A method and system for estimating state of charge of a power lithium battery

Abstract

The present invention provides a method and system for estimating the state of charge of a power lithium battery, establishes a third-order RC equivalent circuit model of a power lithium battery, determines the equation group of the model; transforms the equation group of the third-order RC equivalent circuit model to obtain a transfer function, converts the transfer function into a differential equation, and uses a forgetting factor recursive least squares method to derive the circuit parameters of the third-order RC equivalent circuit model; and in the derivation process, uses a simulated annealing algorithm to optimize the forgetting factor; based on the obtained circuit parameters, uses an adaptive H infinite unscented Kalman filter algorithm to estimate the state of charge of the power lithium battery. The present invention can significantly improve the accuracy of online parameter identification.

Description

Technical Field

[0001] The present invention belongs to the technical field of power lithium battery SOC estimation, and in particular relates to a power lithium battery state of charge estimation method and system. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The Battery Management System (BMS) is one of the three core components of electric vehicles (i.e., batteries, motors, and electronic controls). The battery's state of charge (SOC) parameter is the main indicator for determining the health of power lithium batteries. Accurate SOC data can significantly extend the battery's service life, optimize the battery pack's control strategy to better utilize the power battery's performance, and improve the reliability of electric vehicles. However, the battery's state of charge parameters cannot currently be obtained through direct measurement. The state of charge can only be estimated based on the battery voltage, current, internal resistance, and the battery's environment. Therefore, the charge parameters of power lithium batteries have always been a research hotspot for BMS system manufacturers.

[0004] Affected by the accuracy of the battery model and the stability of the parameters, the accuracy of the battery state of charge estimation of power lithium after multiple working cycles cannot meet the application needs. This often leads to inaccurate mileage indication of new energy vehicles, causing new energy vehicles to break down on the road. The more traffic is congested during holidays, the more serious the breakdown phenomenon becomes, causing huge safety hazards. Although the classic Thevenin battery model currently used by most BMS manufacturers also includes the polarization of the battery, its accuracy is not high and it is difficult to accurately obtain the chemical state of the battery. Although the prediction accuracy of the model can be improved by the joint improvement of the second-order RC network and the intelligent algorithm, it still cannot fundamentally solve the bottleneck problem of the error in the model itself. Although the high-order RC network model can improve the estimation accuracy of the state of charge within a limited range, the data calculation is too complicated, and the computing and storage resources of the processor are extremely high, making it difficult to apply in practice.

[0005] According to the existing public technology and market research, most of the implementation schemes for SOC estimation of power lithium-ion batteries are mainly: ampere-hour integration method, open circuit voltage method, neural network method, Kalman filter method, etc. Among them, the ampere-hour integration method requires a small step size to ensure numerical stability, introduces numerical errors, and may be distorted in long-term simulation. Rigid problems require a smaller step size, and the complexity rises sharply in high-dimensional problems. It is not suitable for non-smooth systems. The disadvantages of the neural network method include big data requirements, black box nature, overfitting risks, high computational complexity, difficulty in hyperparameter adjustment, unsuitable for small samples, and possible bias problems. The Unscented Kalman Filter (UKF) algorithm achieves a more accurate estimation of SOC parameters when the power lithium battery is just put into use, but as the battery usage time increases, due to the irregularity of charging and discharging, the parameters and state changes inside the battery show strong nonlinearity, especially after overcharging, undercharging, and over-discharging operations on the power lithium battery, the SOC parameter estimation accuracy of the Kalman filter will be significantly reduced. Therefore, there is an urgent need to develop a battery model that does not require high processor computing resources and can meet the requirements of an SOC estimation algorithm that corrects the nonlinearity of power lithium batteries caused by various irregular operations. Summary of the invention

[0006] In order to solve the above problems, the present invention proposes a method and system for estimating the state of charge of a power lithium battery. The present invention can significantly improve the accuracy of online parameter identification.

