Double-path battery health state estimation method based on electrical coupling model

Abstract

The invention provides a dual-path battery health state estimation method based on an electrical coupling model, which comprises the following steps: step 1, establishing an electrical coupling model of a battery, the electrical coupling model comprising a gas-liquid dynamic model and an equivalent circuit model; 2, performing parameter identification on the gas-liquid dynamic model and the equivalent circuit model respectively; 3, constructing a joint estimation state equation based on the key parameters obtained through identification, and carrying out fusion estimation on the battery state; and 4, providing a battery health state estimation method based on the gas-liquid coupling model. According to the method, through dual prediction of the gas-liquid dynamic model and the equivalent circuit model, high-precision parameters of the gas-liquid dynamic model and structural simplicity and convenience of the equivalent circuit are considered, and accuracy and robustness of a prediction result are ensured; meanwhile, the parameter identification of the gas-liquid dynamic model is convenient, the parameter is used as the aging characteristic of the battery, the physical significance of the model is reflected, and errors generated by model prediction iteration can be avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of battery health management, and particularly to a double-path battery state of health estimation method based on an electrical coupling model. BACKGROUND

[0002] In the whole life cycle of the battery, the battery performance will inevitably age due to complex electrochemical side reactions, loss of active lithium ions, structural degradation of electrode materials, etc. This aging phenomenon is intuitively manifested in two aspects: one is the decay of electrical energy storage capacity, i.e., the actual dischargeable capacity of the battery decreases; the other is the decline of fast charge and discharge capability, i.e., the internal resistance of the battery increases, resulting in reduced charge and discharge efficiency and increased thermal management pressure. In order to quantify the degree of battery aging, the industry introduces the health state as a key indicator. The state of health (SOH) is a percentage indicator used to evaluate the current performance of the power battery relative to its brand-new state, and is a core parameter for evaluating the current use of the battery, predicting the remaining service life, and ensuring the safe operation of the battery.

[0003] Accurate and reliable estimation of SOH has become one of the key technical challenges of battery management systems. Current estimation methods can be divided into model-based and data-driven methods. Among them, the electrochemical model and the equivalent circuit model are representative of the model-based method: the electrochemical model starts from the reaction mechanism inside the battery, has clear physical meaning and high accuracy, and can reflect the mapping relationship between the electrochemical parameters and the external voltage, but the model is complex, the calculation burden is large, and it is difficult to realize real-time estimation under dynamic working conditions; the equivalent circuit model describes the polarization and hysteresis of the battery through an RC network, has the advantages of simple structure, easy parameter identification, and high computational efficiency, but the model parameters often lack a clear correspondence with the SOH reflecting the internal health state of the battery, and because the parameters lack actual constraints in the identification process, the identified parameters may deviate from the actual values, thus limiting the estimation accuracy and robustness in practical applications.

[0004] In summary, the existing battery state of health estimation is not ideal in practical applications, while the double-path estimation technology based on the gas-liquid dynamics model and the equivalent circuit model provided by the present application can well fill or compensate for the shortcomings of existing battery state of health estimation technologies. SUMMARY

[0005] The present application couples the gas-liquid dynamics model and the equivalent circuit model, proposes an electrical coupling model, and realizes a double-path battery state of health estimation method; the gas-liquid dynamics model reflects the change in battery capacity, the equivalent circuit model reflects the change in battery internal resistance, and the double-path estimation improves the estimation accuracy and robustness;

[0006] The technical scheme of the present application is: a double-path battery state of health estimation method based on an electrical coupling model, characterized in that it comprises the following steps:

[0007] Step one: constructing an electrical coupling model of the battery, which comprises a gas-liquid dynamics model and an equivalent circuit model;

[0008] Step two: respectively performing parameter identification on the gas-liquid dynamics model and the equivalent circuit model to obtain key parameters of the models;

[0009] Step three: based on the identified key parameters, constructing a joint estimation state equation to perform fusion estimation on the battery state;

[0010] Step four: proposing a battery state of health estimation method based on the gas-liquid coupling model;

