An adaptive collaborative estimation method for SOC and SOH of lithium-ion batteries

Abstract

The present invention discloses a method for adaptive collaborative estimation of the SOC and SOH of a lithium-ion battery, which belongs to the technical field of lithium-ion battery state estimation. The steps are as follows: first, the terminal voltage and load current data of the lithium-ion battery are measured through a mixed power pulse characteristic experiment; then, a fractional-order equivalent circuit model of the lithium-ion battery is established offline; then, according to the state space of the established model, a dual adaptive square root volume Kalman filter for lithium-ion battery state estimation is constructed; finally, the actual working conditions are simulated through a random walk charge and discharge experiment, and the SOC and SOH of the lithium-ion battery are collaboratively estimated online. The present invention realizes high-precision SOC and SOH estimation while considering the influence of battery aging, solves the problem that the filter covariance matrix is ​​easy to lose semi-positive definiteness in collaborative estimation and the square root volume Kalman algorithm is not applicable, has strong robustness and generalization ability, and reduces the computational cost in a more mechanism-oriented way.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium-ion battery state estimation, and in particular to a method for adaptively co-estimating the SOC and SOH of a lithium-ion battery. Background Art

[0002] Currently, approaches to lithium-ion battery state estimation are primarily categorized into data-driven and model-driven methods. Data-driven methods are diverse and do not consider the battery's inherent operating principles and internal reaction mechanisms, resulting in a wide range of applications and strong learning capabilities for highly nonlinear data. However, these methods come at the cost of computational cost, requiring a large amount of training data. Furthermore, their generalization capabilities are constrained by the scope of the training data. Model-driven methods, on the other hand, construct an electrochemical model or equivalent circuit model of the lithium battery, establishing a nonlinear dynamic relationship between the battery state and observable signals, thereby constructing the battery's state space. Once the state space is constructed, these methods estimate the battery state using a state observer or filter. While limited to certain extent by model accuracy, these methods can reflect the battery's internal characteristics and degradation mechanisms, achieving accurate estimation results while maintaining robustness and adaptability to diverse and complex operating conditions.

[0003] During lithium-ion battery operation, battery model parameters gradually drift, affecting the reliability of the battery model and hindering battery state estimation, leading to cumulative errors. Traditional model-based methods estimate the state of charge (SOC) solely, ignoring the tightly coupled relationship between SOC, state of health (SOH), and model parameters. Consequently, collaborative estimation methods for lithium-ion battery state of charge (SOC) have attracted considerable attention in recent years. Collaborative estimation simultaneously estimates SOC while simultaneously updating the actual battery capacity and model parameter data to align with the current battery state, resulting in more reliable and accurate estimates. However, compared to intra-cycle SOC variations, battery model parameter drift is much slower and does not require recursive updates. Traditional collaborative estimation methods, which simultaneously update the battery state and model parameters, incur a significant computational burden. Furthermore, the model parameter covariance matrix tends to lose semi-positive definiteness with recursion, impacting the stability and accuracy of the estimation. Therefore, it is necessary to distinguish between different timescales for the estimation. Currently, manually specifying model parameter update conditions has been proven feasible. This approach achieves adaptive collaborative estimation by setting a threshold for the error between the predicted and observed values ​​or limiting the model parameter update step size, significantly reducing the computational cost of online estimation. However, the robustness of such estimation methods, which rely on manually defined adaptive update conditions, is limited. Furthermore, since battery model parameters do not need to be updated when the error is within a specified range, this adaptive mechanism can introduce certain estimation biases. Furthermore, to overcome the highly nonlinear state estimation of batteries, the aforementioned methods use either the extended Kalman function, which approximates linearity via Taylor decomposition, or the unscented Kalman function, which approximates the probability density using uniformly selected sigma points near the expected value. This results in a certain loss of estimation accuracy. Summary of the Invention

[0004] In response to the problems pointed out in the above background technology, the present invention provides a method for adaptively co-estimating the SOC and SOH of a lithium-ion battery in order to accurately and stably estimate the SOC and SOH of the lithium-ion battery. The technical solution adopted by the present invention includes the following steps:

[0005] Step 1: Measure the terminal voltage and load current data of the lithium-ion battery through a mixed power pulse characteristic experiment;

[0006] Step 2: Establish a fractional-order equivalent circuit model of the lithium-ion battery offline. The specific steps are as follows:

[0007] Step 2.1: According to Kirchhoff's voltage law, the input-output relationship equation of the lithium-ion battery system can be obtained:

[0008] U T (k)=OCV[SOC(k)]-U1(k)-U2(k)-R0I(k) (1)

[0009] Where U1 and U2 are the load voltages of fractional-order components C1 and C2, R0 and I are the ohmic internal resistance and load current, respectively. OCV[SOC(k)] is the OCV-SOC polynomial. SOC(k) is expressed using the ampere-hour integration method as:

[0010]

[0011] Among them C p is the actual capacity of the battery, η is the coulombic efficiency of the battery, T s is the sampling time;

[0012] Using the Grünwald-Letnikov fractional order discretization definition:

[0013]

[0014] in is a fractional-order operator, and α is the fractional order of the corresponding component;

[0015] Step 2.2: Establish the Kirchhoff current relationship equation for the lithium-ion battery model:

[0016]

[0017] Where R1 and R2 represent the electrochemical polarization resistance and concentration polarization resistance. According to the definition of formula (3), the discretization of formula (4) is as follows:

[0018]

[0019] In the formula, the fractional order differential is truncated using the short memory criterion, and the upper bound of the sum is set to 1;

[0020] Step 2.3: Let the load current I be the input and the terminal voltage U T For output, define the state vector x of the first filter state space = [SOC, U1, U2, R0, 1 / C p ] T , the state vector of the second filter state space is θ = [1 / R1, 1 / C1, α, 1 / R2, 1 / C2, β] T , n x and n θ are the dimensions of x and θ, u(k)=I(k), y(k)=U T (k), n d For the observation dimension, the discrete state space expression of the fractional-order model is established. The state equation and observation equation of the first filter are as follows:

[0021]

[0022] Where xi represents the i-th row element of x, w x and v represent the state noise and observation noise of the state vector x, respectively. They have zero mean and variance matrices Q x and R's uncorrelated white noise, the state transfer matrix and control matrix are shown as follows:

[0023]

[0024]

[0025] Where θ i represents the i-th row element of θ;

[0026] The state equation and observation equation of the second filter are as follows:

[0027]

[0028] Where w θ represents the state noise of the state vector θ, which has a mean of zero and a variance matrix of Q θ White noise that is uncorrelated with v;

[0029] Step 2.4: In the data measured in step 1, the terminal voltage at the end of each static moment in the entire experimental cycle is taken as a sampling point. The corresponding SOC of this point is obtained by formula (2). A total of 13 sampling points are taken. The curvefitting tool in Matlab is used to fit the OCV-SOC polynomial shown below:

[0030]

[0031] Step 3: Based on the state space of the model established in step 2, a dual adaptive square root volumetric Kalman filter is constructed for lithium-ion battery state estimation. The first filter estimates the SOC and the SOH represented by the ohmic internal resistance, and the second filter adaptively updates the lithium-ion battery model parameters according to the aging dormant zone. The specific steps are as follows:

[0032] Step 3.1: The first filter performs forward estimation, one-step prediction of the state and updates the covariance matrix:

[0033]

[0034] Where, qr represents QR decomposition, S(k|k-1) is the decomposed lower triangular matrix;

[0035] Step 3.2: The first filter performs forward estimation, estimates the volume point and propagates the volume point:

[0036]

[0037] Y i (k|k-1)=g(X i (k|k-1),u(k)) (14)

[0038] Where, ξ i is the volume point, where i = 1, 2, ..., n x , e i For [I x , -I x ] column i, I x n x Level unit array;

[0039] Step 3.3: The first filter performs forward estimation and calculates the measurement estimate:

[0040]

[0041] Step 3.4: The first filter performs forward estimation and constructs the weighted center matrices of the state and observation as the square root factors of the covariance matrix:

[0042]

[0043]

[0044] Step 3.5: The first filter performs forward estimation and calculates S by qr decomposition. xy , S yy and the square root of the updated covariance matrix:

[0045]

[0046] In the formula S(k|k) is the updated value of the square root of the covariance matrix;

[0047] Step 3.6: The first filter performs forward estimation, calculates the filter gain and updates the state x:

[0048]

[0049] Step 3.7: The first filter performs forward estimation and calculates SOH:

[0050]

[0051] Where R NEW is the initial ohmic resistance of the battery, R EOL is the ohmic resistance of the battery at the end of its life;