[0007] According to some embodiments, the present invention adopts the following technical solutions:

[0008] A method for estimating the state of charge of a power lithium battery comprises the following steps:

[0009] Establish a third-order RC equivalent circuit model of a power lithium battery and determine the equation group of the model;

[0010] The equation group of the third-order RC equivalent circuit model is transformed to obtain the transfer function, which is then converted into a differential equation, and the circuit parameters of the third-order RC equivalent circuit model are derived using the forgetting factor recursive least squares method.

[0011] And in the derivation process, the simulated annealing algorithm is used to optimize the forgetting factor;

[0012] Based on the obtained circuit parameters, the adaptive H infinite unscented Kalman filter algorithm is used to estimate the state of charge of the power lithium battery.

[0013] As an optional implementation, the third-order RC equivalent circuit model is:

[0014]

[0015] Among them, U ocrepresents the open circuit voltage of the power supply, R0 represents the ohmic internal resistance of the battery, R1 represents the electrochemical polarization internal resistance, R2 represents the electrochemical polarization internal resistance, R3 represents the concentration difference polarization internal resistance, C1 represents the electrochemical polarization capacitance, C2 represents the electrochemical polarization capacitance, C3 represents the concentration difference polarization capacitance, U t represents the terminal voltage, I is the current, and U1, U2, and U3 represent the three loop terminal voltages of capacitors C1, C2, and C3 respectively;

[0016] The state of charge SOC is determined using the ampere-hour integration method.

[0017] Where t0 is the start time of discharge, t is the current time, η is the coulombic efficiency, which is measured by calculating the ratio of the discharge capacity to the charge capacity of the battery, Q C It represents the rated capacity of the battery. After discretizing the above, we can get the state equation and output equation of the model.

[0018] As an optional implementation, the specific process of deriving the circuit parameters of the third-order RC equivalent circuit model using the forgetting factor recursive least squares method includes:

[0019] The forgetting factor λ is introduced into the recursive least squares parameter identification to change in real time according to the current. When the current fluctuation is greater than the set value, λ is minimized as much as possible, and vice versa, to enhance the stability of the identification process. The least squares parameter identification equation is:

[0020]

[0021] Among them, K is the correction weight value, Φ is the pre-measurement vector, is the parameter to be identified before decoding, and P(k) is the covariance matrix.

[0022] As an optional implementation method, in the specific process of deriving the circuit parameters of the third-order RC equivalent circuit model using the forgetting factor recursive least squares method, the difference between the open-circuit voltage and the terminal voltage at each moment and the battery current value are used as input values, and the covariance matrix and the weight factor matrix are corrected through continuous iterations, and the parameter matrix is ​​calculated. The parameters in the matrix are the coefficients of each link in the Z domain of the transfer function of the equivalent circuit model. According to the third-order RC equivalent circuit model of the battery, the battery parameters are identified.

[0023] As an optional implementation, the specific process of optimizing the forgetting factor using the simulated annealing algorithm includes taking the minimum terminal voltage error as the optimization goal, performing iterative optimization with the forgetting factor as the optimization variable within the optional range of the set forgetting factor, and making the best estimate of the parameters.

[0024] As an optional implementation, the specific process of estimating the state of charge of the power lithium battery using the adaptive H infinite unscented Kalman filter algorithm includes:

[0025] Initialize parameters;

[0026] Apply singular value decomposition to the covariance matrix to calculate the sampling points and corresponding weight values;

[0027] Time update, the state variables are needed to calculate the expected value and error covariance matrix of the prior state estimate variables;

[0028] Measure updates, calculate output predictions and covariance matrices;

[0029] Correct the posterior state estimate;

[0030] The noise covariance matrix is ​​corrected using an adaptive algorithm.

[0031] As a further implementation method, the specific process of correcting the posterior state estimate includes obtaining, according to the characteristics of the lithium battery state equation, Among them, γ 2 Take the largest eigenvalue, By regulating γ 2 Improve the robustness to non-Gaussian noise and correct the ill-conditioned matrix encountered when calculating the covariance matrix, where The posterior estimate of the covariance, is the posterior estimate of the covariance, γ 2 is the cost function in the H-infinity algorithm, P xy,k is the mutual covariance between the sampling point observation prediction and the state prediction, and R is the noise covariance matrix.