[0011] In the above scheme, the state equation for constructing the gas-liquid dynamics model in step one is formula one to formula two: In the formula, SOC q is the SOC estimated by the gas-liquid dynamics model at time t, U(t) is the terminal voltage of the gas-liquid dynamics model at time t, k1, k2, k3, and k4 are to-be-identified parameters, T1 and T2 are the initial temperature and the temperature at the current sampling time, OCV is the open-circuit voltage, I is the current, η is the coulomb efficiency, C0 is the rated capacitance, OCV1 is the initial open-circuit voltage, || is the absolute value symbol, is the derivative of SOC with respect to time t;

[0012] The state equation for constructing the equivalent circuit model in step one is formula three to formula four: V(t)=OCV(SOC d (t))+V1(t)+R0I(t) (formula four) In the formula, SOC d is the SOC estimated by the equivalent circuit model, V1 is the polarization voltage, R0 is the ohmic resistance, R1 and C1 are the polarization resistance and the polarization capacitance, V(t) is the terminal voltage of the gas-liquid dynamics model at time t, is the derivative of V1 with respect to time t;

[0013] The parameter identification method of the equivalent circuit model based on the least square method in the second step is formula five to formula seventeen; in order to facilitate online parameter identification, the discrete model formula five of the battery is established according to formula five, the voltage equation of formula five is reconstructed into formula six, a recursive form with a forgetting factor, which contains three parameters α, β and γ; then, formula seven defines the system output as the difference between the battery terminal voltage and the open circuit voltage under the current SOC, and a standard linear regression model formula eight is constructed, formula nine shows that the observation vector of the equivalent circuit model is composed of historical system output values, current and historical current values, and formula ten shows that the parameters are contained;

[0014] The core of parameter identification is to use the recursive least square algorithm for online real-time update: first, formula eleven calculates the prior error, that is, the deviation between the current measurement output and the predicted value based on the last time parameter; then, formula twelve calculates the gain vector, which weighs the weight in parameter update; then, formula thirteen updates the parameter estimation vector using the gain vector and the prior error; finally, formula fourteen updates the covariance matrix to ensure the convergence and stability of the algorithm; after obtaining the updated linear parameters, the formula ten is converted into the equivalent circuit model element parameters with clear physical meaning through the preset parameter mapping relationship, including formula fifteen representing the polarization resistance, formula sixteen representing the ohmic resistance and formula seventeen representing the polarization capacitance; V(k) = OCV(SOC d (k)) + V1(k) + R0I(k) (formula five) V(k) = αV(k-1) + βI(k) + γI(k-1) + (1-α)OCV(SOC d (k)) (formula six) y(k) = V(k) - OCV(SOC d (k)) (formula seven) y(k) = φ(k) T θ (formula eight) φ(k) = [y(k-1), I(k), I(k-1)] T (formula nine) θ = [α, β, γ] T (formula ten) e(k) = y(k) - φ(k) T θ(k-1) (formula eleven) θ(k) = θ(k-1) + K(k)e(k) (formula thirteen) where V(k) is the terminal voltage of the equivalent circuit at discrete point k, SOC d (k) is the state of charge of the equivalent circuit at discrete point k, I(k) is the current at discrete point k, V1(k) is the polarization voltage of the equivalent circuit at discrete point k; α is an autoregressive coefficient, β is a current coefficient, and γ is a historical current coefficient; y(k) is a system output, φ(k) is an observation vector, e(k) is a prior error, K(k) is a gain vector, θ(k) is a parameter estimation vector, P(k) is a covariance matrix, λ is a forgetting factor; lnα is a natural logarithm of α;

[0015] The parameter identification method based on the gas-liquid dynamic model and the extended Kalman filter is formula eighteen to formula twenty-two:

[0016] State vector: x q = [SOC q , k1, k2, k3, k4] T (formula eighteen) x q (k) = f(x q (k-1), I(k-1)) + ω(k-1) (formula nineteen) where: where SOC q is the charge state estimated by the gas-liquid dynamics model; ω(k) is process noise;

[0017] Observation equation: y q (k) = U(k) = h(x q (k), I(k)) + υ(k) (formula twenty-one) where: where υ(k) is observation noise;

[0018] ① Parameter initial state: x q_0 = [SOC q_0 , k 1_0 , k 2_0 , k 3_0 , k 4_0 ] T , Q q , R q , P q (formula twenty-three) where x q_0 , as a model state vector, includes the battery state of charge and four key physical parameters, and the subscript _0 of the letter indicates the initial value of the parameter; Q q is a covariance matrix of process noise, and R qis the covariance matrix of the observation noise, P q is the error covariance matrix;

