cooperative adaptive mechanism for state and parameters includes:

[0038] When the voltage residual variance of the state square root volume Kalman filter channel continuously meets the preset abnormal conditions, the process noise weight of the parameter square root volume Kalman filter channel is increased to promote parameter updates; when the capacity update magnitude of the parameter square root volume Kalman filter channel meets the preset abrupt change conditions, the process noise weight of the state square root volume Kalman filter channel is increased to adapt to the state estimation requirements.

[0039] Furthermore, the dual adaptive square root capacitive Kalman filter channels all update the covariance in square root form to maintain the positive definiteness of the covariance matrix and improve numerical stability.

[0040] Secondly, this solution discloses a battery state estimation system, including:

[0041] The data acquisition module is used to collect real-time battery current and terminal voltage data;

[0042] The parameter identification module is used to identify the parameters of the second-order RC equivalent circuit model online based on the recursive least squares method.

[0043] The influence factor correction module is used to correct adaptive factors related to SOC range and current conditions in real time based on filter performance indicators.

[0044] The dual adaptive noise adjustment module is used to calculate the process noise covariance and measurement noise covariance of the state filter channel and the parameter filter channel respectively based on the corrected adaptive factor.

[0045] The dual square root capacitive Kalman filter estimation module includes a state square root capacitive Kalman filter unit and a parameter square root capacitive Kalman filter unit, which are used for output voltage and SOC estimation results, as well as dynamic internal resistance and maximum available capacity estimation results, respectively.

[0046] The results output module is used to calculate the health status based on the maximum available capacity estimation results and output the battery terminal voltage, SOC, maximum available capacity and health status.

[0047] Furthermore, the result output module smooths the maximum available capacity estimation result before calculating the health status.

[0048] Thirdly, this solution discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0049] Compared with existing technologies, the battery state estimation method, system, and medium described in this invention have the following advantages:

[0050] (1) This invention decouples state and parameter estimation by constructing micro-state SRCKF channels and macro-parameter SRCKF channels, reducing coupling interference. Furthermore, the channel noise covariance adaptive logic can better match the rapid state changes and the slow time-varying parameter characteristics, thereby improving the estimation accuracy of SOC, terminal voltage, internal resistance and capacity.

[0051] (2) The present invention corrects the adaptive factor in real time based on the filtering performance index, and combines the optimization strategy of SOC partitioning and current partitioning to make the noise covariance adjustment more suitable for complex working conditions such as low SOC and high current, thereby enhancing the robustness and stable output capability under all working conditions.

[0052] (3) Both filtering channels in this invention update the covariance matrix in the form of square root, and maintain the positive definiteness of the matrix by means of Cholesky decomposition and QR decomposition, thereby reducing the risk of numerical divergence and improving the numerical stability of the algorithm.

[0053] (4) In this invention, RLS and dual SRCKF form a closed-loop optimization mechanism. RLS updates the key parameters of the second-order RC model in real time, and dual SRCKF tracks the changes in internal resistance and capacity during the battery aging process, so that the estimation of key states such as SOC and SOH is more suitable for the monitoring needs of the entire life cycle of power batteries. Attached Figure Description

[0054] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0055] Figure 1 This is a schematic diagram of the method described in an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram comparing the terminal voltage estimation results described in an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of the terminal voltage estimation error according to an embodiment of the present invention;

[0058] Figure 4 This is a schematic diagram comparing the SOC estimation results described in the embodiments of the present invention;

[0059] Figure 5 This is a schematic diagram of the SOC estimation error according to an embodiment of the present invention;

[0060] Figure 6 This is a schematic diagram of the SOH estimation results described in an embodiment of the present invention. Detailed Implementation

[0061] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0062] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0063] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0064] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0065] This invention addresses the problems of low estimation accuracy, poor robustness, and insufficient adaptability to all operating conditions in existing battery state estimation methods. It provides a battery state estimation method and system based on the fusion of dual adaptive square root capacitive Kalman filtering and RLS. By decoupling state and parameter estimation through dual filtering channels and combining real-time correction of influence factors and adaptive mechanism of channel noise covariance, it achieves accurate and coordinated estimation of battery SOC, SOH, terminal voltage, and key parameters.