[0052] Step 3.8: Backward estimation is used to construct the aging dormant region and calculate the filter backward gain as shown in the following formula:

[0053]

[0054] Step 3.9: Backward estimation: Construct the aging dormant region and estimate the backward state vector:

[0055]

[0056] Step 3.10: Backward estimation is used to construct the aging dormancy region, and the threshold of the aging dormancy region is established, which increases with the aging of the lithium-ion battery:

[0057]

[0058] Step 3.11: Backward estimation constructs the aging dormant region, and the second filter works adaptively:

[0059]

[0060] When equation (23) does not hold, θ does not need to be updated. After equation (23) is completed, the next moment estimation is entered and θ maintains the current moment value. When equation (23) holds, the model parameter θ is no longer adapted to the current lithium-ion battery, so θ is estimated and updated after equation (23);

[0061] Step 3.12: The second filter estimates and updates the model parameters, predicts the model parameters in one step, and updates the covariance matrix:

[0062]

[0063] Where,

[0064] Step 3.13: The second filter estimates and updates the model parameters to estimate the volume points:

[0065]

[0066] Where, ξ j is the volume point, where j = 1, 2, ..., n θ , e j For [I θ , -I θ ]'s jth column, I θ n θ Level unit array;

[0067] Step 3.14: The second filter estimates and updates the model parameters to calculate the broadcast volume point:

[0068]

[0069] Different from the traditional method, in collaborative estimation, the model parameter θ as a variable in the state space driving matrix cannot directly obtain the value of the propagation volume point, but needs to be calculated by mapping θ to x;

[0070] Step 3.15: The second filter estimates and updates the model parameters, calculating the measurement estimate:

[0071]

[0072] Step 3.16, the second filter estimates and updates the model parameters, respectively constructing the weighted center matrix of the model parameters and the corresponding observations as the square root factor of the covariance matrix:

[0073]

[0074] Step 3.17, the second filter estimates the updated model parameters and calculates S by qr decomposition θy ,S θθ and the square root of the updated covariance matrix:

[0075]

[0076] Step 3.18, the second filter estimates and updates the model parameters, calculates the filter gain and updates the model parameters θ:

[0077]

[0078] Step 4: Simulate actual operating conditions through a random walk charge and discharge experiment, and use the filter constructed in step 3 to perform online collaborative estimation of the lithium-ion battery's SOC and SOH. The random walk charge and discharge experiment simulates actual operating conditions by randomly selecting one of the following excitation currents: -4.5A, -3.75A, -3A, -2.25A, -1.5A, -0.75A, 0.75A, 1.5A, 2.25A, 3A, 3.75A, or 4.5A. Negative current indicates charging, while positive current indicates discharging. The charge and discharge time is 5 minutes. After each charge or discharge cycle, there will be a rest period of less than 1 second. The cutoff voltage is 3.2V.

[0079] The beneficial effects of the present invention are:

[0080] 1. This invention takes into account the impact of battery aging on state estimation and implements adaptive SOC and SOH collaborative estimation by establishing an aging dormant zone. This not only improves the accuracy of SOC and SOH estimation, but also effectively reduces the computational cost of state estimation in a more systematic way.

[0081] 2. The present invention solves the problem that the square root volume Kalman filter is not suitable for battery state estimation by propagating volume points twice, ensuring the semi-positive definiteness of the filter covariance matrix in collaborative estimation, and improving the robustness and generalization ability of the method;

[0082] 3. This estimation algorithm can make the estimated value converge quickly to the correct value under certain initial deviations, effectively realize online estimation of SOC and SOH of lithium-ion batteries, and has engineering value. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 Schematic diagram of the overall process of the method of the present invention.

[0084] Figure 2 is the fractional-order equivalent circuit model of lithium-ion battery.

[0085] Figure 3 This is the SOC estimation result diagram of the 18650RW09 lithium-ion battery.

[0086] Figure 4 This is the terminal voltage prediction result diagram of 18650RW09 lithium-ion battery.