[0019] ② Time update:

[0020] ③ Measurement update:

[0021] ④ Calculate the Kalman gain:

[0022] ⑤ Update the state estimate:

[0023] ⑥ Update the error covariance

[0024] The SOC of the gas-liquid dynamics model is obtained through the extended Kalman iteration q k1, k2, k3, k4 provide important information for subsequent SOH estimation;

[0025] Step three: Establish the state equation of the joint estimation of the gas-liquid dynamics model and the equivalent circuit model; x[t] = [SOC, k3, R0, R1] T (Equation thirty) SOC - (k) = A(k-1)SOC(k-1)A T (k-1) + Q (Equation thirty-two) K(k) = SOC - (k)C T (k) [C(k) SOC - (k)C T (k) + R] -1 (Equation thirty-three) SOC(k) = [I - K(k)C(k)] SOC - (k) (Equation thirty-five) In the formula, x[t] constructs a four-dimensional joint state vector integrating the battery core state and key model parameters, indicates the state prediction equation, K(k) indicates the Kalman gain K(k), and equation thirty-five updates the state Where A(k) and C(k) are Jacobian matrices;

[0026] Step four proposes a battery state of health estimation method based on a gas-liquid coupling model;

[0027] The ohmic internal resistance R0 is identified based on an equivalent circuit model, and the internal resistance method state of health SOH is calculated R : In the formula, R 0,initial is the initial internal resistance of the battery, R 0,end is the internal resistance at the end of life, and R0 is the ohmic internal resistance value identified online at the current time;

[0028] Based on the gas-liquid dynamics model parameters, the capacity method state of health SOH is calculated by actual capacity estimation c ;

[0029] SOH R and SOH c are fused by adaptive weights to obtain the final SOH estimation; In the formula, λ R and λ C are adaptive weights, is the variance of the internal resistance method SOH estimation, is the variance of the capacity method SOH estimation, and ε is a constant, which is 10 -6 ;

[0030] The gas-liquid dynamics model and the equivalent circuit model are coupled in the above scheme, and the battery SOH is estimated; only the open circuit voltage curve of the battery is relied on to complete the parameter identification of the two models, the k3 and SOC of the gas-liquid dynamics model are relied on to complete the capacity method estimation, and the R0 of the equivalent circuit model is relied on to complete the battery internal resistance method estimation;

[0031] Compared with the prior art, the beneficial effects of the present application are: through the double-path prediction of the gas-liquid dynamics model and the equivalent circuit model, the high-precision parameters of the gas-liquid dynamics model and the structural simplicity of the equivalent circuit are taken into account, the accuracy and robustness of the prediction result are ensured; at the same time, the parameter identification of the gas-liquid dynamics model is convenient, the identified parameters extracted can be independently analyzed, as the aging characteristics of the battery, can reflect the physical meaning of the model, and can avoid the error generated by model prediction iteration; the present application can be extended to health state estimation under battery state of charge estimation, rely on the offline open circuit voltage curve of the battery, and perform online estimation, which is simple and convenient to realize in the battery management system, and makes the data analysis structure of the management system more complete. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 Flow of the present application based on the health state estimation of the electrical coupling model;

[0033] Figure 2 SOC and OCV curves of the 5-string battery module in different aging stages provided by the present application for parameter identification;

[0034] Figure 3 Parameter identification results of R0 in the present application;

[0035] Figure 4 SOH estimation results according to internal resistance and capacity in the present application. Specific implementation

[0036] The present application will be further described in detail below in combination with the drawings and specific embodiments, but the protection scope of the present application is not limited thereto;

[0037] The technical solution of the present application is: a dual-path battery health state estimation method coupling a gas-liquid dynamics model and an equivalent circuit model, referring to Figure 1 , characterized in that:

[0038] Step one: constructing an electrical coupling model of the battery, the electrical coupling model comprising a gas-liquid dynamics model and an equivalent circuit model;

[0039] Step two: respectively identifying parameters of the gas-liquid dynamics model and the equivalent circuit model to obtain key parameters of the models;