[0066] To achieve the above objectives, the technical solution adopted by the present invention mainly includes the following steps:

[0067] 1. Establish a second-order RC equivalent circuit model: Use a second-order RC equivalent circuit model to describe the dynamic characteristics of the battery, define the state vector and parameter vector, and establish the system state equation, measurement equation and parameter evolution equation;

[0068] 2. Online parameter identification of RLS: Based on the recursive least squares method with forgetting factor, the key parameters of the second-order RC model are identified online, providing an accurate model foundation for the dual SRCKF;

[0069] 3. Construction of Dual Adaptive SRCKF Framework: Construct two independent channels, a micro (state) SRCKF and a macro (parameter) SRCKF, define adaptive factors related to state and parameters, and adjust the factors in real time based on filtering performance indicators;

[0070] 4. Adaptive adjustment of noise covariance for each channel: To address the differentiated requirements of the two SRCKF channels, adaptive logic for Q and R is designed separately, and a collaborative adaptive mechanism of state and parameters is introduced to dynamically update the noise covariance.

[0071] 5. Dual SRCKF Joint Estimation: State of Charge (SOC) and terminal voltage are estimated using the state SRCKF, while internal resistance and maximum available capacity are estimated using the parameter SRCKF.

[0072] 6. SOH Calculation and Output: SOH is calculated based on the estimated maximum available capacity, and the estimated results are smoothed and output in real time.

[0073] Specifically, the establishment of the second-order RC equivalent circuit model is as follows:

[0074] The second-order RC equivalent circuit model includes the ohmic internal resistance. Two RC polarization branches and open-circuit voltage ,in:

[0075] State vector , , These are the polarization voltages of the two RC branches, and SOC is the remaining charge.

[0076] Parameter vector , For dynamic internal resistance, Maximum available capacity;

[0077] Equations of state: , where A is the state transition matrix, B is the input matrix, Ts is the sampling interval, I(k) is the real-time current, and w_state(k) is the state process noise;

[0078] Measurement equation: Where C = [-1, -1, 0], The experimental data were obtained by fitting the data with an 8th-order polynomial. Noise for state measurement;

[0079] Parameter evolution equation: The noise is for the parameter process, assuming the parameter is a slowly time-varying process.

[0080] Specifically, in the online parameter identification of RLS, the following is a detailed explanation:

[0081] The RLS algorithm with a forgetting factor is used to identify second-order RC models. Parameters, specific process:

[0082] 1. Initialize the parameter vector θ = [ The covariance matrix is ​​P=1000·I, and the forgetting factor is λ=0.995. For the first / second polarization branch resistors and capacitors;

[0083] 2. Calculating the regression vector based on numerical differentiation The partial derivatives are calculated using the method of small perturbations;

[0084] 3. Calculate the RLS gain moment: ;

[0085] 4. Update parameter estimates ,in This is the measured terminal voltage. Predict the terminal voltage for the model;

[0086] 5. Update the covariance matrix ;

[0087] Specifically, in the dual adaptive SRCKF framework and noise covariance adjustment, the details are as follows:

[0088] The dual adaptive SRCKF framework includes micro (state) SRCKF and macro (parameter) SRCKF channels, adjusting the noise covariance through a combination of channel-specific adaptation and cross-channel collaboration.

[0089] (1) Adaptive Q and R behavior of microscopic (state) SRCKF

[0090] The SRCKF state is responsible for SOC and terminal voltage estimation. (Process noise) and The adaptive logic for (measuring noise) is as follows:

[0091] Adaptive: Adjustments are made based on the SOC level and current magnitude, introducing adaptive parameters. , , When SOC < 0.2, enlarge To adapt to the high slope characteristics of OCV-SOC; the larger the current, Further enlargement To compensate for polarization effect errors, The formula for calculation is:

[0092]

[0093] The covariance of the basic process noise in the state channel. For SOC weight adaptive factor, The SOC attenuation adaptive factor. This is the current weight adaptive factor. For real-time charging and discharging current, This is the battery's rated maximum current.

[0094] Adaptive: Based on the sliding window variance adjustment of the voltage residual, the variance σ² is calculated using the voltage residuals of 50 sampling points.

[0095]

[0096] like ,but ;otherwise = Abnormal interference can be avoided by reducing the measurement weight.

