A mechanism- and data-driven joint estimation method for SOC-SOH of lithium batteries

Abstract

This invention discloses a mechanism- and data-driven joint estimation method for lithium-ion batteries' State of Charge (SOC) and State of Health (SOH), comprising the following steps: First, a least squares support vector machine (LSSVM) prediction model is trained using a multidimensional dataset reconstructed from the phase space. Then, a state equation is established based on the ampere-hour integral method and an empirical model of capacity decay. An interim particle filter (IUPF) observation equation is established using the LSSVM prediction model, and the improved interim particle filter (IUPF) algorithm is substituted into the model for state estimation and parameter updates. Finally, the updated LSSVM model parameters are iteratively applied to the training dataset, achieving joint prediction of battery SOC and SOH and online updating of the data-driven model. This invention proposes a mechanism- and data-driven joint estimation method for lithium-ion batteries' SOC and SOH, achieving joint estimation of the battery's state of charge (SOC) and state of health (SOH), and considering system state noise and measurement noise to ensure the accuracy and stability of the battery state estimation results.

Description

TECHNICAL FIELD

[0001] The application relates to a lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving and belongs to the technical field of electric power energy storage. BACKGROUND

[0002] With the continuous promotion of the double carbon target, the installed capacity of intermittent new energy power generation represented by wind power and photovoltaic power generation in the power system has been significantly improved, and the flexible regulation demand of the power system such as frequency modulation and peak regulation has also increased. The conventional pumped storage regulation mode cannot be constructed and developed on a large scale due to geographical environmental constraints, and the electrochemical energy storage system represented by lithium ion batteries has become an important regulation resource of the new power system dominated by new energy due to its flexible configuration, convenient installation and rapid regulation. At the same time, due to the closed nature of the physical structure of lithium batteries as an electrochemical reaction system, it is difficult to monitor and diagnose the operating state of lithium batteries. Traditional lithium battery state of charge (SOC) estimation is mainly based on model-based methods combined with Kalman filter series algorithms, which are limited in terms of environmental dependence and model accuracy, and have problems such as low estimation accuracy, weak model generalization ability, and sensitivity to initial conditions. Most lithium battery state of health (SOH) estimation uses neural network algorithms, but neural networks may not be suitable for other types or batches of batteries, and as a "black box" model, its internal decision-making process is difficult to explain and limited, and there are problems in model generalization ability and interpretability. Moreover, the separate battery state estimation of lithium battery SOC and SOH ignores the coupling between the two, so that the speed and accuracy of SOC and SOH estimation cannot meet the needs of practical applications. Especially in some extreme accidents, thermal triggers, mechanical collisions, and the like, timely early warning cannot be made, thereby causing the spread and expansion of battery system failures and causing great harm to people's life and property safety. Therefore, it is necessary to consider the coupling relationship between the operating SOC and SOH of lithium batteries, combine the mechanism model with the data model, carry out joint estimation of the SOC and SOH of lithium batteries, and further improve the level of lithium battery operating state monitoring, identification and diagnosis warning. SUMMARY

[0003] The application aims at the defects and deficiencies of the prior art, and provides a lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving. Based on the least square support vector machine-improved unscented Kalman particle filter (LSSVM-IUPF) framework, the three-layer IUPF algorithm is used to achieve accurate joint estimation of SOC-SOH with optimal state estimation parameters as the target, so as to reduce the estimation error of the battery health state SOH and the battery capacity state SOC.

[0004] The technical scheme adopted by the present application to solve the technical problems is: a lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving, comprising the following steps:

[0005] Step 1: Construct a multi-dimensional data set of the time sequence of the SOC value and the current maximum available capacity of the lithium battery respectively, reconstruct the training data from the initial time to the k-1 time using the Takens embedding method, and generate a d-dimensional vector in the phase space.

[0006] Step 2: Reconstruct the phase space, and reconstruct the multi-dimensional data set into a phase space with the same topological meaning as the dynamic system.

[0007] Step 3: Offline training of the reconstructed least square support vector machine (LSSVM, Least Square Support Vector Machines) prediction model.

[0008] Step 4: According to the ampere-hour integration method and the battery capacity degradation empirical model, the state equations for SOC and SOH estimation are established respectively; the observation equations for SOC and SOH estimation are established by the LSSVM online prediction model.