[0040] Step three: based on the identified key parameters, constructing a joint estimation state equation to fuse and estimate the battery state;

[0041] Step four: proposing a battery health state estimation method based on the gas-liquid coupling model;

[0042] In the above scheme, the state equation of the gas-liquid dynamics model constructed in step one is formula one to formula two: In the formula, SOC q is the SOC estimated by the gas-liquid dynamics model at time t, U(t) is the terminal voltage of the gas-liquid dynamics model at time t, k1, k2, k3, k4 are to-be-identified parameters, T1 and T2 are the initial temperature and the temperature at the current sampling time, OCV is the open circuit voltage, I is the current, η is the coulomb efficiency, C0 is the rated capacitance, OCV1 is the initial open circuit voltage, || is the absolute value symbol, is the derivative of SOC with respect to time t;

[0043] The state equation of the equivalent circuit model constructed in step one is formula three to formula five: V(t) = OCV(SOC d (t)) + V1(t) + R0I(t) (Equation Four) d where SOC is the estimated SOC by the equivalent circuit model, V1 is the polarization voltage, R0 is the ohmic resistance, R1 and C1 are the polarization resistance and polarization capacitance, V(t) is the terminal voltage of the gas-liquid dynamics model at time t, is the derivative of V1 with respect to time t;

[0044] Five ternary batteries with a rated capacity of 29 Ah were selected and connected in series to form a 5-string battery module, Figure 2 are the SOC and OCV curves of the 5-string battery module at different aging stages, which will be used as parameter identification data; the parameter identification method of the equivalent circuit model based on the least squares method in step two is Equations Five to Seventeen; first, to facilitate online parameter identification, a discrete model of the battery is established according to Equation Five, and the voltage equation of Equation Five is reconstructed into Equation Six, which is a recursive form with a forgetting factor, containing three parameters α, β, and γ; then, Equation Seven defines the system output as the difference between the battery terminal voltage and the open-circuit voltage at the current SOC, and a standard linear regression model is constructed in Equation Eight; Equation Nine shows that the observation vector of the equivalent circuit model is composed of historical system output values, current and historical current values, and Equation Ten shows the parameters contained therein;

[0045] Figure 3 is the parameter identification result of R0, reflecting the change curve of the estimated R0; the core of parameter identification is to update the parameters in real time online using the recursive least squares algorithm: first, Equation Eleven calculates the a priori error, i.e., the deviation between the current measurement output and the predicted value based on the parameters at the last time; then, Equation Twelve calculates the gain vector, which weighs the weight in parameter updating; subsequently, Equation Thirteen updates the parameter estimation vector using the gain vector and the a priori error; finally, Equation Fourteen updates the covariance matrix to ensure the convergence and stability of the algorithm; after obtaining the updated linear parameters, Equation Ten is converted into the equivalent circuit model element parameters with explicit physical meaning through a preset parameter mapping relationship, including Equation Fifteen representing the polarization resistance, Equation Sixteen representing the ohmic internal resistance, and Equation Seventeen representing the polarization capacitance; V(k) = OCV(SOC d (k)) + V1(k) + R0I(k) (Equation Five) V(k) = αV(k-1) + βI(k) + γI(k-1) + (1-α)OCV(SOC d (k)) (Equation Six) y(k) = V(k) - OCV(SOC d (k)) (Equation Seven) y(k) = φ(k) T θ (Equation Eight) φ(k) = [y(k-1), I(k), I(k-1)] T (Equation Nine) θ = [α, β, γ] T (Equation Ten) e(k) = y(k) - φ(k) T θ(k-1) (Equation Eleven) θ(k) = θ(k-1) + K(k)e(k) (Equation Thirteen) In the formula, V(k) is the terminal voltage of the equivalent circuit at discrete point k, SOC d (k) is the state of charge of the equivalent circuit at discrete point k, I(k) is the current at discrete point k, V1(k) is the polarization voltage of the equivalent circuit at discrete point k; α is the autoregressive coefficient, taking the value 0.95; β is the current coefficient, taking the value 0.01; γ is the historical current coefficient, taking the value 0.001; y(k) is the system output, φ(k) is the observation vector, e(k) is the prior error, K(k) is the gain vector, θ(k) is the parameter estimation vector, P(k) is the covariance matrix, λ is the forgetting factor, taking the value 0.95; lnα is the natural logarithm of α;