[0097] (2) Adaptive Q and R of macroscopic (parameter) SRCKF

[0098] The parameter SRCKF is responsible for estimating internal resistance and capacitance. (Process noise) and The adaptive logic for (measuring noise) is as follows:

[0099] Adaptive: Adjusts based on the variance of the parameter change rate, calculates the variance of the internal resistance change over 10 adjacent sampling points. and variance of capacity change ,

[0100]

[0101]

[0102] For the basic process noise covariance of the parameter channel, The average variance of the rate of change of internal resistance / capacitance. / The variance of the change in internal resistance / capacitance. This serves as a reference variance for parameter stability.

[0103] Adaptive: Adjusts the variance based on the voltage residuals during parameter updates, using a 50-sliding-window residual calculation for variance. ,

[0104]

[0105] Update the voltage residual variance for the parameters. .

[0106] (3) Cooperative adaptation of state and parameters

[0107] To achieve cross-channel optimization, collaborative logic is designed as follows:

[0108] When the voltage residual variance of state SRCKF is sampled for 5 consecutive points When this occurs, it is determined to be a parameter mismatch, prompting parameter updates;

[0109]

[0110] Scaling factor for the enhanced parametric process noise covariance. .

[0111] When the parameter SRCKF At that time, it was determined to be a sudden change in battery characteristics, and The scaling factor is increased by an additional 1.1 times to better suit state estimation requirements.

[0112]

[0113] To enhance the state-process noise covariance, Scaling factor for battery rated capacity .

[0114] (4) Real-time correction of adaptive factors

[0115] Define filter performance metrics

[0116] Filtering performance indicators The mean, first-order autocorrelation coefficient, and variance of the combined voltage residual are used to quantify the stability and error convergence of the filter. The value range is [0,1], and the closer it is to 1, the better the filtering performance.

[0117]

[0118] Mean of voltage residual, The normalization coefficient for the mean sub-indicator. The first-order autocorrelation coefficient of the voltage residual. The normalized coefficient is the sub-index of the autocorrelation coefficient. This serves as a reference threshold for the variance sub-index. Let V be the variance of the voltage residual.

[0119] The adaptive factor is adjusted based on this indicator:

[0120] (SOC weighting factor) correction: adapting to nonlinearity in low SOC range

[0121] Its core function is to amplify the process noise covariance in the low SOC range. This compensates for the nonlinear error caused by the high slope of the OCV-SOC curve. The correction logic is based on a comparison between performance in the low SOC range and performance in the high SOC range; if the performance in the low SOC range is insufficient, the error is increased. .

[0122] (1) SOC interval partitioning and partition performance calculation

[0123] First, divide the SOC data within the window into three intervals, and calculate the filtering performance for each interval:

[0124]

[0125] , , For datasets in the low / medium / high SOC range, , The filtering performance is measured in the low / high SOC range.

[0126] (2) Correction formula

[0127] When the performance in the low SOC range is less than 80% of the performance in the high SOC range (preset ratio), the gradient descent method is used to increase the efficiency. Otherwise, remain unchanged:

[0128]

[0129] The SOC weighting factor is corrected at time k. This is the historical value at time k-1. for Learning rate =0.01, The preset threshold ratio for performance comparison =0.8.

[0130] 2. (Current weighting factor) correction: Compensating for large current polarization effects

[0131] Its core function is to amplify the process noise covariance under high current conditions. This compensates for model errors caused by intensified polarization effects. The correction logic is based on a comparison of performance in the high-current range and performance in the low-current range; if the high-current performance is insufficient, the error is increased. .

[0132] (1) Current range division and zone performance calculation

[0133] First, divide the current data within the window into three intervals based on their absolute values ​​(to avoid the influence of charging and discharging signs), and then calculate the filtering performance of each interval:

[0134]

[0135] , , Data sets for small / medium / large current ranges. Rated current, , The filtering performance is measured in the small / large current range.

[0136] (2) Correction formula

[0137] When the performance in the high-current range is lower than 80% (preset ratio) of the performance in the low-current range, the gradient descent method is used to increase the current. Otherwise, remain unchanged:

[0138]

[0139] This is the current weighting factor corrected at time k. This is the historical value at time k-1. for Learning rate =0.02, The preset threshold ratio for performance comparison =0.8.

[0140] (SOC decay factor) correction: balancing the performance gradient across the SOC range.