[0032] As a further implementation method, the specific process of correcting the noise covariance matrix using the adaptive algorithm includes: calculating the terminal voltage error value output at the model k moment, calculating the approximate value of the terminal voltage error covariance of the model k moment, and updating the noise coefficient Q by obtaining the covariance of the terminal voltage difference value. k and R k .

[0033] A power lithium battery charge state estimation system, comprising:

[0034] A model building module is configured to establish a third-order RC equivalent circuit model of a power lithium battery and determine a set of equations of the model;

[0035] a parameter identification module, configured to transform the equation group of the third-order RC equivalent circuit model to obtain a transfer function, transform the transfer function into a differential equation, and derive circuit parameters of the third-order RC equivalent circuit model using a forgetting factor recursive least squares method;

[0036] A forgetting factor optimization module is configured to optimize the forgetting factor using a simulated annealing algorithm during the derivation process;

[0037] The state of charge estimation module is configured to estimate the state of charge of the power lithium battery based on the obtained circuit parameters using an adaptive H infinite unscented Kalman filter algorithm.

[0038] An electronic device comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the above method are completed.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention proposes an equivalent circuit model of a third-order RC network of a power lithium battery, and uses a simulated annealing optimization algorithm to optimize the forgetting factor in the forgetting factor least squares method (FFRLS). Taking the minimum terminal voltage error as the optimization goal, the appropriate forgetting factor is dynamically optimized in real time within the optional range of the set forgetting factor, avoiding the data saturation problem caused by poor forgetting factor selection, thereby improving the accuracy of online parameter identification.

[0041] The present invention converts the state estimation expectation value and the error covariance matrix into a third-order form based on the third-order RC network equivalent circuit model of the power lithium battery. This can not only fully take into account the polarization characteristics of the battery, but also can reflect the state information of the lithium-ion battery in real time and accurately, making the dynamic and static characteristics of the power lithium battery charge and discharge more feasible.

[0042] The present invention also adopts a simulated annealing optimization algorithm in parameter identification to optimize the forgetting factor in the forgetting factor least squares method (FFRLS). Then, the H-infinity filtering algorithm is introduced for the noise in the UKF algorithm, and the non-Gaussian noise is filtered by minimizing the cost function, and the singular value decomposition (Singularly Valuable Decomposition, SVD) method is used to solve the problem of the ill-conditioned matrix generated by the PKCholesky decomposition of the covariance matrix. Finally, an adaptive link is added to solve the problem of the uncertainty of the initial values ​​of the state noise covariance matrix and the measurement noise covariance matrix. It can effectively improve the SOC estimation accuracy of the power lithium battery, can be better applied to nonlinear time-varying systems, and effectively enhance the robustness of the estimation algorithm, thereby ensuring the accuracy and feasibility of the lithium battery SOC estimation, and further improving the reliability of the new energy vehicle.

[0043] The present invention can not only comprehensively consider the polarization characteristics of the battery and has the dual advantages of capacitance and resistance, but also can accurately reflect the state information of the lithium-ion battery in real time, and can fully express the dynamic and static characteristics of the lithium battery charging and discharging, fundamentally solving the bottleneck problem of the error existing in the model itself.

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0046] Figure 1 This is an algorithm flow chart of a power lithium battery SOC estimation method of this embodiment;

[0047] Figure 2 The battery SOC prediction of different model algorithms of this embodiment;

[0048] Figure 3 is the SOC prediction error of different model algorithms in this embodiment;

[0049] Figure 4 are the voltage values ​​of different model algorithm terminals in this embodiment;

[0050] Figure 5 is the voltage value error of different model algorithms in this embodiment. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0053] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0054] Embodiment 1

[0055] To facilitate understanding by those skilled in the art, the entire technical solution of this embodiment is described below in steps.

[0056] A method for estimating SOC of a power lithium battery, such as Figure 1 As shown, the following steps are included:

[0057] Step 1: Simulate and experimentally design the third-order RC equivalent circuit model of the lithium battery and construct a set of equations as shown in formula (1).