[0009] Step 5: According to the training data from the initial time to the k-1 time, the offline model of LSSVM is trained, and the LSSVM parameters of the regression prediction model at the k time are estimated using the improved unscented particle filter algorithm (IUPF, Intermediate UPF).

[0010] Step 6: The IUPF algorithm observes new data at the k time, and the system uses an incremental learning method to learn and update only the information of the new data, updates the particle weight, and combines the LSSVM online model to give further prediction values of SOC, SOH and LSSVM parameters.

[0011] Step 7: The updated LSSVM model parameters are iterated to the training data set for state estimation at the k+1 time, so as to realize the joint prediction of the battery SOC and SOH and the online update of the data driven model.

[0012] Among them, step 4 further comprises the following steps:

[0013] Step 4.1: According to the ampere-hour integration method and the battery capacity degradation empirical model, the IUPF state equations for SOC and SOH estimation are established respectively;

[0014] Step 4.2: The SOC and SOH obtained by offline LSSVM training are used as virtual observation values, and the LSSVM regression model is used as an observation equation.

[0015] Further, the method for establishing the state equation of the model, the state equation of the SOC estimation established based on the ampere-hour integral method is discretely expressed as follows:

[0016]

[0017] SOC k is the SOC value of the lithium battery at time k; Δt k is the system sampling time; C k is the available capacity of the battery at time k; I k is the system input current at time k.

[0018] Further, the method for establishing the observation equation of the model, the discretization expression of the state equation of the capacity based on the empirical model of capacity attenuation is as follows:

[0019] C k+1 = η c C k + β1exp(-β2 / Δt k ) + ξ k

[0020] In the formula, C k is the available capacity of the battery at time k; U oc,k is the open circuit voltage; ξ k is the state noise; β1 and β2 are coefficients to be determined.

[0021] Further, the SOC-SOH joint estimation method of multi-method fusion, the definition of the state of health SOH of the lithium ion battery is as follows:

[0022]

[0023] In the formula, SOH k is the SOH value calculated at time k, C new is the factory capacity of the battery; C current is the current maximum available capacity of the battery.

[0024] Further, the SOC-SOH joint estimation method of multi-method fusion, the discretization expression of the state equation of the SOC and SOH established by the LSSVM observation equation is as follows:

[0025] y k = f(x k-1 , s k ) + ι k

[0026] In the formula, x i is the state vector, s k is the LSSVM parameter, l k is the Gaussian white noise.

[0027] wherein step 6 further comprises the following steps:

[0028] Step 6.1: initialize n initial state particles x k ;

[0029] Step 6.2: apply the improved UKF algorithm to each sampling particle for weight optimization;

[0030] Step 6.3: calculate the importance weight of the particle and normalize it;

[0031] Step 6.4: use the resampling algorithm to duplicate and eliminate the particles, and recalculate the weight;

[0032] Step 6.5: perform state estimation of the corresponding SOC or SOH;

[0033] Step 6.6: perform online parameter update of LSSVM.

[0034] An SOC-SOH joint estimation model system, comprising:

[0035] The SOC and the current capacity state equation of the lithium ion battery are constructed by using the ampere-hour integration method and the battery capacity attenuation method, and the IUPF-LSSVM framework is used to realize online training of the data-driven model and further prediction of the battery state parameters through iteration.

[0036] Further, the SOC-SOH joint estimation model system described above, the IUPF-LSSVM framework module is specifically used to realize online training of the data-driven model and further prediction of the battery parameters through iteration.

[0037] Further, the SOC-SOH joint estimation model system described above, the optimal state estimation module outputs the estimated SOH value and SOC value.

[0038] A computer readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the SOC-SOH joint estimation method based on multi-method fusion described above.

[0039] An apparatus (controller / computer device / electronic device) comprising: a memory for storing data; a processor for executing the algorithm instructions to enable the apparatus to perform the operations of a mechanism and data fusion driven lithium battery SOC-SOH joint estimation method as described above.

