Lithium battery parameter and state estimation method, system and device based on multi-scale Kalman filtering and medium

Abstract

The invention provides a lithium battery parameter and state estimation method, system and device based on multi-scale Kalman filtering, and a medium. The method comprises the following steps: S1, establishing a battery model; s2, testing the open-circuit voltage of the lithium ion battery to obtain an OCV-SOC fitting curve; s3, discretizing the model state equation of the battery model to obtain a battery model state equation; and S4, acquiring sampling data of the lithium battery, inputting the sampling data into a battery model state equation, estimating the SOC according to a calculation result of the battery model state equation by using an MIUKF algorithm in each time period of the microscopic time scale, obtaining a corresponding OCV according to the OCV-SOC fitting curve and the SOC, performing battery parameter online identification by using an EKF algorithm at each time of the macroscopic time scale, and obtaining a battery parameter online identification result. And updating the battery model state equation by using the identified battery parameter estimation value. According to the method, the SOC estimation error is obviously reduced, the problem of poor SOC estimation precision caused by parameter change in the working process of the lithium battery is solved, and the precision and robustness are improved.

Description

Technical Field

[0001] The present invention relates to the field of battery state of charge estimation, and in particular to a lithium battery parameter and state estimation method, system, device and medium using multi-scale Kalman filtering. Background Art

[0002] SOC (State of Charge) is defined as the battery's state of charge, which indicates the remaining charge. Accurately estimating lithium-ion battery SOC plays a vital role in improving battery reliability, extending battery life, preventing overcharge and discharge, and ensuring the battery's endurance and safe driving.

[0003] Common SOC estimation methods can be divided into four categories: 1) the ampere-hour integration method; 2) the open-circuit voltage method; 3) the data-driven method; and 4) the adaptive filter method. The adaptive filter method is the most widely used in battery SOC estimation. Currently, battery SOC estimation typically combines battery parameter identification and SOC estimation on the same time scale. However, SOC is a rapidly changing quantity, while battery parameters are more slowly changing. Using the same time scale results in high processor computation load and poses the problem of poor SOC estimation accuracy due to parameter variations during lithium battery operation. Summary of the Invention

[0004] The first object of the present invention is to estimate the SOC and battery parameters of a lithium battery respectively at different time scales to improve the SOC estimation accuracy.

[0005] To this end, the above-mentioned purpose of the present invention is achieved through the following technical solutions:

[0006] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0007] A multi-scale Kalman filter method for estimating lithium battery parameters and states comprises the following steps:

[0008] S1. Establish a battery model;

[0009] S2. Perform an open circuit voltage test on the lithium-ion battery to obtain an OCV-SOC fitting curve;

[0010] S3. Discretize the model state equation of the battery model according to the macro time scale and the micro time scale to obtain the battery model state equation, where adjacent moments in the macro time scale are the start and end moments of a period in the micro time scale;

[0011] S4. Obtain the sampling data of the lithium battery and input the battery model state equation. Use the MIUKF algorithm to estimate the SOC based on the calculation results of the battery model state equation at each time period of the microscopic time scale, and obtain the corresponding OCV based on the OCV-SOC fitting curve and SOC. Use the EKF algorithm to perform online battery parameter identification at each moment of the macroscopic time scale, and update the battery model state equation with the identified battery parameter estimation values.

[0012] While adopting the above technical solutions, the present invention may also adopt or combine the following technical solutions:

[0013] As a preferred technical solution of the present invention: in step S1, the battery model is a second-order RC equivalent circuit model.

[0014] As a preferred technical solution of the present invention: in step S2, the open circuit voltage test of the lithium-ion battery is performed by a discharge static method.

[0015] As a preferred technical solution of the present invention: in step S3, the battery model state equation is expressed as:

[0016]

[0017] In the above formula, X k,l 、X k,l+1 They are t k,l Time and t k,l+1 The system state vector, t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l are the time indices of macro time scale and micro time scale respectively, T is the fixed sampling interval between two adjacent measurement points, u k,l and Y k,l Represents t k,l The system input vector and system observation vector at time η is the battery coulomb efficiency, Q N is the rated capacity of the battery, f(SOC k,l ) is the open circuit voltage, is the polarization resistance R1 in the battery model at t k,l The voltage across the terminals is is the polarization resistance R2 in the battery model at t k,l The voltage across the terminals, SOC k,l t k,l The state of charge value at t k,l The voltage across the polarization resistor R1 in the battery model is k,l t k,lThe current value when θ k t k,l The battery parameter values ​​at this time, where: R0 is the battery ohmic internal resistance in the battery model, R1, C1 and R2, C2 constitute two RC networks in the battery model. In order to simplify the function, the F() function and G() function are used to simplify the equation respectively.

[0018] As a preferred technical solution of the present invention: in step S4, before obtaining the sampling data of the lithium battery, an initialization step is also included, specifically including: initializing the state variable State variable error covariance Battery parameter state variables Battery parameter state variable error covariance The expression is as follows:

[0019]

[0020] In the above formula, x 0,0 is the initial value of the state variable, θ is the battery parameter, and E[] represents the calculation of the covariance.