[0141] Its core function is regulation Decay rate as SOC increases: The larger, The decay rate increases with increasing SOC; The smaller the value, the slower the decay. The correction logic is based on the "performance gradient between SOC intervals" to avoid abrupt performance changes between low and medium SOC intervals, and between medium and high SOC intervals.

[0142] (1) Calculation of performance gradient in SOC interval

[0143] First, calculate the performance difference (gradient) in the low → medium and medium → high SOC ranges to reflect the smoothness of performance changes with SOC:

[0144]

[0145]

[0146] This represents the performance gradient in the low to medium SOC range. This represents the performance gradient in the medium to high SOC range. This represents the filtering performance in the SOC range.

[0147] (2) Correction formula

[0148] Adjust according to the anomalies of the two gradients respectively. :

[0149]

[0150] The SOC decay factor is corrected at time k. =0.005 is learning rate, =0.2 is the upper limit threshold for the low → medium SOC gradient. =-0.1 is the lower limit threshold for the medium to high SOC gradient.

[0151] 4. Modified boundary constraints: to prevent factor runaway.

[0152] The corrected adaptive factor must be forced to meet the boundary range between physical meaning and engineering reality to prevent the factor from becoming too large / too small due to abnormal data, thereby causing problems. abnormal: ∈[0.1,2.0]、 ∈[0.05,0.5]、 ∈[0.1,1.0]. If the corrected factor exceeds the boundary, the corresponding boundary value is taken as the final result, for example... When the revised value is 2.1, it should be forcibly set to 2.0. This constraint prevents the factor from increasing abnormally. Excessive magnification or abnormal reduction leads to Insufficient adaptation ensures the stability of noise covariance adjustment.

[0153] Specifically, the joint estimation process using two SRCKFs is as follows:

[0154] State SRCKF estimation

[0155] 1. Initialization Phase

[0156]

[0157] Initial estimate vector of state channels, , These are the initial estimates of the capacitor voltages in the two polarization branches. This is the initial estimate of SOC. Let the initial covariance square root matrix be... Let be the initial state covariance matrix.

[0158] 2. Adaptive noise covariance calculation

[0159] Process noise covariance of the state channel and measurement noise covariance Based on SOC, current and voltage residual adaptive adjustment:

[0160]

[0161] The covariance of the basic process noise in the state channel. For SOC weight adaptive factor, The SOC attenuation adaptive factor. This is the current weight adaptive factor. For real-time charging and discharging current, This is the battery's rated maximum current. This is the battery's rated maximum current. The variance of the voltage residual sliding window is given.

[0162] 3. Volume point generation

[0163]

[0164]

[0165] For the j-th state volume point, These are the standard volume point basis vectors. The dimension of the state vector; for 3D identity matrix.

[0166] 4. Time Update

[0167] (1) Volume point state transfer

[0168]

[0169] The state transition matrix A and the input matrix B are:

[0170]

[0171] Here is the state transition matrix. For the input matrix, The sampling time interval, For battery coulomb efficiency, This is the maximum usable capacity of the battery. / For the first / second polarization branch resistance, / For the first / second polarization branch capacitors.

[0172] (2) Prior state estimation

[0173]

[0174] These are prior state estimates. The weights are the state volume points.

[0175] (3) Covariance Square Root Update

[0176]

[0177]

[0178] Let be the prior square root of covariance matrix. The volume point matrix after time update. Let be the square root matrix of the process noise covariance.

[0179] 5. Measurement Update

[0180] (1) Observation volume point transfer

[0181]

[0182] Where the observation matrix Open circuit voltage It is an 8th-order polynomial:

[0183]

[0184] For the observed predicted value of the j-th volume point, The prior SOC value of the j-th volume point For ohmic internal resistance, These are the polynomial fitting coefficients, obtained through experimental calibration.

[0185] (2) Observation, prediction and information calculation

[0186]

[0187]

[0188] This is the a priori predicted value of the terminal voltage. For voltage information, This is the measured value of the terminal voltage at time k+1.

[0189] (3) Covariance and Kalman gain

[0190]

[0191]

[0192]

[0193] To observe the square root of the covariance matrix, To observe the volume point matrix, To measure the square root of the noise covariance, The state-observation cross-covariance matrix, For the state channel Kalman gain, It is the inverse of the transpose matrix.

[0194] (4) State and covariance update

[0195]

[0196] Output the optimal state estimate. Let be the posterior covariance square root matrix.