[0040] Advantages:

[0041] 1、The model of the joint state estimation of the current capacity state SOC and the health state SOH of the battery is proposed, the LSSVM-IPUF algorithm is autonomously designed, the problems of low estimation accuracy and slow convergence speed of single algorithm are overcome, higher accuracy of the SOC and SOH value estimation of the lithium battery is realized, and stable operation of the battery is ensured.

[0042] 2、The present application aims to make the lithium ion battery have accurate estimation of the state of charge and the state of health due to the possible overcharge or overdischarge of the lithium ion battery, optimize the limitations of single algorithm, realize more accurate estimation by using double-layer algorithm, make the estimation of the current capacity and the health state of the lithium ion battery have higher accuracy and robustness, ensure the safe and flexible operation of the power system, and provide technical support for efficient use of lithium batteries.

[0043] 3、The system architecture proposed in the present application reduces the SOH estimation error caused by the increase of ohmic resistance due to the thickening of SEI film layer caused by multiple charging and discharging by real-time updating of the SOC value and the terminal voltage into the estimation of SOH, and real-time updating of the ohmic resistance into the state estimation of SOC, realizes bidirectional flow of information, and can make the state estimation still have the advantages of stability and small estimation error under complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0044] Figure 1 Fig. 1 is a schematic diagram of the SOC-SOH joint estimation block diagram based on the LSSVM-IUPF algorithm of the present application;

[0045] Figure 2 Fig. 2 is a schematic diagram of the LSSVM-IUPF algorithm flow of the present application. DETAILED DESCRIPTION

[0046] It should be noted that:

[0047] The Panasonic 18650 type ternary lithium battery with rated capacity 1800mAh, maximum voltage 4.2V and cut-off voltage 2.5V is taken as the experimental object.

[0048] The technical scheme of the present application will be described in detail below through the drawings and specific embodiments, and it should be understood that the specific features in the embodiments and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, and are not limitations of the technical scheme of the present application, and the technical features in the embodiments and the embodiments can be combined with each other without conflict.

[0049] Embodiment 1

[0050] The embodiment provides a mechanism and data fusion driven minimum vector machine-incremental unscented Kalman particle filtering (LSSVM-IUPF) framework, and comprises the following steps.

[0051] Step 1: respectively constructing time series multidimensional data sets of lithium battery SOC values and current maximum available capacities, reconstructing training data from an initial time to k-1 time by using a Takens embedding method, and generating d-dimensional vectors in a phase space;

[0052] In step 1, the following contents are included.

[0053] Step 1.1: respectively constructing time series multidimensional data sets of lithium battery SOC values and current maximum available capacities, initializing N initial state particles Wherein, i represents the i th particle, represents the SOC value at the initial time, represents the current maximum available capacity at the initial time, and x k obeys a proposal distribution function.

[0054] Step 1.2: the Takens embedding method is as follows.

[0055] The relationship between one-dimensional nonlinear lithium battery SOC values and time and the relationship between the previous maximum available capacity and time, Takens proposes that the phase trajectory of a chaotic system still has a certain structure, which can be used to identify and analyze chaotic behavior, assuming that a single variable time series under a nonlinear dynamic system When the phase space embedding dimension d is greater than a certain value, there is a smooth mapping f (.), and the generated d-dimensional vector is as follows:

[0056] z k+1 =f(z k ,z k-1 ,z k-2 …,z k-d+1 )

[0057] Wherein, z k represents particle training data at k time, and k is in the range of 1 to N.

[0058] Step 2: reconstructing the phase space to reconstruct the phase space with the same topological meaning as the dynamic system from the multidimensional data set.

[0059] In step 2, the following contents are included.

[0060] Step 2.1: reconstructing the phase space to reconstruct the phase space with the same topological meaning as the dynamic system from the multidimensional data set, and the phase space reconstruction process is as follows.

[0061]

[0062] where x n is the input vector of the smooth map f(.), d+1 z d+2 ,...,z N ) T is the output vector of the corresponding smooth map f(.).

[0063] Step 2.2: Determine the minimum dimension d of phase space reconstruction by the false neighbor method.

[0064] Step 3: Put the reconstructed data into the LSSVM prediction model for offline training;

[0065] wherein step 3 comprises the following contents:

[0066] Step 3.1: The LSSVM model can be expressed according to the statistical theory and Gaussian kernel function as follows:

[0067]

[0068] wherein a = [a1, a2,..., a n ] is the support vector factor, x k = (x1, x2,..., x N ) is the support vector parameter, and b is the bias.