[0058]

[0059] Where U oc U represents the open circuit voltage of the power supply, R0 represents the ohmic internal resistance of the battery, R1 represents the electrochemical polarization internal resistance, R2 represents the electrochemical polarization internal resistance, R3 represents the concentration difference polarization internal resistance, C1 represents the electrochemical polarization capacitance, C2 represents the electrochemical polarization capacitance, and C3 represents the concentration difference polarization capacitance. t represents the end voltage, I is the current, and U1, U2, and U3 represent the three loop end voltages of capacitors C1, C2, and C3 respectively.

[0060] The SOC determined by the ampere-hour integration method is as follows:

[0061]

[0062] Where t0 is the discharge start time, t is the current time, η is the coulombic efficiency which is measured by calculating the ratio of the discharge capacity to the charge capacity of the battery, and Q is C It indicates the rated capacity of the battery.

[0063] By discretizing the above equation, we can obtain the state equation and output equation of the model.

[0064]

[0065] Where Ts is the sampling time, ω(k) is the process noise, and it obeys a Gaussian distribution with a mean of 0 and a covariance of Q(k).

[0066] U t (k) = U oc (SOC(k))-R0I(k)-U1(k)-U2(k)-U3(k)+v(k) (4)

[0067] Where v(k) is the observation noise, which obeys a Gaussian distribution with a mean of 0 and a covariance of R(k).

[0068] Step 2: Use the Forgetting Factor Recursive Least Squares (FFRLS) method to identify parameters.

[0069] The least squares parameter identification is as follows:

[0070]

[0071] In the above formula, K is the correction weight value, Φ is the pre-measurement vector, is the parameter to be identified before decoding, and P(k) is the covariance matrix.

[0072] It is worth mentioning that the forgetting factor λ introduced in this embodiment on the basis of the recursive least squares method is adjusted in real time according to the change of current size. When the current fluctuation is large, λ decreases, and vice versa. Through experimental tests, it is found that this scheme can effectively enhance the stability of the identification process and improve the output accuracy.

[0073] The formula is:

[0074] P -1 (k+1)=λP -1 (k)+Φ(k+1)Φ T (k+1) (7)

[0075] So we get the formula:

[0076]

[0077] In FFRLS, the difference between the open-circuit voltage and the terminal voltage at each moment and the battery current value are taken as input values. The covariance matrix and the weight factor matrix are corrected through continuous iteration, and the parameter matrix is ​​calculated. The parameters in the matrix are the coefficients of each link in the Z domain of the transfer function of the equivalent circuit model. According to the third-order RC equivalent circuit model of the battery, the battery parameters are identified.

[0078] The specific implementation process includes the following steps:

[0079] Step 2-1, perform Laplace transform on the set of equations constructed by analyzing the third-order RC equivalent circuit model of the lithium battery:

[0080]

[0081] The transfer function of the system is obtained:

[0082]

[0083] in:

[0084]

[0085] Step 2-2, use the bilinear transformation method to discretize the transfer function of the system, take the open circuit voltage and terminal voltage as the output of the system, and the current value as the input value of the system, let: We can get:

[0086]

[0087] With the open circuit voltage and terminal voltage as the output of the system and the current value as the input value of the system, equation (12) is transformed into a differential equation:

[0088] U(k)=α1U(k-1)+α2U(k-2)+α3U(k-3)+β0I(k)+β1I(k-1)+β2I(k-2)+β3I(k-3) (13)

[0090] Let: φ(k)=[U(k-1)U(k-1)U(k-1)I(k)I(k-1)I(k-2)I(k-3)] T

[0091]

[0092] Then the system can be expressed as:

[0093] Analyze the above parameter values ​​and perform inverse transformation on the transfer function in the Z domain, and let:

[0094]

[0095] Substitute into formula (12):

[0096]

[0097] By coefficient comparison:

[0098]

[0099] Combining the above formula derivation, the resistance, capacitance and other parameters of the third-order RC equivalent circuit model can be identified.