[0021] As a preferred technical solution of the present invention: in step S4, estimating the SOC using the MIUKF algorithm according to the calculation results of the battery model state equation in each time period of the microscopic time scale includes the following steps:

[0022] Calculate the sampling point set and sampling point weight values ​​at time l in the current period;

[0023] Calculate the predicted value of the state variable based on the sampling point set and the sampling point weight value and the predicted value of the system variance The expression is as follows:

[0024]

[0025] In the above formula, t k-1,l The i-th sampling point in the state variable matrix at time , t k-1,l-1 The i-th sampling point in the state variable matrix at time , t k-1,l-1 The battery parameter matrix, u k-1,l-1 t k-1,l-1 The input matrix when is the mean weight of the i-th sampling point, is the variance weight of the i-th sampling point, t k-1,l-1 The state variable prediction value matrix at time , t k,l-1 The state variable prediction value matrix, Qk-1,l-1 t k-1,l-1 The covariance matrix of the process noise at time , n is the length of the system state vector;

[0026] Calculate observation values ​​based on the sampling point set and sampling point weight values and the observed variance predicted value The expression is as follows:

[0027]

[0028] In the above formula, t k-1,l-1 The i-th sampling point in the state variable matrix at time , t k-1,l-1 The battery parameter matrix, u k-1,l-1 t k-1,l-1 The input matrix when is the mean weight of the i-th sampling point, t k-1,l-1 The output variable matrix when is the variance weight of the i-th sampling point, t k-1,l The output variable prediction value matrix, R k-1,l t k-1,l The covariance matrix of the observation noise at time , n is the length of the system state vector;

[0029] Predicting values ​​based on state variables System variance prediction value Observations and the observed variance predicted value Calculate state variable covariance MIUKF Gain K k-1,l , estimated values ​​of state variables and the state variable error covariance estimate The expression is as follows:

[0030]

[0031] In the above formula, n is the length of the system state vector, is the variance weight of the i-th sampling point, t k-1,l The output variable prediction value matrix when , t k-1,l The i-th sampling point in the state variable matrix at time K k-1,j t k-1,l The Kalman coefficient, e k-j+1 is the error update at time k-j+1.

[0032] If l=L, L is the end time of the current period, l is set to zero and the step of using the EKF algorithm to perform online battery parameter identification is executed. If l is less than L, l is incremented by 1 and the step of calculating the sampling point set and sampling point weight value at time l in the current period is executed.

[0033] As a preferred technical solution of the present invention: the sampling point set and the sampling point weight value expression at time l are as follows:

[0034]

[0035] In the above formula, is the estimated value of the state variable obtained in the last calculation, k and l are the time indices of the macro time scale and micro time scale respectively, is the system variance prediction value obtained in the last calculation, t k-1,l The 0th sampling point in the state variable matrix at time , t k-1,l The i-th sampling point in the state variable matrix at time , n is the length of the system state vector, κ is the scaling factor, is the mean weight of the 0th sampling point, is the variance weight of the 0th sampling point, is the mean weight of the i-th sampling point, is the variance weight of the i-th sampling point, and α and β are weight coefficients respectively.

[0036] As a preferred technical solution of the present invention: in step S4, the step of performing online identification of battery parameters using the EKF algorithm at each moment of the macroscopic time scale includes:

[0037] Initialize t k,0 The estimated value of the state variable, the estimated value of the state variable error covariance, the system input vector and the system observation vector at time t k-1,L The estimated value of the state variables, the estimated value of the state variable error covariance, the system input vector and the system observation vector at time t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l are time indices of macro time scale and micro time scale respectively;

[0038] According to t k,0 The estimated values ​​of the state variables, the estimated values ​​of the state variable error covariance, the system input vector and the system observation vector are used to calculate the estimated values ​​of the battery parameters. and the battery parameter state error covariance estimate The expression is as follows:

[0039]

[0040] In the above formula, Y k,0 t k,0 The system observation vector at time , t k,0 The estimated value of the state variable at time u k,0 t k,0 The system input vector at time , is the predicted value of the battery parameter matrix, is the Kalman coefficient, is the Kalman output matrix, is the system covariance matrix, is the covariance matrix of the process noise.

[0041] The second object of the present invention is to provide a lithium battery parameter and state estimation system based on a multi-scale Kalman filter, comprising the following modules:

[0042] - a battery model building module, the battery model building module is used to build a battery model;

[0043] -OCV-SOC fitting curve acquisition module, the OCV-SOC fitting curve acquisition module is used to perform an open circuit voltage test on a lithium-ion battery to obtain an OCV-SOC fitting curve;

[0044] - A battery model state equation acquisition module, which is used to discretize the model state equation of the battery model according to the macroscopic time scale and the microscopic time scale to obtain the battery model state equation, where adjacent moments in the macroscopic time scale are the start and end moments of a period in the microscopic time scale;

[0045] - A battery parameter online identification and state estimation module, which is used to obtain sampling data of the lithium battery and input the battery model state equation. The battery model state equation is obtained through the battery model state equation acquisition module. The SOC is estimated based on the calculation results of the battery model state equation using the MIUKF algorithm at each time period on the microscopic time scale, and the corresponding OCV is obtained based on the OCV-SOC fitting curve and the SOC. The EKF algorithm is used to perform online battery parameter identification at each moment on the macroscopic time scale, and the battery model state equation is updated with the identified battery parameter estimates.