[0197] Parameter SRCKF estimation

[0198] 1. Initialization Phase

[0199]

[0200] The initial covariance square root matrix is ​​the parameter.

[0201] 2. Adaptive noise covariance calculation

[0202]

[0203] For the basic process noise covariance of the parameter channel, The average variance of the rate of change of internal resistance / capacitance. / The variance of the change in internal resistance / capacitance. For parameter stability, refer to the variance. Update the voltage residual variance for the parameters.

[0204] 3. Volume point generation

[0205]

[0206] For the j-th parameter volume point, For the dimension of the parameter vector, The standard volume point basis vector is the parameter.

[0207] 4. Time Update

[0208] (1) Transfer of volume point parameters

[0209]

[0210] (2) Prior parameter estimation

[0211]

[0212] (3) Covariance Square Root Update

[0213]

[0214] 5. Measurement Update

[0215] (1) Observation volume point transfer

[0216]

[0217] These are the observed and predicted values ​​for the parameter volume points. The Jacobian matrix is ​​used to measure the parameters of the equation.

[0218] (2) Observation, prediction and news

[0219]

[0220]

[0221] This is for information about the parameter channel.

[0222] (3) Kalman gain and parameter update

[0223]

[0224] For the parameter channel Kalman gain, This is for the optimal parameter estimation.

[0225] Specifically, the SOH calculation and result output are as follows:

[0226] 1. Calculation :

[0227]

[0228] This refers to the battery's rated capacity.

[0229] 2. Output results: Real-time output , , and Store historical data on factor correction and filtering performance.

[0230] Three improved A-DSRCKF-RLS algorithms (the improved algorithm of this invention patent) are compared. The results of A-DSRCKF and A-DSRCKF-RLS are compared as follows: Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) of terminal voltage and SOC are compared:

[0231]

[0232] On the other hand, the system architecture corresponding to this solution includes the following modules:

[0233] 1. Data Acquisition Module: Employs high-precision current sensors (accuracy ≤ 0.05%) and voltage sensors (accuracy ≤ 0.1%) to acquire real-time battery current and terminal voltage data at 0.1s sampling intervals. The data is then filtered and preprocessed before being transmitted to subsequent modules.

[0234] 2. Parameter Identification Module: Based on the RLS algorithm, it receives raw data from the data acquisition module and error feedback from the dual SRCKF estimation module, and identifies the parameters of the second-order RC model online. The parameters are output to the dual SRCKF estimation module;

[0235] 3. Impact Factor Correction Module: Stores filter error data, calculates filter performance indicators, and corrects them in real time based on regional performance differences (SOC region, current region). , An adaptive factor is output to the dual adaptive noise adjustment module.

[0236] 4. Dual Adaptive Noise Adjustment Module: Receives the corrected adaptive factor, real-time SOC, and current data, and calculates according to the channel-specific adaptive logic. , The result is converted to square root form through Cholesky decomposition and output to the dual SRCKF estimation module.

[0237] 5. Dual SRCKF estimation module: includes a state estimation SRCKF unit and a parameter estimation SRCKF unit, which respectively receive model parameters from the parameter identification module and covariance data from the noise adjustment module, complete state and parameter estimation, and output to the result fusion output module;

[0238] 6. Result Fusion Output Module: Smoothes the capacity estimate, calculates the State of Health (SOH), outputs the estimate via CAN bus or serial port, supports LCD screen visualization, and reserves a data upload interface.

[0239] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A battery state estimation method, characterized in that, include: A second-order RC equivalent circuit model is established to determine the battery state equation, measurement equation and parameter evolution equation. A state vector is defined to characterize the polarization voltage and remaining charge, and a parameter vector is defined to characterize the dynamic internal resistance and maximum usable capacity. The polarization resistance and polarization capacitance of the second-order RC equivalent circuit model are identified online using the recursive least squares method with a forgetting factor to obtain model parameters for subsequent filtering. A dual adaptive square root capacitive Kalman filter framework is constructed, which includes a state square root capacitive Kalman filter channel for battery state estimation and a parameter square root capacitive Kalman filter channel for battery parameter estimation. Define adaptive factors related to SOC range and current conditions, and correct the adaptive factors in real time based on filter performance indicators; Based on the corrected adaptive factor, the process noise covariance and measurement noise covariance of the state square root volume Kalman filter channel and the process noise covariance and measurement noise covariance of the parameter square root volume Kalman filter channel are respectively adaptively adjusted for each channel, and a collaborative adaptive mechanism of state and parameter is introduced. The battery terminal voltage estimate and remaining power estimate are output through the state square root volume Kalman filter channel, and the dynamic internal resistance estimate and maximum available capacity estimate are output through the parameter square root volume Kalman filter channel. The battery health status is calculated based on the estimated maximum available capacity, and the estimated results of terminal voltage, remaining charge, maximum available capacity, and health status are output.