[0069] In addition, σ represents the radial basis kernel width;

[0070] Step 3.2: The further prediction model can be obtained by bringing in the reconstructed data as follows:

[0071]

[0072] Step 4: According to the ampere-hour integration method and the battery capacity degradation empirical model, the state equations for SOC and SOH estimation are respectively established; and the observation equations for SOC and SOH estimation are respectively established through the LSSVM online prediction model.

[0073] wherein step 4 comprises the following steps:

[0074] Step 4.1: According to the ampere-hour integration method and the battery capacity degradation empirical model, the IUPF state equations for SOC and SOH estimation are respectively established;

[0075] Further, the discretized expression of the state equation for the current capacity SOC estimation of the lithium ion battery based on the ampere-hour integration method in step 4.1 is as follows:

[0076]

[0077] SOC kLet Δt be the SOC value of the lithium battery at time k; k C is the system sampling time; k I represents the available battery capacity at time k. k Let k be the system input current at time k.

[0078] Furthermore, the discretized representation of the state equation for estimating the current capacity of the lithium-ion battery based on the empirical equation for battery capacity degradation in step 4.1 is as follows:

[0079] C k+1 =η c C k +β1exp(-β2 / Δt k )+ξ k

[0080] Among them, C k Let ξ be the available battery capacity at time k; k β1 and β2 are state noise; β1 and β2 are coefficients to be determined.

[0081] Step 4.2: The SOC and SOH obtained from offline LSSVM training are used as virtual observations, and the LSSVM regression model is used as the observation equation.

[0082] Furthermore, the LSSVM observation equation in step 4.4 is as follows:

[0083] y k =h(x k-1 ,s k )+ι k

[0084] Where, x i It is a state vector, s k These are LSSVM parameters, l k It is Gaussian white noise.

[0085] Step 5: Train the offline LSSVM model based on the training data from the initial time to time k-1, and estimate the LSSVM parameters of the regression prediction model at time k using the improved unscented particle filter algorithm (IUPF, Intermediate UPF).

[0086] Step 6: When the IUPF algorithm observes new data at time k, the system uses an incremental learning method to learn and update only the information of the new data, update the particle weights, and combine the LSSVM online model to give further predicted values ​​of SOC, SOH and LSSVM parameters.

[0087] Step 6 includes the following steps:

[0088] Step 6.1: Initialize n initial state particles xk :

[0089] When k = 0, N particles are selected in the data sample x0 for initialization, and the particles satisfy the prior distribution function, that is,

[0090]

[0091]

[0092] wherein the value range of i is 1, 2,..., N. represents the mean value calculated according to the data sample, represents the covariance calculated according to the data sample.

[0093] Step 6.2: Apply the improved UPF algorithm to each sampling particle for weight optimization;

[0094] When k = 0, the particle mean and covariance are calculated based on the UKF algorithm:

[0095]

[0096] When k ≠ 0, importance sampling is performed, and the improved UKF algorithm is applied to optimize the covariance matrix calculation problem.

[0097] 1) Perform sigma point calculation:

[0098]

[0099] wherein, is the mean value of the weight, is the covariance of the weight, and β is 2 when solving the Gaussian problem. Proportional correction coefficient λ = α 2 (L + κ) - L, κ is a quadratic sampling factor.

[0100] Since there is no valid value for the singular value square root, the improved UKF algorithm will perform singular value decomposition (SVD) on the UKF algorithm:

[0101]

[0102] wherein U is the left singular vector, S is called the singular value matrix, and V is the right singular vector.

[0103] After decomposition, the sigma point is calculated through the following formula:

[0104]

[0105] where j takes values 1, 2,..., 2L, L is the dimension of the sigma points.

