Lithium battery SOC and SOE joint estimation method for solving particle depletion

Abstract

The invention relates to a lithium battery SOC and SOE joint estimation method for solving particle depletion, and provides a novel genetic marginal particle filtering method for accurately and jointly estimating the SOC and SOE of a lithium ion battery under multiple temperature scales. The problem of particle depletion in a traditional particle filtering algorithm is solved; then a Rao-Blackwell theory in statistics is adopted, partial linear state variables are marginalized in the calculation process of particle filtering, posterior distribution is approximated to single Gaussian distribution, a Kalman filter is used for updating under the condition that other nonlinear variables are known, a genetic marginalized particle filtering algorithm is obtained, and prediction accuracy is improved; for a ternary lithium ion battery, a second-order Thevenin equivalent circuit model is established, and the model can accurately simulate the polarization effect of the battery and accurately characterize the battery. The method has the effects of timely offline parameter identification data processing and real-time characterization.

Description

Technical Field

[0001] The present application relates to the field of new energy measurement and control, and in particular to a method for jointly estimating SOC and SOE of a lithium battery for solving particle depletion. Background Art

[0002] Lithium-ion batteries have become the main component of energy storage system applications due to their advantages such as long life, high stability, high energy density and moderate price. With the continuous expansion of the battery market, more and more energy storage fields have begun to use ternary lithium-ion batteries with higher energy density for better product performance and user experience. The lithium-ion battery management system (Battery Management System, BMS) as the main object of battery energy management and safety monitoring has become an indispensable core in the use of lithium-ion batteries as energy and power.

[0003] A BMS with excellent performance can not only accurately measure key reference information such as battery voltage, current and temperature, perform system safety diagnosis and hazard alarm at the same time, but also predict the state of charge (SOC) and energy state (SOE), and manage the battery energy consumption based on actual conditions to improve the battery's energy utilization efficiency. Since state parameters such as SOC and SOE are hidden quantities that cannot be measured directly, and the estimation results are easily affected by noise, resulting in large estimation errors, accurately and efficiently estimating battery state parameters has become a major research hotspot in today's battery management systems.

[0004] SOC is a direct display of the remaining power of a power lithium-ion battery. Generally speaking, it is defined as the ratio of the power that can be discharged from the current state to the voltage to the rated power. Accurate and reliable SOC is the basis for ensuring normal use by users. It can not only optimize the charging and discharging strategy of lithium-ion batteries, but also be used to calculate other state parameters of power batteries. SOC can provide an effective reference for the state of the battery, provide data and correction solutions for the energy management of power lithium-ion batteries, and is the cornerstone of BMS for predictive analysis. Reliable SOC value estimation can help the BMS system better manage energy and avoid the occurrence of thermal runaway of lithium-ion batteries and a sudden reduction in cycle life. Accurately estimating SO The C value is the key to ensuring the working performance of the battery pack, improving the cycle life and reducing the cost of use. As a typical secondary battery, the essence of the charging and discharging of lithium-ion batteries is the intercalation and deintercalation movement of lithium ions between the positive and negative plates. At the same time, a series of side reactions will occur between the positive and negative plates and the electrolyte. Therefore, there are multiple mutually coupled electrochemical reactions inside the battery. However, in actual applications, due to the complexity and diversity of the battery working environment, as well as aging problems caused by frequent use, overcharging and over-discharging and other factors, the internal reaction process of lithium-ion batteries will be affected, thereby greatly increasing the difficulty of estimating the battery SOC. Therefore, accurate and reliable SOC estimation strategies have become one of the problems that BMS urgently needs to solve.