[0046] A third object of the present invention is to provide an electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus, and wherein:

[0047] a memory for storing a computer program;

[0048] A processor is used to execute the computer program stored in the memory to implement the lithium battery parameter and state estimation method steps of the multi-scale Kalman filter as described above.

[0049] The present invention also provides a non-volatile storage medium, characterized in that: the non-volatile storage medium stores an executable program, and when the executable program is executed by a processor, it implements the lithium battery parameter and state estimation method steps of the multi-scale Kalman filter as described above.

[0050] The present invention provides a lithium battery parameter and state estimation method, system, device, and medium using a multi-scale Kalman filter. The method comprises the following steps: S1, establishing a battery model; S2, performing an open-circuit voltage test on the lithium-ion battery to obtain an OCV-SOC fitting curve; S3, discretizing the model state equation of the battery model based on macroscopic and microscopic time scales to obtain the battery model state equation; S4, obtaining sampled data of the lithium battery and inputting it into the battery model state equation; estimating the SOC using the MIUKF algorithm based on the calculation results of the battery model state equation at each time period on the microscopic time scale, and obtaining the corresponding OCV based on the OCV-SOC fitting curve and the SOC; performing online battery parameter identification using the EKF algorithm at each moment on the macroscopic time scale, and updating the battery model state equation using the identified battery parameter estimates. The present invention estimates the SOC of the lithium battery on a microscopic time scale and estimates the battery parameters on a macroscopic time scale, significantly reducing the error in SOC estimation, solving the problem of poor SOC estimation accuracy caused by parameter changes during the lithium battery operation process, and improving accuracy and robustness.

[0051] Specifically, the present invention has the following beneficial effects:

[0052] The present invention uses the MIUKF (Multi-Innovation Unscented Kalman Filter) to estimate battery SOC at a microscopic time scale and the EKF (Extended Kalman Filter) for online battery parameter identification at a macroscopic time scale. The identified battery parameter values ​​are used to update the battery model state equation. The MIUKF uses multiple innovations to correct state variables and improves the estimation accuracy of the UKF (Unscented Kalman Filter) by reusing old information, significantly reducing the error in SOC estimation. Online battery parameter identification at a macroscopic time scale also avoids frequent adjustments to the model state equation, thereby improving robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1This is a simplified flowchart of the steps of the lithium battery parameter and state estimation method using multi-scale Kalman filtering provided by the present invention.

[0054] Figure 2 This is the battery model circuit diagram in Example 1.

[0055] Figure 3 This is the OCV-SOC fitting curve in Example 1.

[0056] Figure 4 This is a block diagram of the algorithm structure in Example 1.

[0057] Figure 5a-5b They are the actual waveforms of battery current and voltage in Example 1 respectively.

[0058] Figures 6a-6e is the online identification result of the battery parameters in Example 1, wherein: Figure 6a This is the online identification result of the resistor R1 in the second-order RC model of the battery; Figure 6b This is the online identification result of the battery's ohmic internal resistance R0; Figure 6c This is the online identification result of battery capacitor C1; Figure 6d This is the online identification result of the resistor R2 in the second-order RC model of the battery; Figure 6e This is the online identification result of battery capacitor C2.

[0059] Figures 7a-7b They are respectively the comparison results between the voltage in the battery parameter online identification result and the actual voltage in Example 1.

[0060] Figure 8 This is a comparison chart of the SOC estimation results of the lithium battery parameter and state estimation method using the multi-scale Kalman filter provided in Example 1 and other methods. DETAILED DESCRIPTION

[0061] The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.

[0062] Example 1

[0063] To address the time-varying internal parameters of lithium batteries and inaccurate SOC estimation, an algorithm for online identification of battery model parameters and joint SOC estimation is proposed. Based on a second-order RC equivalent circuit model, this joint algorithm uses the EKF (Extended Kalman Filter) algorithm to online identify battery model parameters at a macroscopic time scale, and combines it with the MIUKF (Multi-Innovation Unscented Kalman Filter) algorithm at a microscopic time scale to estimate the lithium battery SOC. The multi-innovation unscented Kalman filter uses data information from multiple moments to expand the residual scalar in the algorithm into an innovation matrix, enabling the reuse of historical data. This can improve the adaptability and convergence of the UKF (Unscented Kalman Filter) in nonlinear systems.