[0106] 2) The state value and the predicted covariance of the particle are obtained from the further prediction of the sigma sampling points:

[0107]

[0108] 3) The observation prediction value is obtained from the further state prediction value:

[0109]

[0110] 4) The self-covariance and cross-covariance are calculated with the weighted values:

[0111]

[0112] 5) The Kalman gain is calculated, and the state quantity and the covariance are updated:

[0113]

[0114] 6) The important density function is generated, and the particles are sampled to generate:

[0115]

[0116] Step 6.3: Calculate the importance weight of the particle and normalize it to make the sum equal to 1;

[0117] 1) Calculate the weight of each particle:

[0118]

[0119] 2) Normalize:

[0120]

[0121] Step 6.4: Use the resampling algorithm to set the threshold value N max to determine the number of effective particles N eff , duplicate and eliminate the particles, and recalculate the weight;

[0122]

[0123] Step 6.5: Perform the corresponding SOC or state estimation of the current capacity of the particle update;

[0124]

[0125] Step 6.6: Perform the parameter update of the online estimation LSSVM.

[0126] Further, the state equation and observation equation of the IUPF algorithm framework of step 6.6 are as follows:

[0127]

[0128] where s k = [γ k σ k ], γ k is a regularization term coefficient, σ k is a kernel parameter, s k is a state vector in the IUPF state space equation, ω k , ξ k are mutually independent Gaussian white noise.

[0129] Step 7: The updated LSSVM model parameters are iterated to the training data set for state estimation at time k+1, so as to realize the joint prediction of the battery SOC and SOH and the online update of the data-driven model.

[0130] Embodiment 2

[0131] The embodiment provides a SOC-SOH joint estimation model, and the steps are as follows:

[0132] The state equations of the SOC and the current maximum available capacity of the lithium ion battery are respectively constructed by using the ampere-hour integration method and the battery capacity attenuation method, and the online training data-driven model and the further prediction of the battery state parameters are realized by using the IUPF-LSSVM framework through iteration, and the state estimation of the SOC value and the SOH value is realized by using the optimal state estimation module.

[0133] Step 1: Initialize the voltage and current state and the LSSVM parameter vector and weight;

[0134] Step 2: The state equations of the SOC and the current maximum available capacity of the lithium ion battery are respectively constructed by using the ampere-hour integration method and the battery capacity attenuation method, and the measurement equation of the lithium ion battery SOC and the current maximum available capacity is constructed by using the LSSVM, and the SOC and the current maximum available capacity are transmitted to the IUPF algorithm;

[0135] Step 3: The SOC value and the SOH value at time k+1 are calculated through the prediction and update module of the IUPF algorithm;

[0136] Step 4: The value of the current maximum available capacity is updated and fed back to the state equation of the SOC state space mathematical model, and the LSSVM parameter vector at the next time is updated;

[0137] Step 5: The SOC value and the SOH value of the lithium ion battery are realized to be jointly predicted and iterated repeatedly.

[0138] Those skilled in the art will appreciate that embodiments of the present application can be readily used as software, hardware, or a combination of software and hardware. In a software embodiment, the methods can be tangibly embodied in a machine-readable storage medium having stored thereon instructions that can be used to program a processing system to perform the methods. The machine-readable storage medium can be magnetic (e.g., magnetic disks), optical (e.g., optical discs), electrical (e.g., memory devices), or any combination thereof. Such software can be transmitted using any apparatus adapted to transfer a set of instructions from one place to another. A machine-readable storage medium is a computer-related media that stores computer-readable instructions, data structures, program modules or other data. Examples of computer- readable media include RAM, ROM, EPROM, EEPROM, floppy disks, CD-ROMs, DVD-ROMs, SAS, SCSI, data signals, data streams, and any other medium suitable for storing computer program code. The software can be transmitted using any data transmission techniques, e.g., communication network, wireless network, and the like.

[0139] The present application is described in relation to flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It is understood that each block of the flow diagrams and / or block diagrams, and combinations of blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks. Figure 1 means for performing each of the functions specified in the flow diagram and / or block diagram block or blocks.

[0140] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks. Figure 1 means for performing each of the functions specified in the flow diagram and / or block diagram block or blocks.

[0141] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more functions specified in the flow diagram and / or block diagram block or blocks. Figure 1 means for performing each of the functions specified in the flow diagram and / or block diagram block or blocks.

[0142] The embodiments of the present application described above are illustrative, and are not meant to limit the scope of the present application to the exact construction details shown and described. Those skilled in the art will be able to affect various changes and modifications without departing from the scope and spirit of the present application, which are intended to be limited only by the scope of the appended claims.