[0005] With respect to the above-mentioned related technologies, the inventors believe that there are the following technical defects: unlike SOC, SOE represents the change of battery energy state, and it takes into account the change of battery voltage on the basis of SOC; although SOE estimation is difficult, it has greater advantages than SOC in terms of the remaining discharge time and remaining energy of subsequent power lithium-ion batteries; since power lithium-ion batteries have broad application prospects in various industries in the future, BMS systems need to have higher accuracy and stability, so it is very meaningful to enhance the research on SOE estimation algorithms on the basis of SOC estimation algorithms; at the same time, in the current use of power lithium-ion batteries, SOE is an important basis for estimating the remaining use time or mileage; in addition, if the SOE estimation result is greater than the actual result, it will cause the power lithium-ion battery to be over-discharged, and the over-discharge of the power lithium-ion battery will cause irreversible damage to the battery itself, which will not only reduce the service life of the battery, but also bring great safety problems; if the estimated value of SOE is less than the actual value, it will cause the problem of reduced use efficiency, and the use efficiency of the power lithium-ion battery cannot be effectively exerted. Summary of the invention

[0006] In order to improve the problems of difficulty in jointly estimating SOC and SOE of existing lithium-ion batteries and low estimation accuracy under various ambient temperatures, the present application provides a method for jointly estimating SOC and SOE of lithium batteries that solves particle depletion.

[0007] The present application provides a method for jointly estimating SOC and SOE of a lithium battery to solve particle depletion, which adopts the following technical solution:

[0008] A method for jointly estimating SOC and SOE of lithium batteries to solve particle depletion is disclosed. A second-order Thevenin equivalent circuit model is established for ternary lithium-ion batteries. The parameters of the second-order Thevenin equivalent model are identified online by using a forgetting factor recursive least squares algorithm (FFRLS). The resampling process in the traditional particle filter algorithm is optimized and improved by a genetic algorithm. The Rao-Blackwell theory in statistics is used to marginalize some linear state variables in the particle filter calculation process, and the posterior distribution is approximated as a single Gaussian distribution. The Kalman filter is used to update the remaining nonlinear state variables under the condition that they are known, so as to realize the establishment of a lithium-ion battery pack estimation model and the reliable operation of the mathematical iterative operation algorithm of the SOC value and the SOE value. The proposed genetic marginalized particle filter algorithm is verified under HPPC and BBDST conditions at temperatures of 15°C, 25°C and 35°C, respectively, to solve the problems of difficult joint estimation of SOC and SOE, low precision and poor adaptability to ambient temperature of the traditional particle filter algorithm.

[0009] By adopting the above technical scheme, a second-order Thevenin equivalent circuit model is established, and the forgetting factor recursive least squares algorithm (FFRLS) is used to realize the online identification of model parameters. The resampling process of the traditional particle filter algorithm is optimized by combining the genetic algorithm, and some linear state variables in the particle filter are marginalized by the Rao-Blackwell theory. The posterior distribution is approximated as a single Gaussian distribution, and the update is completed through the Kalman filter under the condition that the nonlinear state variables are known. The problems of insufficient accuracy, poor adaptability to ambient temperature and difficulty in application under complex working conditions in the joint estimation of lithium battery SOC (state of charge) and SOE (state of energy) of the traditional particle filter algorithm are solved. The high-precision joint estimation of SOC and SOE in HPPC (hybrid pulse power characteristic) and BBDST (power battery test) conditions under multiple temperature conditions (15℃, 25℃, 35℃) and the stable and reliable operation of the model are realized. The algorithm has strong environmental adaptability and algorithm optimization capabilities, and meets the efficient and accurate requirements of lithium battery pack state evaluation under complex working conditions.

[0010] Optionally, a second-order Thevenin equivalent circuit model is established for ternary lithium-ion batteries. This model can accurately simulate the polarization effect of the battery, accurately characterize the battery, and the related calculations are relatively simple and easy to implement in engineering.

[0011] By adopting the above technical solution and establishing a second-order Thevenin equivalent circuit model, the polarization effect of the ternary lithium-ion battery can be accurately simulated to achieve accurate characterization of the battery characteristics. At the same time, the advantages of the model's simple calculation and easy parameter identification are utilized to support efficient estimation of the battery state. It has high precision, dynamic tracking of the battery state and convenience of engineering implementation, meeting application requirements under multiple working conditions and temperature environments.