[0100] Step 3: Use simulated annealing algorithm to optimize the forgetting factor.

[0101] In order to optimize the algorithm disclosed in the embodiment and enhance its application adaptability in the field of power lithium batteries, the present invention introduces a forgetting algorithm based on the forgetting factor recursive least squares algorithm. The forgetting algorithm takes the minimum terminal voltage error as the optimization goal, and dynamically optimizes the appropriate forgetting factor in real time within the optional range of the set forgetting factor, avoiding the "data saturation" problem caused by poor forgetting factor selection, thereby improving the accuracy of online parameter identification.

[0102] The objective function is:

[0103]

[0104] Where U t(k) is the measured value of the terminal voltage, is the terminal voltage value estimated by the recursive least squares method, and the objective function is the absolute value of the terminal voltage error.

[0105] Taking the forgetting factor λ as the optimization variable, iterative optimization is performed with the goal of minimizing J, and then the parameters are optimally estimated.

[0106] Step 4: Estimate the battery SOC using an adaptive H infinite unscented Kalman filter algorithm.

[0107] In order to solve the problem that the noise of UKF algorithm must be white noise, the H-infinity filtering algorithm is introduced to filter out non-Gaussian noise by minimizing the cost function, and the singular value decomposition (SVD) method is used to solve the problem of ill-conditioned matrix generated by PKCholesky decomposition of covariance matrix. Finally, the adaptive link is added to solve the problem of uncertain initial values ​​of state noise covariance matrix and measurement noise covariance matrix. The identification accuracy is improved and the accurate estimation of power battery state of charge is achieved.

[0108] In particular, this embodiment also improves some problems in the algorithm:

[0109] (1) Use the singularly valuable decomposition (SVD) method to solve the problem of ill-conditioned matrix generated by PKCholesky decomposition of the covariance matrix.

[0110] (2) An adaptive method is used to solve the problem of uncertain initial values ​​of the state noise covariance matrix and the measurement noise covariance matrix.

[0111] The SVD principle and the algorithm improvement method of the present invention are explained below with specific examples.

[0112] Assume that matrix A is an m×n matrix, and the SVD decomposition definition of matrix A is:

[0113]

[0114] The matrices U and V are orthogonal matrices, and the column vector u of U i is the matrix AA T The eigenvector of V, the column vector v i is the matrix A T The eigenvector of A; σ i are singular values, is the matrix AA T The characteristic value of .

[0115]

[0116] The specific steps of the adaptive H infinite unscented Kalman filter algorithm proposed in this embodiment are as follows:

[0117] (1) Initialization

[0118]

[0119] In the formula, represents the initial estimate before any measurement is used, E(x0) represents the expected value of x when x=x0, represents the covariance of the estimation error at the initial moment.

[0120] (2) P k Perform SVD decomposition to calculate sampling points and corresponding weight values

[0121] [u S v] = SVD(P k ) (20)

[0122]

[0123]

[0124] In the formula, w m and w c represents the corresponding weight, and λ is the adjustment parameter. α controls the distribution distance of the sampling points. In the present invention, α is 1. β is the prior information integrated into the random variable x. In the present invention, β is 2. L is the dimension of the input variable. is a set of weights used for state mean estimation, is a set of weights for the state covariance estimate, α is the sigma point scaling parameter used to minimize possible high-order effects, and β is a non-negative weight term used to adjust the weight of the 0th sigma point covariance. It is a series of 2L+1 sampling points calculated based on the state estimation at time k-1. In the embodiment of the present invention, the value of L is 4, and 2L+1 represents the number of sigma point sets.

[0125] (3) Time update, calculation of the expected value of the prior state estimation variable and the error covariance matrix requires state variables

[0126]

[0127]

[0128]

[0129] In the formula, f(·) represents the nonlinear system equation, and the sigma point is transformed into Represents the merge vector To obtain the prior state estimate at time k, is the prior estimate of the covariance, Q k is the process noise, as shown in equation (23), the sigma point It is first obtained by time updating and sampling through the state function f.