[0087] Figure 5 This is the SOH result diagram of 18650RW09 lithium-ion battery. DETAILED DESCRIPTION

[0088] An adaptive collaborative estimation method of SOC and SOH for lithium-ion batteries, such as Figure 1 As shown in the figure: The terminal voltage and load current data of the lithium-ion battery are measured by the hybrid power pulse characteristic experiment to obtain the corresponding lithium-ion battery data set; the fractional-order equivalent circuit model of the lithium-ion battery is established offline based on the obtained data set. The model is as follows: Figure 2 As shown in the figure; based on the state space of the established model, a dual adaptive square root volume Kalman filter for lithium-ion battery state estimation is constructed; the actual working conditions are simulated by random walk charge and discharge experiments, and the constructed filter is used to perform online collaborative estimation of the SOC and SOH of the lithium-ion battery, and the SOC and SOH estimation results are analyzed.

[0089] The following describes the specific embodiments of the present invention in detail with reference to the accompanying drawings and examples. Three lithium-ion batteries, 18650RW09, 18650RW10, and 18650RW11, are selected as specific examples.

[0090] Step 1: Measure the terminal voltage and load current data of the lithium-ion battery through a mixed power pulse characteristic experiment;

[0091] Step 2: Offline establish the fractional order equivalent circuit model of lithium-ion battery. Figure 2The specific steps are as follows:

[0092] Step 2.1: According to Kirchhoff's voltage law, the input-output relationship equation of the lithium-ion battery system can be obtained:

[0093] U T (k)=OCV[SOC(k)]-U1(k)-U2(k)-R0I(k) (1)

[0094] Where U1 and U2 are the load voltages of fractional-order components C1 and C2, R0 and I are the ohmic internal resistance and load current, respectively. OCV[SOC(k)] is the OCV-SOC polynomial. SOC(k) is expressed using the ampere-hour integration method as:

[0095]

[0096] Among them C p is the actual capacity of the battery, η is the coulombic efficiency of the battery, T s is the sampling time;

[0097] Using the Grünwald-Letnikov fractional order discretization definition:

[0098]

[0099] in is a fractional-order operator, and α is the fractional order of the corresponding component;

[0100] Step 2.2: Establish the Kirchhoff current relationship equation for the lithium-ion battery model:

[0101]

[0102] Where R1 and R2 represent the electrochemical polarization resistance and concentration polarization resistance. According to the definition of formula (3), the discretization of formula (4) is as follows:

[0103]

[0104] In the formula, the fractional order differential is truncated using the short memory criterion, and the upper bound of the sum is set to 1;

[0105] Step 2.3: Let the load current I be the input and the terminal voltage U T For output, define the state vector x of the first filter state space = [SOC, U1, U2, R0, 1 / C p ] T , the state vector of the second filter state space is θ = [1 / R1, 1 / C1, α, 1 / R2, 1 / C2, β] T , n x and n θare the dimensions of x and θ, u(k)=I(k), y(k)=U T (k), n d For the observation dimension, the discrete state space expression of the fractional-order model is established. The state equation and observation equation of the first filter are as follows:

[0106]

[0107] Where x i represents the i-th row element of x, w x and v represent the state noise and observation noise of the state vector x, respectively. They have zero mean and variance matrices Q x and R's uncorrelated white noise, the state transfer matrix and control matrix are shown as follows:

[0108]

[0109] Where θ i represents the i-th row element of θ;

[0110] The state equation and observation equation of the second filter are as follows:

[0111]

[0112] Where w θ represents the state noise of the state vector θ, which has a mean of zero and a variance matrix of Q θ White noise that is uncorrelated with v;

[0113] Step 2.4: In the data measured in step 1, the terminal voltage at the end of each static moment in the entire experimental cycle is taken as a sampling point. The corresponding SOC of this point is obtained by formula (2). A total of 13 sampling points are taken. The curvefitting tool in Matlab is used to fit the OCV-SOC polynomial shown below:

[0114]

[0115] When the order n is 10, the accuracy meets the fitting requirements. Then the genetic algorithm is used to determine the parameters of the fractional-order equivalent model offline. The obtained three battery model parameters are shown in Table 1;

[0116] Table 1 Parameters of fractional-order model for lithium-ion batteries

[0117]

[0118] Step 3: Based on the state space of the model established in step 2, a dual adaptive square root volumetric Kalman filter is constructed for lithium-ion battery state estimation. The first filter estimates the SOC and the SOH represented by the ohmic internal resistance, and the second filter adaptively updates the lithium-ion battery model parameters according to the aging dormant zone. The specific steps are as follows:

[0119] Step 3.1: The first filter performs forward estimation, one-step prediction of the state and updates the covariance matrix:

[0120]

[0121] Where, qr represents QR decomposition, S(k|k-1) is the decomposed lower triangular matrix;

[0122] Step 3.2: The first filter performs forward estimation, estimates the volume point and propagates the volume point:

[0123]

[0124] Y i (k|k-1)=g(X i (k|k-1),u(k)) (14)

[0125] Where, ξ i is the volume point, where i = 1, 2, ..., n x , e i For [I x , -I x ] column i, I x n x Level unit array;

[0126] Step 3.3: The first filter performs forward estimation and calculates the measurement estimate:

[0127]

[0128] Step 3.4: The first filter performs forward estimation and constructs the weighted center matrices of the state and observation as the square root factors of the covariance matrix:

[0129]

[0130] Step 3.5: The first filter performs forward estimation and calculates S by qr decomposition. xy , S yy and the square root of the updated covariance matrix:

[0131]

[0132] In the formula S(k|k) is the updated value of the square root of the covariance matrix;

[0133] Step 3.6: The first filter performs forward estimation, calculates the filter gain and updates the state x:

[0134]

[0135]

[0136] Step 3.7: The first filter performs forward estimation and calculates SOH:

[0137]

[0138] Where R NEW is the initial ohmic resistance of the battery, R EOL is the ohmic resistance of the battery at the end of its life;

[0139] Step 3.8: Backward estimation is used to construct the aging dormant region and calculate the filter backward gain as shown in the following formula:

[0140]

[0141] Step 3.9: Backward estimation constructs the aging dormant region and estimates the backward state vector:

[0142]

[0143] Step 3.10: Backward estimation is used to construct the aging dormancy region, and the threshold of the aging dormancy region is established, which increases with the aging of the lithium-ion battery:

[0144]

[0145] Step 3.11: Backward estimation constructs the aging dormant region, and the second filter works adaptively:

[0146]

[0147] When equation (23) does not hold, θ does not need to be updated. After equation (23) is completed, the next moment estimation is entered and θ maintains the current moment value. When equation (23) holds, the model parameter θ is no longer adapted to the current lithium-ion battery, so θ is estimated and updated after equation (23);

[0148] Step 3.12: The second filter estimates and updates the model parameters, predicts the model parameters in one step, and updates the covariance matrix:

[0149]

[0150] Where,

[0151] Step 3.13: The second filter estimates and updates the model parameters to estimate the volume points:

[0152]

[0153] Where, ξ j is the volume point, where j = 1, 2, ..., n θ , e j For [I θ , -I θ ]'s jth column, I θ n θ Level unit array;

[0154] Step 3.14: The second filter estimates and updates the model parameters to calculate the broadcast volume point:

[0155]

[0156] Different from the traditional method, in collaborative estimation, the model parameter θ as a variable in the state space driving matrix cannot directly obtain the value of the propagation volume point, but needs to be calculated by mapping θ to x;

[0157] Step 3.15: The second filter estimates and updates the model parameters, calculating the measurement estimate:

[0158]

[0159] Step 3.16, the second filter estimates and updates the model parameters, respectively constructing the weighted center matrix of the model parameters and the corresponding observations as the square root factor of the covariance matrix:

[0160]

[0161] Step 3.17, the second filter estimates the updated model parameters and calculates S by qr decomposition θy ,S θθ and the square root of the updated covariance matrix:

[0162]

[0163] Step 3.18, the second filter estimates and updates the model parameters, calculates the filter gain and updates the model parameters θ:

[0164]

[0165] Step 4: Simulate actual operating conditions through random walk charge and discharge experiments. Specifically, the excitation current is randomly selected from -4.5A, -3.75A, -3A, -2.25A, -1.5A, -0.75A, 0.75A, 1.5A, 2.25A, 3A, 3.75A, and 4.5A. Negative current indicates charging, and positive current indicates discharging. The charge and discharge time is 5 minutes. After each charge or discharge cycle, there is a rest period of less than 1 second. The cutoff voltage is 3.2V. The filter ADSRCKF constructed in Step 3 is used to perform online collaborative estimation of the lithium-ion battery's SOC and SOH. The SOC estimated by the coulomb counting method is used as the reference SOC value. The terminal voltage is predicted using its actual measured value as the reference value. The initial SOC value is set to 90%, with a deviation of 10% from the actual value to verify the robustness of the method. The estimation results of SOC and SOH are evaluated using RMSE and MAE as performance indicators. The statistical results are shown in Table 2. Taking 18650RW09 lithium-ion battery as an example, its SOC estimation results are shown in Table 2. Figure 3 As shown, the terminal voltage prediction results are as follows Figure 4 As shown, the SOH estimation results are as follows Figure 5 shown.