[0012] Optionally, for the established second-order Thevenin equivalent circuit model, a forgetting factor recursive least squares algorithm (FFRLS) is used to perform online identification of the parameters of the second-order Thevenin equivalent model to solve the problem of untimely data processing and inability to represent in real time in offline parameter identification.

[0013] By adopting the above technical solution, for the second-order Thevenin equivalent circuit model, the forgetting factor recursive least squares algorithm (FFRLS) is used for online parameter identification, which realizes the real-time update of the battery model parameters, solves the problem of offline parameter identification processing lag and inability to characterize the dynamic characteristics of the battery in real time, and has the technical function of quickly responding to changes in battery status and improving estimation accuracy, meeting the battery management needs under dynamic conditions.

[0014] Optionally, a genetic algorithm is used to optimize and improve the resampling process in the traditional particle filter algorithm, solving the particle depletion problem in the traditional particle filter algorithm. The Rao-Blackwell theory in statistics is used to marginalize some linear state variables in the particle filter calculation process, approximate the posterior distribution to a single Gaussian distribution, and use the Kalman filter to update the remaining nonlinear state variables when they are known, thereby improving the prediction accuracy.

[0015] By adopting the above technical solution, the traditional particle filter resampling process is optimized through genetic algorithm to solve the particle depletion problem, and the linear state variables are marginalized in combination with Rao-Blackwell theory. The Kalman filter is used to update the nonlinear state variables, which improves the prediction accuracy and computational efficiency of the filtering algorithm. It has the technical functions of high-precision state estimation, adaptation to complex working conditions and optimization of particle distribution.

[0016] Optionally, verification experiments are carried out on the proposed genetic marginalized particle filter algorithm under HPPC and BBDST conditions at temperatures of 15°C, 25°C and 35°C, respectively, to solve the problems of difficult estimation, low accuracy and poor adaptability under different ambient temperatures in the traditional particle filter algorithm for the joint estimation of SOC and SOE of lithium-ion batteries.

[0017] By adopting the above technical solution, for the genetic marginalized particle filter algorithm, through HPPC and BBDST working condition verification experiments at 15℃, 25℃ and 35℃, the problems of low accuracy and poor adaptability of traditional particle filtering in the joint estimation of SOC and SOE of lithium-ion batteries are solved. It has the technical functions of high-precision estimation across temperature environments, strong environmental adaptability and applicability to complex working conditions.

[0018] In summary, the present application includes at least one of the following beneficial technical effects:

[0019] 1. By establishing a second-order Thevenin equivalent circuit model, the polarization effect of ternary lithium-ion batteries is accurately simulated to achieve accurate characterization of battery characteristics, simplify calculations, and support efficient engineering implementation. The forgetting factor recursive least squares algorithm (FFRLS) is used for online identification of model parameters, real-time update of battery model parameters, solving the problem of offline identification lag, and improving the response speed and estimation accuracy of battery dynamic state;

[0020] 2. Use genetic algorithm to optimize the resampling process of traditional particle filter algorithm, combine Rao-Blackwell theory to marginalize linear state variables, and use Kalman filter to update nonlinear state variables to improve particle distribution optimization and prediction accuracy of filter algorithm;

[0021] 3. The genetic marginalized particle filter algorithm was verified under multiple temperature conditions (15°C, 25°C, 35°C) and multiple working conditions (HPPC and BBDST) to solve the problems of low accuracy and poor adaptability in the joint estimation of SOC and SOE, and achieve high-precision estimation across environments;

[0022] 4. Improve the robustness, dynamic adaptability and environmental adaptability of lithium battery SOC and SOE estimation models to meet the requirements of efficient and accurate applications under complex working conditions and different temperature conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of the method of this application.

[0024] Figure 2 It is a schematic diagram of the second-order Thevenin equivalent circuit model in the example of this application. DETAILED DESCRIPTION

[0025] The following is combined with Figure 1-2 This application is described in further detail.

[0026] The embodiment of the present application discloses a method for jointly estimating SOC and SOE of a lithium battery to solve particle depletion.