[0064] Based on the above concept, this embodiment proposes a lithium battery parameter and state estimation method using a multi-scale Kalman filter, such as Figure 1 As shown, the following steps are included:

[0065] S1. Build a battery model. The circuit diagram of the battery model is as follows: Figure 2 As shown;

[0066] S2. Perform an open circuit voltage test on the lithium-ion battery to obtain an OCV-SOC fitting curve. The fitting curve is as follows: Figure 3 As shown;

[0067] S3. Discretize the model state equation of the battery model according to the macroscopic time scale and the microscopic time scale to obtain the battery model state equation, where adjacent moments in the macroscopic time scale are the start and end moments of a period in the microscopic time scale;

[0068] S4, obtain the sampling data of the lithium battery and input the battery model state equation, use the MIUKF algorithm to estimate the SOC according to the calculation results of the battery model state equation at each time period of the microscopic time scale, and calculate the SOC according to the Figure 3 The obtained OCV-SOC fitting curve is used to obtain the corresponding OCV from the SOC. The EKF algorithm is used to perform online battery parameter identification at each moment in the macroscopic time scale, and the battery model state equation is updated with the identified battery parameter estimates.

[0069] In step S1 of this embodiment, the selection of the battery model and the accuracy of the battery model parameter identification will affect the accuracy of the battery SOC estimation. The widely used equivalent circuits include the Rint model, the Thevenin model, the PNGV model, and the GNL model. Among them, the second-order RC equivalent circuit model better reflects the dynamic characteristics of the battery while having a relatively small amount of calculation compared to other models. Therefore, the battery model of this embodiment is a second-order RC equivalent circuit model, such as Figure 2As shown in the figure, U OC is the open circuit voltage of the battery; U t is the battery terminal voltage; R0 is the battery ohmic internal resistance; I is the current, which is positive for charging and negative for discharging; R1, C1 and R2, C2 constitute the two RC networks in the battery model, which are used to describe the electrochemical polarization and concentration difference polarization inside the battery respectively; U1 and U2 are the voltages across the two polarization resistors R1 and R2 respectively. Based on the model structure, the model state equation can be derived as follows:

[0070]

[0071] U t =U oc (SOC)-U1-U2-IR0 (2)

[0072] In formula (1) and formula (2), Indicates the state of the voltage across the polarized resistor R1, Indicates the voltage state across the polarization resistor R2, SOC0 indicates the initial value of the battery's state of charge, I t represents the battery current, η is the battery coulombic efficiency, Q N is the rated capacity of the battery.

[0073] In step S2 of this embodiment, the open circuit voltage represents the dynamic characteristics of the battery, which changes with the electrode material, number of cycles, ambient temperature, etc. At the same time, the battery open circuit voltage U oc There is a nonlinear functional relationship with the state of charge SOC. The open circuit voltage of the lithium-ion battery is tested by the discharge static method, and the battery open circuit voltage U is fitted by a polynomial. oc Experimental value. The single cell battery selected in this embodiment is a ternary lithium battery. The rated capacity is 3Ah, the charging cut-off voltage is 4.2V, the discharging cut-off voltage is 2.5V, the battery voltage is 3.7V, and the maximum discharge current is 10A. Before conducting the discharge static experiment, first fully charge the lithium battery, let it stand for 4 hours, and record the battery open circuit voltage when SOC=1. Then discharge it with a discharge current of 1.5A for 12 minutes, let it stand for 4 hours, and record the battery open circuit voltage when SOC=0.9. Repeat the discharge process until the open circuit voltage of SOC=0 is recorded. Use an 8th-order polynomial to fit the measured data to obtain the OCV-SOC fitting curve of the battery open circuit voltage and state of charge as shown below. Figure 3 As shown, the polynomial expression is:

[0074] f(SOC)=139.9SOC 8 -601.7SOC 7 +1071.1SOC 6 -1007.8SOC 5 +528.1SOC4 -147.4SOC 3 +18.46SOC 2 +0.1626SOC+3.423 (3)

[0075] For step S3 of this embodiment, in order to facilitate the algorithm design of the battery equivalent model, the battery state equation needs to be discretized. In view of the slow-changing characteristics of battery parameters and the fast-changing characteristics of battery SOC, a multi-scale method is used to construct a discrete time state-space equation. The battery parameters are estimated at a macro scale, and the battery state variables are estimated at a micro scale. Combined with the changes in lithium battery parameters, the state space equation based on lithium battery parameters is obtained:

[0076]

[0077] In formula (4), X k,l+1 It is t k,l The system state vector at time t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l represent the time index of macro time scale and micro time scale. T is the fixed sampling interval between two adjacent measurement points. u k,l and Y k,l Represents t k,l The system input vector and system observation vector at time w k,l 、v k,l is the process noise and observation noise of the system, and their noise covariances are Q k,l and R k,l . b k is the model parameter process noise, and its noise covariance is

[0078] According to formula (4), formula (1) and formula (2) are discretized, and the established battery model state equation expression is:

[0079]

[0080] In formula (5) and formula (6), X k,l+1 It is t k,l The system state vector at time t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l are the time indices of macro time scale and micro time scale respectively, T is the fixed sampling interval between two adjacent measurement points, uk,l and Y k,l Represents t k,l The system input vector and system observation vector at time η is the battery coulomb efficiency, Q N is the rated capacity of the battery, f(SOC k,l ) is the open circuit voltage, is the polarization resistance R1 of the battery model at t k,l The voltage across the two ends of the is the polarization resistance R2 of the battery model at t k,l The voltage across the terminals, SOC k,l t k,l The state of charge value at t k,l The voltage across the polarization resistor R1 in the battery model is k,l t k,l The current value when θ k t k,l The battery parameter values ​​at this time are: R0 is the battery ohmic internal resistance in the battery model, and R1, C1 and R2, C2 constitute two RC networks in the battery model.