[0130] (4) Measure updates and calculate output prediction values and the covariance matrix P yy ,P xy .

[0131]

[0132]

[0133]

[0134]

[0135] Here, the sigma point is converted to Merge Vectors get As the calculated output prediction value, P yy is the estimated measurement-prediction covariance, P xy For estimation and The covariance between .

[0136] (5) The above steps are the same as the unscented Kalman filter (UKF) algorithm. Next, the posterior state estimate is corrected.

[0137] K k =P xy,k (P yy,k ) -1 (30)

[0138]

[0139]

[0140]

[0141] Among them, K k is the Kalman gain, is the posterior state covariance estimate, R e,k is the noise matrix.

[0142] According to the characteristics of the lithium battery state equation, by substituting formula (25) into formula (32), we can obtain:

[0143]

[0144] in

[0145]

[0146] Inverting both sides of the equation and simplifying it yields

[0147]

[0148] because is symmetric positive definite, so the above formula must be greater than 0, so we can get

[0149]

[0150] γ -2 can be regarded as the eigenvalue of the left side of the above equation, so

[0151]

[0152] γ 2 Taking the largest eigenvalue, we can get

[0153]

[0154] In this embodiment, γ is adjusted 2 It can improve the robustness to non-Gaussian noise and correct the ill-conditioned matrix encountered when calculating the covariance matrix.

[0155] (6) An adaptive algorithm is introduced to correct the noise covariance matrix Q and the R matrix.

[0156] The formula for introducing the adaptive algorithm is as follows:

[0157]

[0158]

[0159]

[0160]

[0161] ΔU is the terminal voltage error value output by the model at time k, k0, k1, k2, k3 and k4 are the parameters of the battery open circuit voltage, z k is the charge state estimate at time step k, F k is the approximate value of the terminal voltage error covariance at time k of the model, is the covariance weight, is the predicted output voltage at the instantaneous sigma point, and the noise factor Q is updated by obtaining the covariance of the terminal voltage difference kand R k .

[0162] The above formula can effectively solve the problem of uncertain initial values ​​of the state noise covariance matrix and the measurement noise covariance matrix. Further introducing an adaptive algorithm to correct the noise covariance matrix Q and R matrix can better improve the accuracy of the estimation. The adaptive algorithm can also automatically adjust the parameters and covariance matrix of the UKF according to the characteristics of the system dynamics and the measurement data to better adapt to the changes in the actual system and enhance the stability of the filter to cope with the uncertainty of the system model and external disturbances. The improved UKF in this embodiment can automatically adjust the weight of the sigma point to cope with the problem of nonlinear changes of the power lithium battery after a period of use, and achieve good system dynamic matching. The above improvements have greatly improved the working stability and robustness of the power lithium battery.

[0163] It is worth mentioning that the introduction of the adaptive link in this embodiment increases the flexibility and robustness of UKF in the estimation process, thereby better adapting to the complex and changeable system environment of power lithium batteries. This is very helpful for improving the accuracy and reliability of power lithium battery SOC estimation, especially in application scenarios that require high adaptability and stability.

[0164] In order to estimate the SOC of a power lithium battery SOC estimation method proposed in this embodiment and verify the state response of the method under complex working conditions, the inventor uses the Beijing bus operating condition (BBDST) to experimentally verify the accuracy and stability of the model.

[0165] Under the BBDST condition, 2RC-UKF, 3RC-ADHFOU, and 3RC-ADHFOU were compared to verify the accuracy of the 3RC-ADHFOU model in estimating the battery SOC. Figure 2 Comparison of the SOC estimation results of three algorithms; Figure 3 Comparison of the SOC prediction errors of the three algorithms; Figure 4 The predicted value of the terminal voltage is also compared and analyzed using three algorithms; Figure 5 The error analysis of the terminal voltage prediction by three algorithms is shown in Figure 1. The initial SOC of the algorithm is set to 80%.