[0027] Reference Figure 1 , Figure 2, Uoc represents the open circuit voltage, U L is the terminal voltage of the lithium-ion battery, R0 is the battery ohmic internal resistance, R1 and C1 are the electrochemical polarization internal resistance and polarization capacitance, R2 and C2 are the concentration polarization resistance and polarization capacitance. I is the load current; U1 and U2 are the voltages of the parallel networks of R1C1 and R2C2, respectively.

[0028] The embodiment of the present application is based on the experimental analysis of the power application requirements and working characteristics of lithium-ion batteries, combined with the research ideas of modern control theory, and is a lithium-ion battery SOC and SOE joint estimation method based on the genetic marginalized particle filter algorithm, which is suitable for a variety of ambient temperatures; refer to Figure 1 , Figure 2 , aiming at the goal of accurate joint estimation of SOC and SOE of lithium-ion batteries under multi-temperature environments, this application establishes a second-order Thevenin equivalent circuit model to accurately characterize the battery polarization effect and internal reaction, and uses the forgetting factor recursive least squares algorithm to identify the model parameters online, solving the problem of untimely data processing and inability to characterize in real time in offline parameter identification; in view of the difficulty of joint estimation and low estimation accuracy caused by particle depletion in traditional particle filtering, the genetic marginalization particle filtering algorithm optimizes the resampling process in the traditional particle filtering algorithm through a genetic algorithm, solving the problem of particle depletion, and at the same time, through the Rao-Blackwell theory, marginalizes some linear state variables in the calculation process of particle filtering, approximates the posterior distribution to a single Gaussian distribution, and uses the Kalman filter to update when the remaining nonlinear state variables are known, thereby improving the prediction accuracy; the present invention has been verified under working conditions at different ambient temperatures, and can provide a method reference for joint estimation of SOC and SOE of lithium-ion batteries in different application scenarios, with the advantages of simple calculation, good adaptability and high accuracy;

[0029] In order to better reflect the present application, in this embodiment, only a ternary lithium-ion battery is used as an example for explanation, but those skilled in the art should be aware that according to the technical idea of ​​the present invention, a variety of lithium-ion batteries can be estimated based on genetic marginalized particle filtering at multiple temperature scales. The following is a detailed description of the steps for implementing a lithium-ion battery SOC and SOE joint estimation method based on genetic marginalized particle filtering at multiple temperature scales for lithium-ion batteries:

[0030] Step S1: First, an equivalent circuit model is established for the ternary lithium-ion battery, namely, a second-order Thevenin equivalent circuit model; according to Kirchhoff's circuit law, the model relationship is converted in combination with the second-order Thevenin equivalent circuit model; Figure 2 In the equation, Uoc represents the open circuit voltage, U Lis the terminal voltage of the lithium-ion battery, R0 is the battery ohmic internal resistance, R1 and C1 are the electrochemical polarization internal resistance and polarization capacitance, R2 and C2 are the concentration polarization resistance and polarization capacitance; I is the load current; U1 and U2 are the voltages of the parallel network of R1C1 and R2C2 respectively; the commonly used expression of SOC is as follows:

[0031]

[0032] In formula (2): SOC t is the SOC value at time t; C n is the rated capacity of the battery; η1 is the charge and discharge efficiency; for the selected second-order equivalent model, [SOC U1 U2] is selected as the state variable. Combining formula (1) and the definition of SOC, after discretization, its discrete state space equation can be listed as shown in formula (2).