[0046] The parameter identification method based on the gas-liquid dynamic model and the extended Kalman filter is Equations Eighteen to Twenty-two:

[0047] State vector: x q = [SOC q , k1, k2, k3, k4] T (Equation Eighteen) x q (k) = f(x q (k-1), I(k-1)) + ω(k-1) (Equation Nineteen) Wherein: In the formula, SOC q is the charge state estimated by the gas-liquid dynamics model; ω(k) is the process noise;

[0048] Observation equation: y q (k) = U(k) = h(x q (k), I(k)) + υ(k) (Equation Twenty-one) Wherein: where υ(k) is the observation noise;

[0049] ① Parameter initial state: x q_0 = [SOC q_0 , k 1_0 , k 2_0 , k 3_0 , k 4_0 ] T , Q q , R q , P q = [0.1, 0.001, 0.001, 0.01, 0.01] q_0 where x q is the model state vector including the battery state of charge and four key physical parameters, the subscript _0 of the letter denotes the initial value of the parameter; Q q is the covariance matrix of process noise, R q is the covariance matrix of observation noise, P q_0 is the initial error covariance matrix; according to the selected lithium-ion battery, the initial values can be provided as SOC 1_0 = 1, k 2_0 = 0.01, k 3_0 = 0.01, k 4_0 = 0.1, k q = 0.1; Q -4 = diag [10 -6 , 10 -6 , 10 -5 , 10 -5 ], diag represents a diagonal matrix; R q = 0.01; P q = [0.1, 0.001, 0.001, 0.01, 0.01] T

[0050] ② Time update:

[0051] ③ Measurement update:

[0052] ④ Calculate the Kalman gain:

[0053] ⑤ Update the state estimate:

[0054] ⑥ Update the error covariance

[0055] SOC of the gas-liquid dynamics model through extended Kalman iteration q k1, k2, k3, k4 provide important information for subsequent SOH estimation;

[0056] Step three, establish the state equation of the joint estimation of gas-liquid dynamics model and equivalent circuit model; x[t] = [SOC, k3, R0, R1] T (Equation thirty) SOC - (k) = A(k-1)SOC(k-1)A T (k-1) + Q (Equation thirty-two) K(k) = SOC - (k)C T (k) [C(k) SOC - (k) C T (k) + R] -1 (Equation thirty-three) SOC(k) = [I - K(k)C(k)] SOC - (k) (Equation thirty-five) In the formula, x[t] constructs a four-dimensional joint state vector integrating the battery core state and key model parameters, indicates the state prediction equation, K(k) indicates the Kalman gain K(k), and equation thirty-five updates the state where A(k) and C(k) are Jacobian matrices;

[0057] Step four, propose a battery health state estimation method based on the gas-liquid coupling model;

[0058] Based on the online identification of the ohmic resistance R0 of the equivalent circuit model, calculate the internal resistance method health state SOH R : In the formula, R 0,initial is the initial internal resistance of the battery, and R 0,initial = 0.0094 is given according to the selected battery; R 0,end is the internal resistance at the end of life, and R 0,end = 0.01181 is given; R0 is the ohmic resistance value obtained by online identification at the current time;

[0059] Based on the gas-liquid dynamics model parameters, calculate the capacity method health state through actual capacity estimation;

[0060] Figure 4 is the SOH estimation made by capacity method and internal resistance method respectively; SOH R and SOH c are fused by adaptive weight to get fuzzy SOH estimation; In the formula, λ R and λ C are adaptive weights, is the variance of internal resistance method SOH estimation, is the variance of capacity method SOH estimation, and ε is a constant, taking 10 -6 ;

[0061] In the above scheme, the battery SOH estimation is made based on the coupling of gas-liquid dynamics model and equivalent circuit model; it only depends on the open circuit voltage curve of the battery to complete the parameter identification of the two kinds of circuits respectively, depends on k3 and SOC of the gas-liquid dynamics model to complete the capacity method estimation, and depends on R0 of the equivalent circuit model to complete the battery internal resistance method estimation.