[0081] In step S4 of this embodiment, in order to realize the joint online estimation of battery model parameters and battery SOC, as shown in FIG. Figure 4 As shown in the figure, the EKF algorithm is used to perform online identification of the battery parameters R0, R1, R2 and C1, C2, and the identified parameter values ​​are used for MIUKF state estimation. The SOC is estimated according to the MIUKF algorithm, and the OCV at this moment is obtained through the OCV-SOC function relationship of formula (3). At this time, the MIUKF algorithm estimates the SOC at the micro time scale. When the micro time scale and the macro time scale are equal, the EKF performs parameter online identification. After the parameter identification is completed, the MIUKF continues to perform SOC estimation at the micro time scale. The specific steps include the following:

[0082] The first step is to initialize the state variables State variable error covariance Battery parameter state variables Battery parameter state variable error covariance The expression is as follows:

[0083]

[0084] In the above formula, x 0,0 is the initial value of the state variable, θ is the battery parameter, and E[] represents the calculation of the covariance.

[0085] The second step is to obtain the sampling data of the lithium battery, which includes the battery voltage and current signals, and input the sampling data into the battery model. At each time period (l = 1, 2, 3, 4, ..., L) on the micro time scale, the MIUKF algorithm is used to estimate the SOC based on the calculation results of the battery model state equation. The steps include:

[0086] 1. Calculate the sampling point set and sampling point weight value at time l in the current period. The sampling point set and sampling point weight value expressions at time l are as follows:

[0087]

[0088] In the above formula, is the estimated value of the state variable obtained in the last calculation, k and l are the time indices of the macro time scale and micro time scale respectively, is the system variance prediction value obtained in the last calculation, t k-1,l The 0th sampling point in the state variable matrix at time , t k-1,l The i-th sampling point in the state variable matrix at time , κ is the scaling factor, is the mean weight of the 0th sampling point, is the variance weight of the 0th sampling point, is the mean weight of the i-th sampling point, is the variance weight of the i-th sampling point, α and β are weight coefficients respectively, n is the length of the system state vector, in this embodiment n is 3, take α = 1e -4 , β=2.

[0089] 2. Calculate the predicted value of the state variable based on the sampling point set and the sampling point weight value and the predicted value of the system variance The expression is as follows:

[0090]

[0091] In the above formula, t k-1,l The i-th sampling point in the state variable matrix at time , t k-1,l-1 The i-th sampling point in the state variable matrix at time , t k-1,l-1 The battery parameter matrix, u k-1,l-1 t k-1,l-1 The input matrix at t k-1,l-1 The input current i k-1,l-1 , is the mean weight of the i-th sampling point, is the variance weight of the i-th sampling point, t k-1,l-1 The state variable prediction value matrix at time t k-1,l The state variable prediction value matrix at the previous moment on the micro time scale, t k,l-1 The state variable prediction value matrix at time t k-1,l-1 The predicted value matrix of the state variables at the next moment under the macro time scale, Q k-1,l-1 t k-1,l-1 The covariance matrix of the process noise at time , n is the length of the system state vector;

[0092] 3. Calculate the observation value based on the sampling point set and the sampling point weight value and the observed variance predicted value The expression is as follows:

[0093]

[0094] In the above formula, t k-1,l-1 The i-th sampling point in the state variable matrix at time , t k-1,l-1 The battery parameter matrix, u k-1,l-1 t k-1,l-1 The input matrix at t k-1,l-1 The input current i k-1,l-1 , is the mean weight of the i-th sampling point, t k-1,l-1 The output variable matrix when is the variance weight of the i-th sampling point, t k-1,l The output variable prediction value matrix, R k-1,l t k-1,l The covariance matrix of the observation noise at time , n is the length of the system state vector;

[0095] 4. Predicting values ​​based on state variables System variance prediction value Observations and the observed variance predicted value Calculate state variable covariance MIUKF Gain K k-1,l , estimated values ​​of state variables and the state variable error covariance estimate The expression is as follows:

[0096]

[0097] In the above formula, n is the length of the system state vector, is the variance weight of the i-th sampling point, t k-1,l The output variable prediction value matrix when , t k-1,l The i-th sampling point in the state variable matrix at time K k-1,j t k-1,l The Kalman coefficient, e k-j+1 is the error innovation at time k-j+1; for nonlinear battery systems, the traditional UKF uses a single innovation to update the state variables of the system, while the MIUKF algorithm uses multiple innovations to correct the state variables, improving the estimation accuracy of the UKF by reusing old information. k =Y k -G(X k ,θ k ,u k ), then e k is the error innovation at time k, representing the error of the observation value. According to the UKF algorithm process, only the error data at the current moment is used in each SOC estimation process. If the error of the observation value is large or the values ​​of the process covariance and the observation covariance do not match the noise model of the system, it is easy to cause slow algorithm convergence and poor estimation accuracy. To fully utilize the information of historical data, the single innovation at the current moment is expanded into a multi-innovation vector containing the innovations of the current and previous moments to increase the amount of error innovation;

[0098] According to the previous text, the system state vector That is, the estimated value of the state variable Including the state of charge SOC, so the state variable estimate is obtained Afterwards, we can estimate the value from the state variable The estimated value of the state of charge SOC is obtained, and the estimated value of the state of charge SOC is further substituted into formula (3) to obtain the estimated result of the battery open circuit voltage OCV. Therefore, each state variable estimated value is obtained. The corresponding estimated value of the state of charge SOC can be obtained.