[0033]

[0034] The process noise and measurement noise in the SOC estimation process are respectively expressed as ω in the above formula. i (i=1,2,3) and ν are used to represent the uncertainty of the model. The parameters that need to be identified in the model are the open circuit voltage Uoc, the ohmic internal resistance R0, the polarization internal resistance R1, R2 and the polarization capacitance C1, C2. The power integration method is similar to the ampere-hour integration method of SOC estimation. It is based on the definition of integrating the power of the battery. The energy change is obtained by integrating the product of the battery current and voltage to obtain the current SOE value of the battery. The calculation expression is:

[0035]

[0036] Among them, E N is the maximum available energy of the battery, η k is the battery charge and discharge efficiency, P(τ) is the battery power, and t is the current working time. According to the above formula, discretization can be obtained:

[0037]

[0038] Where, ΔE k is the energy consumed at time k, I k is the load current of the battery at time k, U L,k is the terminal voltage of the lithium-ion battery at time k, and Δt is the sampling interval;

[0039] Step S2: The parameters of the second-order Thevenin equivalent model are identified online by using the forgetting factor recursive least squares algorithm to solve the problem that offline parameter identification data is not processed in time and cannot be represented in real time; the forgetting factor can assign a greater weight to the new data, strengthen the feedback effect of the new data in parameter identification, and gradually reduce the influence of the old data, thereby avoiding the information of the old data covering the information provided by the new data, strengthening the correction ability of the algorithm, and enabling the algorithm to always achieve the effect of rapid convergence; the recursive formula of the forgetting factor recursive least squares algorithm is shown in Table 5 below:

[0040]

[0041] Among them, the value range of λ is 0~1, generally 0.95~1; the closer λ is to 1, the better the simulation result, and the value of this simulation is 0.98; the smaller λ is, the faster the algorithm is, but it will cause fluctuations in the algorithm; when λ=1, it is a standard least squares recursive algorithm; the least squares parameter identification method with the forgetting factor added is applied to the equivalent model parameter identification of lithium batteries, and the parameters in the second-order RC equivalent circuit model are identified.

[0042]

[0043] Step S3: First, a state space model of the lithium-ion battery is established through the process model and observation model of the lithium-ion battery. The state variables of the lithium-ion battery model are the SOC and SOE of the lithium-ion battery, and the observation variable of the lithium-ion battery is the load voltage of the lithium-ion battery, as shown in the following formula:

[0044]

[0045] Among them, w k is the process noise of the system, ν k is the observation noise of the system, let w k ~N(0,Q),ν k ~N(0,R), Δt is the sampling period of the system; x k is the state variable, f(·) is the state function, y k is the observed variable, h(·) is the observed function; N SOC initial particles are generated using the prior probability p(x0) and N SOE initial particles The particle weights are =1 / N.

[0046] According to the system update equation, the prior probability sample of the next moment is obtained and After the system obtains new observations, a new particle set {x 0,i}+ , and then obtain the predicted value of the observed value {y 0,i} + , and then calculate the error between the observed value and the predicted value of each particle, and then use the error to find the weight of the particle. The particle weight equation is:

[0047]

[0048] Perform normalization:

[0049]

[0050] Use the new random sample distribution generated in the previous step to calculate the effective number of particles If the number of valid particles is less than the set threshold of the number of valid particles, a new particle set is generated by resampling (extracting particles with high weights and eliminating particles with low weights). After sampling, the weight of each particle in the new particle set is 1 / N; then the estimated value is output. is the estimated SOC value, Estimates for SOE:

[0051]

[0052] The genetic algorithm optimizes the particle set in the resampling process, introduces the biological evolution idea in the genetic algorithm into the particle filter, and replaces the resampling process in the particle filter; due to the unique optimization ability of the genetic algorithm, the efficiency of particle use is optimized, and many particles with different weights are added, which effectively suppresses the degradation of particle weights and the impoverishment of samples; the optimization process is as follows:

[0053] Step 1: Sample M particles x from the initial distribution P(x0) of the population i ,i=1,2,3…,n。

[0054] Step 2: According to the particle state transfer equation, get the updated particle x at time k i ,i=1,2,3…,n。

[0055] Step 3: Based on the measurement equation, calculate the weight of each particle in the particle set at time k i=1,2,3…,n.

[0056] Step 4: The fitness of each particle in the particle set is determined by the weight coefficient. After the genetic algorithm's isogenetic operation, the selection of multiple high-weight particles and the diversity of particles are taken into account to iterate a new particle set.