[0166] from Figure 2 It can be clearly seen that the equivalent circuit model using the ADHFOU algorithm has a higher accuracy in SOC prediction than the UKF algorithm, and the convergence speed is faster than the UKF algorithm, and it also has a higher fit in the later stage of SOC estimation, which is very close to the measured value. Figure 3From the SOC error analysis, it can be seen intuitively that the algorithm of the 3RC-ADHFOU model has a very high convergence speed, and the absolute value of the error is always controlled within 2, and it has higher stability than the other two model algorithms.

[0167] from Figure 4 It can be seen that under the BBDST condition, the estimated value of the terminal voltage of the 3RC-ADHFOU model algorithm has the best fit with the measured value, and the measured terminal voltage fluctuates less; Figure 5 It can be seen that the voltage error using the ADHFOU algorithm is smaller than that of the UKF algorithm, and the accuracy of the third-order model is also significantly improved compared to the second-order model.

[0168] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0170] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0171] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0172] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principle of the present invention without creative labor shall be included in the protection scope of the present invention.

Claims

1. A method for estimating the state of charge of a power lithium battery, characterized in that: The following steps are involved: Establish a third-order RC equivalent circuit model of a power lithium battery and determine the equation group of the model; The equation group of the third-order RC equivalent circuit model is transformed to obtain the transfer function, which is then converted into a differential equation, and the circuit parameters of the third-order RC equivalent circuit model are derived using the forgetting factor recursive least squares method. And in the derivation process, the simulated annealing algorithm is used to optimize the forgetting factor; Based on the obtained circuit parameters, the state of charge of the power lithium battery is estimated using the adaptive H infinite unscented Kalman filter algorithm; The specific process of deriving the circuit parameters of the third-order RC equivalent circuit model using the forgetting factor recursive least squares method includes: Introducing Forgetting Factor in Parameter Identification of Recursive Least Squares Method λ , to change in real time according to the current size. When the current fluctuation is greater than the set value, λ Try to reduce it as much as possible, otherwise increase it to enhance the stability of the identification process. The least squares parameter identification equation is: in, K To correct the weight value, Φ is the pre-measurement vector, is the parameter to be identified before decoding, P(k) is the covariance matrix; The adaptive H infinite unscented Kalman filter algorithm is specifically as follows: in view of the problem that the noise of the unscented Kalman filter algorithm must be white noise, the H infinite filter algorithm is introduced to filter out non-Gaussian noise by minimizing the cost function, and the singular value decomposition method is used to solve the problem that the covariance matrix PKCholesky decomposition produces an ill-conditioned matrix; finally, an adaptive link is added to achieve accurate estimation of the state of charge of the power battery; The specific process of estimating the state of charge of a power lithium battery using the adaptive H infinite unscented Kalman filter algorithm includes: Initialize parameters; Apply singular value decomposition to the covariance matrix to calculate the sampling points and corresponding weight values; Time update, the state variables are needed to calculate the expected value and error covariance matrix of the prior state estimate variables; Measure updates, calculate output predictions and covariance matrices; Correct the posterior state estimate; The noise covariance matrix is ​​corrected using an adaptive algorithm.

2. A method for estimating the state of charge of a power lithium battery as claimed in claim 1, characterized in that: The third-order RC equivalent circuit model is: in, U oc represents the open circuit voltage of the power supply, R 0 represents the ohmic internal resistance of the battery, R 1 represents the electrochemical polarization internal resistance, R 2 represents the electrochemical polarization internal resistance, R 3 represents the concentration difference polarization internal resistance, C 1 represents the electrochemical polarization capacitance, C 2 represents the electrochemical polarization capacitance, C 3 represents the concentration difference polarization capacitance, U t represents the terminal voltage, I is the current, U 1. U 2. U 3 represent capacitors C 1. C 2. C The three loop terminal voltages of 3; The state of charge SOC is determined using the ampere-hour integration method. In the formula, t 0 is the discharge start time, t At this moment, η Coulomb efficiency is measured by calculating the ratio of the discharge capacity to the charge capacity of the battery. Q C It represents the rated capacity of the battery. After discretizing the above, we can get the state equation and output equation of the model.