[0099] 5. If l = L, where L is the end time of the current period, set l to zero and execute the step of online identification of battery parameters using the EKF algorithm. If l is less than L, add 1 to l and execute the step of calculating the sampling point set and sampling point weight value at time l in the current period.

[0100] The third step is to use the EKF algorithm to perform online identification of battery parameters at each moment of the macroscopic time scale. The steps include:

[0101] Initialize t k,0 The estimated value of the state variable when State variable error covariance estimate System input vector u k,0 and the system observation vector Y k,0 t k-1,L The estimated value of the state variable when State variable error covariance estimate System input vector u k-1,L and the system observation vector Y k-1,L ,Right now:

[0102]

[0103] As mentioned above, t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l are time indices of macro time scale and micro time scale respectively;

[0104] According to t k,0 The estimated value of the state variable when State variable error covariance estimate System input vector u k,0 and the system observation vector Y k,0 Calculate battery parameter estimates and the battery parameter state error covariance estimate The expression is as follows:

[0105]

[0106]

[0107] In the above formula, Y k,0 t k,0 The system observation vector at time , t k,0 The estimated value of the state variable at time u k,0 t k,0 The system input vector at time , is the predicted value of the battery parameter matrix, is the Kalman coefficient, is the Kalman output matrix, is the system covariance matrix, is the covariance matrix of the process noise.

[0108] The fourth step is to calculate the estimated battery parameters Substitute θ into equations (5) and (6) k , thereby updating the battery model state equation, and the updated battery model state equation is used to estimate the SOC of the corresponding period of the microscopic time scale starting from the current moment of the macroscopic time scale, and tk,0 The estimated value of the state variable when State variable error covariance estimate System input vector u k,0 and the system observation vector Y k,0 As the initial value of SOC estimation at time 0 on the microscopic time scale.

[0109] The following experimental verification and analysis results illustrate the effects of the lithium battery parameter and state estimation method using the multi-scale Kalman filter of this embodiment.

[0110] The battery test can charge or discharge the battery according to the set working conditions. In order to verify the effectiveness of the proposed algorithm under complex dynamic conditions, the dynamic working condition test is used with a total time of 20,000 seconds. The battery current and the measurement terminal voltage are as follows: Figure 5a-5b As shown, import Figure 5a-5b The dynamic working condition data of the battery is obtained by setting the macro time scale to 60 and the micro time scale to 1 for experimental simulation. The battery is identified online. The parameter identification results are shown in the figure. Figure 6a-6e As shown, Figure 6a This is the online identification result of the resistor R1 in the second-order RC model of the battery; Figure 6b This is the online identification result of the battery's ohmic internal resistance R0; Figure 6c This is the online identification result of battery capacitor C1; Figure 6d This is the online identification result of the resistor R2 in the second-order RC model of the battery; Figure 6e The online identification result of the battery capacitor C2 is shown in Figure 2. In order to verify the accuracy of the algorithm's online identification parameters, the actual voltage under dynamic conditions is compared with the voltage in the battery parameter value estimated by the method of this embodiment. The results are shown in Figure 2. Figure 7a-7b The figure shows a dynamic comparison between the voltage and the actual voltage in the online battery parameter identification results. After removing the first large error caused by an incorrect initial SOC value, the maximum absolute error is 0.02V. The average absolute error is 0.0050V, and the relative average absolute error is 0.05%. The actual terminal voltage is very consistent with the estimated voltage, demonstrating the effectiveness of the method of this embodiment for online battery parameter identification.

[0111] For the SOC estimation result, the method of this embodiment is compared with the UKF-EKF, DEKF and other algorithms. After inputting the current and voltage data under dynamic working conditions, this paper uses the ampere-hour integration method as the true value of SOC. The error of the SOC estimation result is as follows: Figure 8 As shown, Figure 8The accuracy of the SOC estimation method proposed in this embodiment was further compared with traditional algorithms such as UKF-EKF and DEKF. It can be seen that the SOC estimation error of the method of this embodiment is mostly within -0.8%, which is smaller than the other two methods. In addition, the computational efficiency of the three methods was compared. In order to minimize the impact of randomness, the three methods were performed 5 times, and the average values ​​were then taken for comparison. The average computation time of the method of this embodiment, UKF-EKF, and DEKF were 0.472s, 0.522s, and 0.552s, respectively. It can be seen that the method of this embodiment consumes less computation time. Therefore, the method of this embodiment can not only improve the accuracy and performance of SOC estimation, but also reduce the computation time.