Claims

1. A lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving, characterized in that, Comprising the following steps: Step 1: Constructing a multi-dimensional data set of the time sequence of the lithium battery SOC value and the current maximum available capacity respectively, reconstructing the training data from the initial time to the k-1 time in the phase space to generate a d-dimensional vector; Step 2: Reconstructing the multi-dimensional data set into the phase space with the same topological significance as the dynamic system; Step 3: Offline training of the reconstructed least square support vector machine (LSSVM) prediction model; Step 4: Establishing the state equation of SOC and SOH estimation according to the ampere-hour integration method and the battery capacity degradation empirical model respectively; and establishing the observation equation of SOC and SOH estimation by the LSSVM online prediction model; Step 4.1: Establishing the IUPF state equation of SOC and SOH estimation according to the ampere-hour integration method and the battery capacity degradation empirical model respectively; Step 4.2: Using the SOC and SOH obtained by offline LSSVM training as virtual observation values, and using the LSSVM regression model as the observation equation; Step 5: Training the offline model of LSSVM according to the training data from the initial time to the k-1 time, and estimating the LSSVM parameters of the regression prediction model at the k time by using the improved intermediate UPF (IUPF) algorithm; Step 6: The IUPF algorithm observes new data at the k time, and the system uses an incremental learning method to learn and update the data information, updates the particle weight, and combines the LSSVM online model to give further prediction values of SOC, SOH and LSSVM parameters, wherein the state equation and the observation equation of the IUPF algorithm framework are as follows: where s k = [γ k σ k ], γ k is a regularization term coefficient, σ k is a kernel parameter, s k is a state vector in the IUPF state space equation, ω k , ξ k are mutually independent Gaussian white noises; The state equation of SOC estimation based on the ampere-hour integration method is discretely represented as follows: SOC k is the SOC value of the lithium battery at time k; Δt k is the system sampling time; I k is the system input current at time k; Step 7: Iterating the updated LSSVM model parameters to the training data set for state estimation at the k+1 time, so as to realize the joint prediction of the battery SOC and SOH and the online update of the data-driven model.

2. The lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving according to claim 1, characterized in that, The state equation of capacity based on the capacity attenuation empirical model is discretely represented as follows: C k+1 = η c C k + β1exp(-β2 / Δt k )+ ξ k In the formula, C k is the available capacity of the battery at time k; ξ k is the state noise; and β1 and β2 are coefficients to be determined.

3. The SOC-SOH co-estimation method of claim 1, wherein, The definition of the state of health SOH of the lithium ion battery is as follows: In the formula, SOH k is the SOH value calculated at the kth moment, C new is the battery factory capacity; C current is the current maximum available capacity of the battery.

4. The lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving according to claim 1, characterized in that, The LSSVM observation equation constructs the state equation of SOC and SOH discretely represented as follows: y k = f(x k-1 , s k )+i k where s k is the LSSVM parameter, l k is the Gaussian white noise.

5. The lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving according to claim 1, characterized in that, Step 6 further comprises the following steps: Step 6.1: Initialize n initial state particles x k ; Step 6.2: Applying the improved UKF algorithm to each sampling particle for weight optimization; Step 6.3: Calculating the importance weight of the particle and normalizing it; Step 6.4: Using the resampling algorithm to copy and eliminate the particles, and recalculating the weight; Step 6.5: Performing the corresponding state estimation of SOC or SOH; Step 6.6: Performing the parameter update of the online estimation LSSVM.

6. A SOC-SOH combined estimation model system, characterized in that, The system is used to realize the lithium battery SOC-SOH joint estimation method based on mechanism and data fusion driving according to any one of claims 1-5, comprising: The SOC and the current capacity state equation of the lithium ion battery are respectively constructed by using the ampere-hour integration method and the capacity attenuation method, the IUPF-LSSVM framework is used to realize the online training of the data-driven model and the further prediction of the battery state parameters through iteration, and the optimal state estimation module is used to realize the state estimation of the SOC value and the SOH value. The IUPF-LSSVM framework module is specifically used for realizing the online training of the data-driven model and the further prediction of the battery parameters through iteration. The optimal state estimation module outputs the estimated SOH value and SOC value.

Images