[0057] Step 5: Calculate the state estimate and variance estimate at time k:

[0058]

[0059] Step 6: Use the state equation f(·) to predict the particle at the next moment Right now i=1,2,3,…,n.

[0060] Step 7: Let k=k+1, and go to step 3 when the next measurement time comes.

[0061] The edge particle filter algorithm directly samples from the edge distribution of the state vector and uses the edge distribution function P(x 1:k |Z 1:k ,u 1:k ) instead of P(x k |Z 1:k-1 ,u 1:k )The predicted probability density is as follows:

[0062] P(x k |Z 1：k )∞P(z k |x k )∫P(x k |x k-1 )×P(x k-1 |Z 1：k-1 )dx k-1 (12)

[0063] Since the integral result cannot be obtained directly, and P(x k-1 |Z 1:k-1 ) can be represented by the particle set at time k-1 It is estimated that P(x k |Z 1:K-1 ) Available Approximate representation; for ease of calculation, the proposed distribution q(x k |Z 1:k ) adopts a method similar to P(x k |Z 1:k-1 ) in the form of

[0064]

[0065] In summary, the particle weight formula can be rewritten as:

[0066]

[0067] The MPF algorithm directly samples from the marginal distribution function, which can reduce the variance of the particle importance weight and increase the estimated value of effective sampling, thereby effectively suppressing the particle degradation phenomenon.

[0068] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is specifically described in the following embodiments in conjunction with the accompanying drawings; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0069] This case selected HPPC conditions and BBDST conditions at multiple temperatures to verify the new algorithm. In order to verify the estimation accuracy of the GM-PF algorithm for the joint estimation of SOC and SOE, the HPPC condition was used to carry out power lithium-ion battery test experiments at 15℃, 25℃, and 35℃ for the same ternary lithium battery. The discharge current at 15℃ and 25℃ was 1C, while the discharge current at 35℃ was 0.5C. The rated capacity at 15℃ was 65.76Ah, the rated capacity at 25℃ was 66.32Ah, and the rated capacity at 35℃ was 71.62Ah. In the verification experiment, the theoretical value of the battery SOC was calculated by the Ah integration method. According to the definition of SOE, the theoretical value of SOE was calculated by power integration. The estimation results of the three algorithms were compared with the theoretical values, and the estimation errors were integrated and analyzed. The estimation results of the joint estimation of SOC and SOE under different algorithms were compared through experimental data, and the GM-PF algorithm was verified and analyzed based on the comparison results.

[0070] When jointly estimating SOC and SOE, the tracking effect of the MPF algorithm is better than that of the PF algorithm, and the tracking effect of the GMPF algorithm is better than that of the MPF algorithm. The GMPF algorithm has the most stable estimation effect and the smallest error fluctuation. Under the HPPC operating condition of 15℃-35℃, the estimation accuracy of SOC is improved by an average of 77.90%, and the estimation accuracy of SOE is improved by an average of 81.94%. Under the BBDST operating condition of 15℃-35℃, the estimation accuracy of SOC is improved by an average of 84.86%, and the estimation accuracy of SOE is improved by an average of 78.8%. The feasibility of the GMPF algorithm for jointly estimating SOC and SOE of lithium-ion batteries is verified, and compared with the traditional PF algorithm, it has higher estimation accuracy and stronger robustness.