Claims

1. A dual-path battery state of health estimation method based on electrical coupling model, characterized in that, The method comprises the following steps: Step 1: constructing an electrical coupling model of the battery, wherein the electrical coupling model comprises a gas-liquid dynamics model and an equivalent circuit model; Step 2: performing parameter identification on the gas-liquid dynamics model and the equivalent circuit model respectively to obtain key parameters of the models; Step 3: based on the identified key parameters, constructing a joint estimation state equation to perform fusion estimation on the battery state; Step 4: proposing a battery health state estimation method based on the gas-liquid coupling model; In the above scheme, the state equation of the gas-liquid dynamics model constructed in step 1 is formula 1 to formula 2: In the formula, SOC q SOC(t) is the SOC estimated by the gas-liquid dynamics model at time t, U(t) is the terminal voltage of the gas-liquid dynamics model at time t, k1, k2, k3, and k4 are to-be-identified parameters, T1 and T2 are the initial temperature and the temperature at the current sampling time, OCV is the open-circuit voltage, I is the current, η is the coulomb efficiency, C0 is the rated capacitance, OCV1 is the initial open-circuit voltage, || is the absolute value symbol, is the derivative of SOC with respect to time t; The state equation of the equivalent circuit model constructed in step 1 is formula 3 to formula 4: V(t) = OCV(SOC d (t)) + Vi(t) + RoI(t) (Equation Four) where SOC d the SOC estimated by the equivalent circuit model, V1is the polarization voltage, R0is the ohmic resistance, R1and C1are the polarization resistance and the polarization capacitance, V(t) is the end voltage of the gas-liquid dynamics model at time t, is the derivative of V1with respect to time t; In step 2, the parameter identification method of the equivalent circuit model based on the least square method is formula 5 to formula 17; in order to facilitate online parameter identification, a discrete model of the battery is established according to formula 5, the voltage equation of formula 5 is reconfigured into formula 6, which is a recursive form with a forgetting factor, and contains three parameters α, β and γ; then, formula 7 defines the system output as the difference between the battery terminal voltage and the open-circuit voltage under the current SOC, and constructs a standard linear regression model formula 8, formula 9 indicates that the observation vector of the equivalent circuit model is composed of historical system output values, current and historical current values, and formula 10 shows that the parameters are contained; The core of parameter identification is to update the parameters in real time by using the recursive least square algorithm: firstly, formula 11 calculates the prior error, that is, the deviation between the current measurement output and the predicted value based on the parameters at the last moment; Then, formula 12 calculates the gain vector to balance the weight in parameter updating; Then, formula 13 updates the parameter estimation vector by using the gain vector and the prior error; finally, formula 14 updates the covariance matrix to ensure the convergence and stability of the algorithm; After obtaining the updated linear parameters, formula 10 is converted into the element parameters of the equivalent circuit model with clear physical meaning through a preset parameter mapping relationship, including formula 15 indicating the polarization resistance, formula 16 indicating the ohmic resistance and formula 17 indicating the polarization capacitance; V(k) = OCV(SOC) d (k))+V1(k)+R0I(k) (Formula 5) V(k) = aV(k - 1) + βl(k) + γl(k - 1) + (1 - a)OCV(SOC d (k)) (Equation Six) y(k) = V(k) - OCV(SOC d (k)) (Equation Seven) y(k) = φ(k) T θ (Equation Eight) φ(k) = [y(k - 1), I(k), I(k - 1)] T (Equation Nine) θ = [α, β, γ] T (Equation Ten) e(k) = y(k) - φ(k) T θ(k - 1) (Equation Eleven) θ(k)=θ(k-1)+K(k)e(k) (formula 13) where V(k) is the terminal voltage of the equivalent circuit at discrete point k, SOC d (k) is the state of charge of the equivalent circuit at discrete point k, I(k) is the current at discrete point k, V1(k) is the polarization voltage of the equivalent circuit at discrete point k; a is an autoregressive coefficient, β is a current coefficient, γ is a historical current coefficient; y(k) is a system output, φ(k) is an observation vector, e(k) is a prior error, K(k) is a gain vector, θ(k) is a parameter estimation vector, P(k) is a covariance matrix, λ is a forgetting factor; ln a is a natural logarithm of a; The parameter identification method based on the gas-liquid dynamics model and the extended Kalman filter is formula 18 to formula 22: State vector: x q = [SOC q , k1, k2, k3, k4] T (Equation Eighteen) x q (k) = f(x q (k-1), I(k-1)) + ω(k-1) (Equation 19) Wherein: where SOC q is the state-of-charge estimated by the gas-liquid dynamics model; ω(k) is the process noise; Observation equation: y q (k) = U(k) = h(x q (k), I(k)) + v(k) (Equation Twenty-One) Wherein: In the formula, υ(k) is the observation noise; ① Parameter initial state: x q_0 = [SOC q_0 , k 1_0 , k 2_0 , k 3_0 , k 4_0 ] T , Q q , R q , P q (Equation Twenty-three) where x q_0 As model state vector includes battery state of charge and four key physical parameters, the subscript _0 of the letter represents the initial value of the parameter; Q q is the covariance matrix of process noise, R q is the covariance matrix of observation noise, P q is the error covariance matrix; ② Time update: ③ Measurement update: ④ Calculate the Kalman gain: ⑤ Update the state estimation: ⑥ Update the error covariance SOC of the gas-liquid dynamics model is obtained through extended Kalman iteration q k1, k2, k3, k4 provide important information for subsequent SOH estimation; Step 3: establishing a state equation for joint estimation of the gas-liquid dynamics model and the equivalent circuit model; x[t] = [SOC, k3, R0, R1] T (Equation Thirty) SOC - (k) = A(k - 1)SOC(k - 1)A T (k - 1) + Q (Equation Thirty-Two) K(k) = SOC - (k)C T (k) [C(k) SOC - (k)C T (k) + R] -1 (Equation Thirty-Three) SOC(k) = [I - K(k)C(k)]SOC - (k) (Equation Thirty-Five) where x[t] constructs a four-dimensional joint state vector that integrates the battery core state and key model parameters, denotes the state prediction equation, K(k) denotes the Kalman gain K(k), and equation thirty-five performs the state update where A(k) and C(k) are Jacobian matrices; Step 4: proposing a battery health state estimation method based on the gas-liquid coupling model; Online identification of ohmic internal resistance R0 based on equivalent circuit model, internal resistance method state of health SOH R : In the formula, R 0,initial is the initial internal resistance of the battery, R 0,end is the internal resistance at the end of life, R0 is the ohmic internal resistance value obtained by online identification at the current time; Based on gas-liquid dynamics model parameters, the state of health (SOH) is calculated by actual capacity estimation c ; SOH is estimated by adaptive weights R and SOH c fusion, resulting in final SOH estimation; where λ R and λ C are adaptive weights, is the variance of the internal resistance method SOH estimate, is the variance of the capacity method SOH estimate, and ε is a constant.