2. The battery state estimation method according to claim 1, characterized in that, In the second-order RC equivalent circuit model: The state vectors are U1, U2 and SOC, where U1 and U2 are the polarization voltages of the two polarization branches, respectively. The parameter vector is R0 and Qn, where R0 is the dynamic internal resistance and Qn is the maximum available capacity; The parameter evolution equation is a slowly time-varying process of parameters.

3. The battery state estimation method according to claim 1, characterized in that, The recursive least squares method includes: Initialize the parameter vector to be identified and the covariance matrix, and set the forgetting factor; A regression vector is constructed based on the measured terminal voltage values ​​and the model prediction values. Calculate the recursive least squares gain and update the vector of parameters to be identified; Update the covariance matrix; Physical constraints are applied to the parameters to be identified, and smoothing is performed.

4. The battery state estimation method according to claim 1, characterized in that, The relationship between open-circuit voltage and SOC in the measurement equation was obtained by fitting experimental data with a polynomial.

5. The battery state estimation method according to claim 1, characterized in that, The filtering performance metrics are constructed based on the mean, first-order autocorrelation coefficient, and variance of the voltage error, and are used to characterize the stability of the filter and the degree of error convergence.

6. The battery state estimation method according to claim 1, characterized in that, Real-time correction of the adaptive factor includes: Voltage error, SOC, and current data are stored using a sliding window. Calculate the filtering performance in the low SOC range, medium SOC range, high SOC range, and small current range, medium current range, and large current range respectively; When the performance in the low SOC range is lower than the preset ratio of the performance in the high SOC range, the adaptive factor used for SOC weight adjustment is increased. When the performance in the high current range is lower than the preset ratio of the performance in the low current range, the adaptive factor used for current weight adjustment is increased. An adaptive factor for SOC decay control is based on performance gradient adjustment within the SOC range. Boundary constraints are then applied to the modified adaptive factor.

7. The battery state estimation method according to claim 1, characterized in that, The process noise covariance of the parameter square root volume Kalman filter channel is adaptively adjusted based on the statistical characteristics of the dynamic internal resistance change and the capacity change, and the measurement noise covariance of the parameter square root volume Kalman filter channel is adaptively adjusted based on the statistical characteristics of the voltage residual during the parameter update stage.

8. The battery state estimation method according to claim 1, characterized in that, The cooperative adaptive mechanism of state and parameters includes: When the voltage residual variance of the state square root volume Kalman filter channel continuously meets the preset abnormal conditions, the process noise weight of the parameter square root volume Kalman filter channel is increased to promote parameter updates; when the capacity update magnitude of the parameter square root volume Kalman filter channel meets the preset abrupt change conditions, the process noise weight of the state square root volume Kalman filter channel is increased to adapt to the state estimation requirements.

9. A battery state estimation system, characterized in that, include: The data acquisition module is used to collect real-time battery current and terminal voltage data; The parameter identification module is used to identify the parameters of the second-order RC equivalent circuit model online based on the recursive least squares method. The influence factor correction module is used to correct adaptive factors related to SOC range and current conditions in real time based on filter performance indicators. The dual adaptive noise adjustment module is used to calculate the process noise covariance and measurement noise covariance of the state filter channel and the parameter filter channel respectively based on the corrected adaptive factor. The dual square root capacitive Kalman filter estimation module includes a state square root capacitive Kalman filter unit and a parameter square root capacitive Kalman filter unit, which are used for output voltage and SOC estimation results, as well as dynamic internal resistance and maximum available capacity estimation results, respectively. The results output module is used to calculate the health status based on the maximum available capacity estimation results and output the battery terminal voltage, SOC, maximum available capacity and health status.

10. A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the method as claimed in any one of claims 1 to 8.

Images