[0166] Table 2 Analysis of SOC and SOH estimation results

[0167]

[0168] The SOC estimation results obtained by the present invention show that the SOC estimation value is very close to the reference value, and it is almost impossible to distinguish between the two. The initial error of the SOC can be quickly restored to the correct value. The RMSE of the estimation results are 0.118%, 0.241% and 0.239%, respectively, and the MAE are 0.088%, 0.225% and 0.226%, respectively. This result proves that the collaborative estimation method of the present invention can ensure the reliability of the model in real time by making necessary updates to the battery model, and reliably correct the SOC through the closed-loop feedback of the filter based on the estimation error, ensuring the accuracy and robustness of the estimation to achieve accurate SOC estimation, which has engineering value. It can be clearly seen from the SOH estimation results that the change amplitude of SOH is very small throughout the charge and discharge cycle, and the change range is about 1E-5%, which is consistent with the change characteristics of the SOH of actual lithium-ion batteries and has real physical significance. Finally, the reliability of the SOC and SOH estimation was verified by predicting the terminal voltage. It is worth noting that during the estimation process, the prediction error of the terminal voltage mainly comes from the error of the lithium-ion battery model. Due to the limitation of the model accuracy, there is always a certain deviation between the predicted value and the actual measured value. The RMSE of the terminal voltage prediction results obtained by the present invention are 0.103V, 0.085V and 0.067V, respectively, and the MAE are 0.090V, 0.071V and 0.053V, respectively, which are very close to the actual measured values, verifying the effectiveness of the SOC and SOH estimation results obtained by the collaborative estimation method of the present invention.