[0071] The implementation principle of a lithium battery SOC and SOE joint estimation method for solving particle depletion in an embodiment of the present application is as follows: the method is based on a second-order Thevenin equivalent circuit model of a lithium-ion battery, and combines a forgetting factor recursive least squares algorithm to perform online identification of model parameters, thereby ensuring real-time accuracy in a dynamically changing environment. An improved particle filter algorithm, especially a genetic algorithm, is used to optimize the resampling process in a traditional particle filter, thereby effectively avoiding the particle depletion phenomenon and improving the estimation accuracy. First, an equivalent circuit model of a lithium battery is established, and the problem of parameter update lag in the traditional method is solved through online parameter identification. Through a genetic marginalized particle filter algorithm, the model can effectively estimate the battery's SOC (battery state of charge). and SOE (battery state of energy), perform well in multi-temperature environments, and can adapt to battery performance under different temperature conditions such as 15°C, 25°C, and 35°C, ensuring accurate estimation results; experimental verification shows that it can not only improve the SOC and SOE estimation accuracy, but also effectively solve the particle degradation problem in traditional methods. Compared with traditional algorithms, the improved algorithm shows a more stable estimation effect under multi-temperature tests, especially under HPPC conditions, the SOC estimation accuracy is improved by an average of more than 77%, with higher application value and reliability, and provides a more efficient and accurate SOC and SOE joint estimation technology for lithium battery management systems, which is of great significance to battery performance evaluation and management under different environmental conditions.

[0072] The embodiments of the present application only take lithium-ion batteries as an example to illustrate the joint estimation of SOC and SOE of lithium-ion batteries based on genetic marginalized particle filtering. Therefore, all equivalent changes made according to the structure, shape, and principle of the present application should be covered within the scope of protection of the present application.

Claims

1. A method for jointly estimating SOC and SOE of a lithium battery to solve particle depletion, characterized in that: A second-order Thevenin equivalent circuit model is established for ternary lithium-ion batteries. The parameters of the second-order Thevenin equivalent model are identified online by the forgetting factor recursive least squares algorithm (FFRLS). The resampling process in the traditional particle filter algorithm is optimized and improved by the genetic algorithm. The Rao-Blackwell theory in statistics is used to marginalize some linear state variables in the particle filter calculation process, and the posterior distribution is approximated as a single Gaussian distribution. The Kalman filter is used to update the remaining nonlinear state variables when they are known, so as to achieve the establishment of the lithium-ion battery pack estimation model and the reliable operation of the mathematical iterative calculation algorithm for the SOC value and SOE value. The proposed genetic marginalized particle filter algorithm is verified under the HPPC and BBDST conditions at 15℃, 25℃ and 35℃, respectively, which solves the problems of the traditional particle filter algorithm, such as the difficulty in joint estimation of SOC and SOE, low precision and poor adaptability to ambient temperature.

2. A method for jointly estimating SOC and SOE of a lithium battery for solving particle depletion according to claim 1, characterized in that: For ternary lithium-ion batteries, a second-order Thevenin equivalent circuit model is established. This model can accurately simulate the polarization effect of the battery and accurately characterize the battery. The related calculations are relatively simple and easy to implement in engineering.

3. A method for jointly estimating SOC and SOE of a lithium battery for solving particle depletion according to claim 1, characterized in that: For the established second-order Thevenin equivalent circuit model, the forgetting factor recursive least squares algorithm (FFRLS) is used to perform online identification of the parameters of the second-order Thevenin equivalent model to solve the problem of untimely data processing and inability to represent in real time in offline parameter identification.

4. The method for jointly estimating SOC and SOE of a lithium battery for solving particle depletion according to claim 1, characterized in that: The genetic algorithm is used to optimize and improve the resampling process in the traditional particle filter algorithm, solving the particle depletion problem existing in the traditional particle filter algorithm.

5. The method for jointly estimating SOC and SOE of a lithium battery for solving particle depletion according to claim 1, characterized in that: The Rao-Blackwell theory in statistics is used to marginalize some linear state variables in the particle filter calculation process, and the posterior distribution is approximated as a single Gaussian distribution. The Kalman filter is used to update the remaining nonlinear state variables when they are known, thereby improving the prediction accuracy.

6. A method for jointly estimating SOC and SOE of a lithium battery for solving particle depletion according to claim 1, characterized in that: For the proposed genetic marginalized particle filter algorithm, verification experiments were carried out under HPPC and BBDST conditions at temperatures of 15℃, 25℃ and 35℃, respectively. The problems of traditional particle filter algorithm in joint estimation of SOC and SOE of lithium-ion batteries, such as difficult estimation, low accuracy and poor adaptability under different ambient temperatures, were solved.

Images