2. The battery health state estimation method based on the electrical coupling model according to claim 1, wherein in step 2, the four to-be-identified parameters k1, k2, k3 and k4 of the gas-liquid dynamics model are identified by using the Kalman filter, and the R0 and R1 of the equivalent circuit model are identified by using the least square method.

3. The method of claim 1, wherein, The joint state vector is expressed as [SOC, k3, R0, R1] T where SOC is the state of charge estimated jointly, k3 is a kinetic parameter from the gas-liquid dynamics sub-model, R0 is the ohmic resistance, and R1 is the polarization resistance.

4. The electrical coupling-based state of health estimation method of claim 1, wherein The first health state index based on the internal resistance change and the second health state index based on the capacity change are fused by an adaptive weighting fusion module, the fusion weight is dynamically calculated according to the estimation uncertainty of the first health state index and the second health state index, and the two indexes are fused into a final health state estimation value.

5. A dual-path battery state-of-health estimation method implementing any one of claims 1-4 in an electrical coupling model, characterized by, In addition to the selected internal resistance and capacity as intuitive aging feature points, the identification parameters k values in the model can also be included, wherein k1 represents the linear current term coefficient in the terminal voltage in the model; k2 reflects the nonlinear current term coefficient in the terminal voltage; k3 reflects the lithium ion embedding and stripping capacity; and k4 reflects the lithium ion diffusion kinetic characteristics. The changes of k1 and k3 in the aging aspect can better reflect the battery health state.

Images