3. A method for estimating the state of charge of a power lithium battery as claimed in claim 1, characterized in that: In the specific process of deriving the circuit parameters of the third-order RC equivalent circuit model using the forgetting factor recursive least squares method, the difference between the open-circuit voltage and the terminal voltage at each moment and the battery current value are used as input values. The covariance matrix and the weight factor matrix are corrected through continuous iteration, and the parameter matrix is ​​obtained. The parameters in the matrix are the coefficients of each link in the Z domain of the transfer function of the equivalent circuit model. According to the third-order RC equivalent circuit model of the battery, the battery parameters are identified.

4. A method for estimating the state of charge of a power lithium battery as claimed in claim 1, characterized in that: The specific process of optimizing the forgetting factor using the simulated annealing algorithm includes taking the minimum terminal voltage error as the optimization goal, performing iterative optimization with the forgetting factor as the optimization variable within the optional range of the set forgetting factor, and making the best estimate of the parameters.

5. A method for estimating the state of charge of a power lithium battery as claimed in claim 1, characterized in that: The specific process of correcting the posterior state estimate includes obtaining ,in, Take the largest eigenvalue, , by adjusting Improve the robustness to non-Gaussian noise and correct the ill-conditioned matrix encountered when calculating the covariance matrix, where The posterior estimate of the covariance, is the posterior estimate of the covariance, As the cost function in the H-infinity algorithm, is the mutual covariance between the sampling point observation prediction and the state prediction, and R is the noise covariance matrix.

6. A method for estimating the state of charge of a power lithium battery as claimed in claim 1, characterized in that: The specific process of correcting the noise covariance matrix using the adaptive algorithm includes: calculating the terminal voltage error value output at the model k moment, calculating the approximate value of the terminal voltage error covariance of the model k moment, and updating the noise coefficient Q by obtaining the covariance of the terminal voltage difference. k and R k .

7. A power lithium battery charge state estimation system, characterized in that: include: A model building module is configured to establish a third-order RC equivalent circuit model of a power lithium battery and determine a set of equations of the model; a parameter identification module, configured to transform the equation group of the third-order RC equivalent circuit model to obtain a transfer function, transform the transfer function into a differential equation, and derive circuit parameters of the third-order RC equivalent circuit model using a forgetting factor recursive least squares method; A forgetting factor optimization module is configured to optimize the forgetting factor using a simulated annealing algorithm during the derivation process; A state of charge estimation module is configured to estimate the state of charge of the power lithium battery using an adaptive H infinite unscented Kalman filter algorithm based on the obtained circuit parameters; The specific process of deriving the circuit parameters of the third-order RC equivalent circuit model using the forgetting factor recursive least squares method includes: Introducing Forgetting Factor in Parameter Identification of Recursive Least Squares Method λ , to change in real time according to the current size. When the current fluctuation is greater than the set value, λ Try to reduce it as much as possible, otherwise increase it to enhance the stability of the identification process. The least squares parameter identification equation is: in, K To correct the weight value, Φ is the pre-measurement vector, is the parameter to be identified before decoding, P(k) is the covariance matrix; The adaptive H infinite unscented Kalman filter algorithm is specifically as follows: in view of the problem that the noise of the unscented Kalman filter algorithm must be white noise, the H infinite filter algorithm is introduced to filter out non-Gaussian noise by minimizing the cost function, and the singular value decomposition method is used to solve the problem that the covariance matrix PKCholesky decomposition produces an ill-conditioned matrix; finally, an adaptive link is added to achieve accurate estimation of the state of charge of the power battery; The specific process of estimating the state of charge of a power lithium battery using the adaptive H infinite unscented Kalman filter algorithm includes: Initialize parameters; Apply singular value decomposition to the covariance matrix to calculate the sampling points and corresponding weight values; Time update, the state variables are needed to calculate the expected value and error covariance matrix of the prior state estimate variables; Measure updates, calculate output predictions and covariance matrices; Correct the posterior state estimate; The noise covariance matrix is ​​corrected using an adaptive algorithm.

8. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the steps in the method according to any one of claims 1 to 6 are completed.

Images