[0169] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, obvious extensions or modifications made within the spirit of the present invention based on the technical solution of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for adaptive collaborative estimation of SOC and SOH of a lithium-ion battery, characterized in that: The steps include: Step 1: Measure the terminal voltage and load current data of the lithium-ion battery through a mixed power pulse characteristic experiment; Step 2: Offline identification establishes a fractional-order equivalent circuit model of the lithium-ion battery and establishes a state space; Step 3: Based on the state space of the model established in step 2, a dual adaptive square root volume Kalman filter for lithium-ion battery state estimation is constructed; Step 4: simulate the actual working conditions through random walk charge and discharge experiments, and use the filter constructed in step 3 to perform online collaborative estimation of the SOC and SOH of the lithium-ion battery; The offline establishment of the fractional-order equivalent circuit model of the lithium-ion battery in step 2 specifically includes the following steps: Step 2.1: According to Kirchhoff's voltage law, the input-output relationship equation of the lithium-ion battery system is obtained: U T (k)=OCV[SOC(k)]-U1(k)-U2(k)-R0I(k) (1); Where U1 and U2 are the load voltages of fractional-order components C1 and C2, R0 and I are the ohmic internal resistance and load current, respectively. OCV[SOC(k)] is the OCV-SOC polynomial. SOC(k) is expressed using the ampere-hour integration method as: Among them C p is the actual capacity of the battery, η is the coulombic efficiency of the battery, T s is the sampling time; Using the Grünwald-Letnikov fractional order discretization definition: in is a fractional-order operator, and α is the fractional order of the corresponding component; Step 2.2: Establish the Kirchhoff current relationship equation for the lithium-ion battery model: Where R1 and R2 represent the electrochemical polarization resistance and concentration polarization resistance. According to the definition of formula (3), the discretization of formula (4) is as follows: In the formula, the fractional order differential is truncated using the short memory criterion, and the upper bound of the sum is set to 1; Step 2.3: Let the load current I be the input and the terminal voltage U T For output, define the state vector x of the first filter state space = [SOC, U1, U2, R0, 1 / C p ] T , the state vector of the second filter state space is θ = [1 / R1, 1 / C1, α, 1 / R2, 1 / C2, β] T , n x and n θ are the dimensions of x and θ, u(k) = I(k), y(k) = U T (k), n d For the observation dimension, the discrete state space expression of the fractional-order model is established. The state equation and observation equation of the first filter are as follows: Where x i represents the i-th row element of x, w x and v represent the state noise and observation noise of the state vector x, respectively. They have zero mean and variance matrices Q x and R's uncorrelated white noise, the state transfer matrix and control matrix are shown as follows: Where θ i represents the i-th row element of θ; The state equation and observation equation of the second filter are as follows: Where w θ represents the state noise of the state vector θ, which has a mean of zero and a variance matrix of Q θ White noise that is uncorrelated with v; Step 2.4: In the data measured in step 1, the terminal voltage at the end of each static moment in the entire experimental cycle is taken as a sampling point. The corresponding SOC of this point is obtained by formula (2). A total of 13 sampling points are taken. The curve fitting toolbox in Matlab is used to fit the OCV-SOC polynomial shown below: The dual adaptive square root volume Kalman filter in step 3 estimates the SOC and the SOH represented by the ohmic internal resistance in the first filter, and adaptively updates the lithium-ion battery model parameters according to the aging dormant zone in the second filter. The specific steps are as follows: Step 3.1: The first filter performs forward estimation, one-step prediction of the state and updates the covariance matrix: Where, qr represents QR decomposition, S(k|k-1) is the decomposed lower triangular matrix; Step 3.2: The first filter performs forward estimation, estimates the volume point and propagates the volume point: Where, ξ i is the volume point, where i = 1, 2, ..., n x , e i For [I x , -I x ] column i, I x n x Level unit array; Step 3.3: The first filter performs forward estimation and calculates the measurement estimate: Step 3.4: The first filter performs forward estimation and constructs the weighted center matrices of the state and observation as the square root factors of the covariance matrix: Step 3.5: The first filter performs forward estimation and calculates S by qr decomposition. xy , S yy and the square root of the updated covariance matrix: In the formula S(k|k) is the updated value of the square root of the covariance matrix; Step 3.6: The first filter performs forward estimation, calculates the filter gain and updates the state x: Step 3.7: The first filter performs forward estimation and calculates SOH: Where R NEW is the initial ohmic resistance of the battery, R EOL is the ohmic resistance of the battery at the end of its life; Step 3.8: Backward estimation is used to construct the aging dormant region and calculate the filter backward gain as shown in the following formula: Step 3.9: Backward estimation: Construct the aging dormant region and estimate the backward state vector: Step 3.10: Backward estimation is used to construct the aging dormancy region, and the threshold of the aging dormancy region is established, which increases with the aging of the lithium-ion battery: Step 3.11: Backward estimation constructs the aging dormant region, and the second filter works adaptively: When equation (23) does not hold, θ does not need to be updated. After equation (23) is completed, the next moment estimation is entered and θ maintains the current moment value. When equation (23) holds, the model parameter θ is no longer adapted to the current lithium-ion battery, so θ is estimated and updated after equation (23). Step 3.12: The second filter estimates and updates the model parameters, predicts the model parameters in one step, and updates the covariance matrix: Where, Step 3.13: The second filter estimates and updates the model parameters to estimate the volume points: Where, ξ j is the volume point, where j = 1, 2, ..., n θ , e j For [I θ , -I θ ]'s jth column, I θ n θ Level unit array; Step 3.14: The second filter estimates and updates the model parameters to calculate the broadcast volume point: Step 3.15: The second filter estimates and updates the model parameters, calculating the measurement estimate: Step 3.16, the second filter estimates and updates the model parameters, respectively constructing the weighted center matrix of the model parameters and the corresponding observations as the square root factor of the covariance matrix: Step 3.17, the second filter estimates the updated model parameters and calculates S by qr decomposition θy ,S θθ and the square root of the updated covariance matrix: Step 3.18, the second filter estimates and updates the model parameters, calculates the filter gain and updates the model parameters θ:

2. The method for adaptive collaborative estimation of SOC and SOH of a lithium-ion battery according to claim 1, characterized in that: The random walk charge and discharge experiment in step 4 simulates the actual working condition, randomly selecting one of -4.5A, -3.75A, -3A, -2.25A, -1.5A, -0.75A, 0.75A, 1.5A, 2.25A, 3A, 3.75A, and 4.5 as the excitation current, the negative current is charging, the positive current is discharging, the charge and discharge time is 5 minutes, and there is a rest time of less than 1s after each charge or discharge cycle. The cut-off voltage is 3.2V.

Images