[0112] Example 2

[0113] The present invention also provides a lithium battery parameter and state estimation system based on a multi-scale Kalman filter, comprising the following modules:

[0114] - a battery model building module, the battery model building module is used to build a battery model;

[0115] -OCV-SOC fitting curve acquisition module, the OCV-SOC fitting curve acquisition module is used to perform an open circuit voltage test on a lithium-ion battery to obtain an OCV-SOC fitting curve;

[0116] - A battery model state equation acquisition module, which is used to discretize the model state equation of the battery model according to the macroscopic time scale and the microscopic time scale to obtain the battery model state equation, where adjacent moments in the macroscopic time scale are the start and end moments of a period in the microscopic time scale;

[0117] - A battery parameter online identification and state estimation module, which is used to obtain sampling data of the lithium battery and input the battery model state equation. The battery model state equation is obtained through the battery model state equation acquisition module. The SOC is estimated based on the calculation results of the battery model state equation using the MIUKF algorithm at each time period on the microscopic time scale, and the corresponding OCV is obtained based on the OCV-SOC fitting curve and the SOC. The EKF algorithm is used to perform online battery parameter identification at each moment on the macroscopic time scale, and the battery model state equation is updated with the identified battery parameter estimates.

[0118] Example 3

[0119] The present invention also provides an electronic device, comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus.

[0120] a memory for storing a computer program;

[0121] A processor is used to execute the computer program stored in the memory to implement the lithium battery parameter and state estimation method steps of the multi-scale Kalman filter as described above.

[0122] Example 4

[0123] The present invention also provides a non-volatile storage medium, in which an executable program is stored. When the executable program is executed by a processor, the steps of the lithium battery parameter and state estimation method of the multi-scale Kalman filter as described above are implemented.

[0124] The above-mentioned specific implementation methods are used to illustrate the present invention and are only preferred embodiments of the present invention, rather than limiting the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit of the present invention and the scope of protection of the claims shall fall within the scope of protection of the present invention.

Claims

1. A lithium battery parameter and state estimation method based on multi-scale Kalman filtering, characterized in that: The steps include: S1. Establish a battery model; S2. Perform an open circuit voltage test on the lithium-ion battery to obtain an OCV-SOC fitting curve; S3. Discretize the model state equation of the battery model according to the macro time scale and the micro time scale to obtain a discretized battery model state equation, where adjacent moments in the macro time scale are the start and end moments of a period in the micro time scale; S4. Based on the lithium battery sampling data obtained in S3 and inputting the discretized battery model state equation, the MIUKF algorithm is used to estimate the SOC according to the calculation results of the battery model state equation at each time period of the microscopic time scale, and the corresponding OCV is obtained according to the OCV-SOC fitting curve and the SOC. The EKF algorithm is used to perform online identification of battery parameters at each moment of the macroscopic time scale, and the battery model state equation is updated with the estimated values ​​of the identified battery parameters.

2. The method according to claim 1, characterized in that In step S2, the open circuit voltage test of the lithium-ion battery is performed by a discharge static method.

3. The method according to claim 1, characterized in that In step S3, the battery model state equation is expressed as: In the above formula, u k,l =i k,l , X k,l 、X k,l+1 They are t k,l Time and t k,l+1 The system state vector, t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l are the time indices of macro time scale and micro time scale respectively, T is the fixed sampling interval between two adjacent measurement points, u k,l and Y k,l Represents t k,l The system input vector and system observation vector at time η is the battery coulomb efficiency, Q N is the rated capacity of the battery, f(SOC k,l ) is the open circuit voltage, is the polarization resistance R1 in the battery model at t k,l The voltage across the terminals is is the polarization resistance R2 in the battery model at t k,l The voltage across the terminals, SOC k,l t k,l The state of charge value at t k,l The voltage across the polarization resistor R1 in the battery model is k,l t k,l The current value when θ k t k,l The battery parameter values ​​at this time, where: R0 is the battery ohmic internal resistance in the battery model, R1, C1 and R2, C2 constitute two RC networks in the battery model; in order to simplify the function, the F() function and G() function are used to simplify the equation respectively.

4. The method according to claim 1, wherein In step S4, before obtaining the sampling data of the lithium battery, an initialization step is also included, specifically including: initializing the state variable State variable error covariance Battery parameter state variables Battery parameter state variable error covariance The expression is as follows: In the above formula, x 0,0 is the initial value of the state variable, θ is the battery parameter, and E[] represents the calculation of the covariance.

5. The method according to claim 1, wherein In step S4, estimating the SOC using the MIUKF algorithm based on the calculation results of the battery model state equation at each time period at the microscopic time scale includes the following steps: Calculate the sampling point set and sampling point weight values ​​at time l in the current period; Calculate the predicted value of the state variable based on the sampling point set and the sampling point weight value and the predicted value of the system variance The expression is as follows: In the above formula, t k-1,l The i-th sampling point in the state variable matrix at time , t k-1,l-1 The i-th sampling point in the state variable matrix at time , t k-1,l-1 The battery parameter matrix, u k-1,l-1 t k-1,l-1 The input matrix when is the mean weight of the i-th sampling point, is the variance weight of the i-th sampling point, t k-1,l-1 The state variable prediction value matrix at time , t k,l-1 The state variable prediction value matrix, Q k-1,l-1 t k-1,l-1 The covariance matrix of the process noise at time , n is the length of the system state vector; Calculate observation values ​​based on the sampling point set and sampling point weight values and the observed variance predicted value The expression is as follows: In the above formula, t k-1,l-1 The i-th sampling point in the state variable matrix at time , t k-1,l-1 The battery parameter matrix, u k-1,l-1 t k-1,l-1 The input matrix when is the mean weight of the i-th sampling point, t k-1,l-1 The output variable matrix when is the variance weight of the i-th sampling point, t k-1,l The output variable prediction value matrix, R k-1,l t k-1,l The covariance matrix of the observation noise at time , n is the length of the system state vector; Predicting values ​​based on state variables System variance prediction value Observations and the observed variance predicted value Calculate state variable covariance MIUKF Gain K k-1,l , estimated values ​​of state variables and the state variable error covariance estimate The expression is as follows: In the above formula, n is the length of the system state vector, is the variance weight of the i-th sampling point, t k-1,l The output variable prediction value matrix when , t k-1,l The i-th sampling point in the state variable matrix at time K k-1,j t k-1,l The Kalman coefficient, e k-j+1 is the error update at time k-j+1; If l=L, L is the end time of the current period, l is set to zero and the step of using the EKF algorithm to perform online battery parameter identification is executed. If l is less than L, l is incremented by 1 and the step of calculating the sampling point set and sampling point weight value at time l in the current period is executed.

6. The method according to claim 5, characterized in that The sampling point set and sampling point weight value expression at time l are as follows: In the above formula, is the estimated value of the state variable obtained in the last calculation, k and l are the time indices of the macro time scale and micro time scale respectively, is the system variance prediction value obtained in the last calculation, t k-1,l The 0th sampling point in the state variable matrix at time , t k-1,l The i-th sampling point in the state variable matrix at time , n is the length of the system state vector, κ is the scaling factor, is the mean weight of the 0th sampling point, is the variance weight of the 0th sampling point, is the mean weight of the i-th sampling point, is the variance weight of the i-th sampling point, and α and β are weight coefficients respectively.

7. The method according to claim 5, characterized in that In step S4, the steps of performing online identification of battery parameters using the EKF algorithm at each moment of the macroscopic time scale include: Initialize t k,0 The estimated value of the state variable, the estimated value of the state variable error covariance, the system input vector and the system observation vector at time t k-1,L The estimated value of the state variables, the estimated value of the state variable error covariance, the system input vector and the system observation vector at time t k,l =t k,0 +l×T,t k,0 =t k-1,L (l=1,2,3,4,...,L), k and l are time indices of macro time scale and micro time scale respectively; According to t k,0 The estimated values ​​of the state variables, the estimated values ​​of the state variable error covariance, the system input vector and the system observation vector are used to calculate the estimated values ​​of the battery parameters. and the battery parameter state error covariance estimate The expression is as follows: In the above formula, Y k,0 t k,0 The system observation vector at time , t k,0 The estimated value of the state variable at time u k,0 t k,0 The system input vector at time , is the predicted value of the battery parameter matrix, is the Kalman coefficient, is the Kalman output matrix, is the system covariance matrix, is the covariance matrix of the process noise.

8. A multi-scale Kalman filter lithium battery parameter and state estimation system, characterized in that: Includes the following modules: - a battery model building module, the battery model building module is used to build a battery model; -OCV-SOC fitting curve acquisition module, the OCV-SOC fitting curve acquisition module is used to perform an open circuit voltage test on a lithium-ion battery to obtain an OCV-SOC fitting curve; - A battery model state equation acquisition module, which is used to discretize the model state equation of the battery model according to the macroscopic time scale and the microscopic time scale to obtain the battery model state equation, where adjacent moments in the macroscopic time scale are the start and end moments of a period in the microscopic time scale; - A battery parameter online identification and state estimation module, which is used to obtain sampling data of the lithium battery and input the battery model state equation. The battery model state equation is obtained through the battery model state equation acquisition module. The SOC is estimated based on the calculation results of the battery model state equation using the MIUKF algorithm at each time period on the microscopic time scale, and the corresponding OCV is obtained based on the OCV-SOC fitting curve and the SOC. The EKF algorithm is used to perform online battery parameter identification at each moment on the macroscopic time scale, and the battery model state equation is updated with the identified battery parameter estimates.

9. An electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus, wherein: a memory for storing a computer program; A processor, wherein the processor is used to execute a computer program stored in a memory to implement the steps of the lithium battery parameter and state estimation method using multi-scale Kalman filtering as described in any one of claims 1 to 6.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores an executable program, and when the executable program is executed by the processor, it implements the steps of the lithium battery parameter and state estimation method of multi-scale Kalman filtering according to any one of claims 1 to 7.

Images