SOC and SOE joint estimation and prediction method of battery

Abstract

According to the SOC and SOE joint estimation and prediction method of the battery provided by the invention, model parameters are dynamically corrected by introducing aging factors, so that the adaptability and precision of the model in different aging stages are improved; through clear SOCSOE nonlinear mapping, joint estimation is allowed to be performed on the two variables, characterization of the battery state is enriched, the joint estimation enables an estimated value to more comprehensively reflect the battery energy condition, and more comprehensive reference is provided for an energy management strategy; for batteries with different power parameter completeness degrees, direct calculation, an empirical model or a correlation model can be flexibly adopted to carry out energy loss estimation, and the applicability in a complex application scene is improved; model errors and measurement noise are considered through an H-infinity filtering algorithm, so that SOC and SOE estimation and prediction still maintain high precision and robustness under uncertain conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of batteries, and in particular to a method for jointly estimating and predicting a battery's SOC (State Of Charge) and SOE (State of Energy). Background Art

[0002] At present, lithium batteries are widely used as energy storage devices in electric vehicles and energy storage systems. However, lithium batteries will experience aging phenomena such as capacity decay, increased internal resistance and increased polarization during long-term operation. In addition, the complexity of dynamic load conditions such as charging piles makes it challenging to accurately evaluate and predict SOC and SOE. Especially under actual operating conditions, it is inconvenient to directly obtain the measurement of SOC and SOE, and they need to be estimated indirectly through models and algorithms.

[0003] At present, most of them are aimed at battery SOC estimation, and there are fewer convenient solutions for SOE estimation, which cannot fully reflect the battery energy status. In related technologies, the SOC and SOE estimation of batteries are generally estimated separately using machine learning and data-driven methods (such as regression analysis, neural networks, etc.), which have strong adaptability and perform well for data under various working conditions. However, this method requires the use of a large amount of high-quality historical battery data for training, which not only requires high-performance computing resources, but also the generalization ability and robustness of the model depend on the diversity of the data set. It is impossible to make effective predictions for batteries that lack complete data, and the generalization ability and robustness are poor. Summary of the invention

[0004] The present invention is to solve the above technical problems, and the purpose of the present invention is to propose a method for jointly estimating and predicting the SOC and SOE of a battery.

[0005] The technical solution adopted by the present invention is as follows:

[0006] The embodiment of the present invention proposes a method for jointly estimating and predicting the SOC and SOE of a battery, comprising the following steps: establishing an improved equivalent circuit model according to model parameters and an aging factor α, wherein the model parameters include: a voltage source V oc 、Ohmic internal resistance R 0 , polarization resistance R 1 and polarization capacitance C 1, establish a SOC_SOE correlation model of the battery according to the SOC_SOE correlation parameters; obtain multiple groups of HPPC (Hybrid Pulse Power Characteristic) test data of the battery, identify model parameters according to the HPPC test data to generate an equivalent circuit model of the battery, obtain (SOC, SOE) data points covering the full SOC range of the battery, identify SOC_SOE correlation parameters according to the data points, and generate a SOC_SOE correlation model of the battery; establish an energy consumption model of the battery according to the completeness of the power parameters; on the basis of the model parameters and the SOC_SOE correlation parameters, construct a state vector including SOC, SOE, polarization voltage, ohmic internal resistance, and polarization resistance through an H∞ filtering algorithm, establish a discrete time state equation and an observation equation of the state vector according to the equivalent circuit model, the SOC_SOE correlation model and the energy consumption model, construct an H∞ filter according to the discrete time state equation and the observation equation, and use the filtering result of the H∞ filter to jointly estimate and predict the SOC and SOE of the current battery.

[0007] The battery SOC and SOE combined estimation and prediction method proposed in the present invention may also have the following additional technical features:

[0008] According to one embodiment of the present invention, the aging factor α is obtained in the following manner: Establish a battery aging model, where R 0 is the ohmic internal resistance of the battery, n is the number of cycles of the battery, and β are the aging coefficients obtained by fitting the experimental data, The ohmic internal resistance is R 0 The aging factor α corresponding to the battery; dynamically adjust R 0 The value of Among them, R 0,init is the initial internal resistance of the battery.

[0009] According to an embodiment of the present invention, identifying model parameters according to the HPPC test data specifically includes: using the HPPC test data to preliminarily estimate R 0 ,R 1 ,C 1 The initial value of R 0 ,R 1 ,C 1 Adjustments are made to optimize the terminal voltage predicted by the equivalent circuit model using a nonlinear least squares optimization method The terminal voltage V measured by HPPC test data t The error between them is used to obtain the optimal model parameter R 0 ,R1 ,C 1 ; The R corresponding to each aging stage was obtained by repeatedly performing nonlinear least squares fitting on the HPPC test data of different cycle numbers. 0 ,R 1 ,C 1 , the obtained parameter variation trend with the number of cycles N is regressed with the aging factor function to determine the aging factor α; according to R 0 ,R 1 ,C 1 and HPPC test data, calculate the voltage source V oc , according to the voltage source V oc Get V oc Corresponding curve with SOC.

[0010] According to an embodiment of the present invention, the SOC_SOE correlation parameter is identified according to the data point to generate the SOC_SOE correlation model of the battery, specifically including: according to J=∑(SOE i -(k 1 ·SOC i +k 2 ·SOC i 2 +k 3 ·SOC i 3 )) 2 Establish objective function J, SOE i is the energy state value of the ith data point, SOC i is the state of charge value of the ith data point, k 1 , k 2 and k 3 is the first association parameter, the second association parameter and the third association parameter; is k 1 , k 2 , k 3 Set a reasonable initial value and use an iterative optimization algorithm to minimize the objective function J. In each iteration, the algorithm calculates the gradient of the objective function relative to each parameter, adjusts the parameter value, and gradually reduces the objective function value until the preset convergence standard is reached and k is obtained. 1 , k 2 and k 3 The optimal solution for k 1 , k 2 and k 3 Substitute the optimal solution into SOE=k 1 ·SOC+k 2 ·SOC 2 +k 3 ·SOC 3 , generate the SOC_SOE correlation model of the battery.

[0011] According to one embodiment of the present invention, an energy consumption model of a battery is established according to the completeness of power parameters, specifically including: if the complete power parameters of the battery can be obtained in real time, the energy consumption and SOE prediction value of the battery are directly calculated using the known power parameters; if some power parameters of the battery are missing, the currently known partial data are matched with a typical operating condition library to find the operating condition type most similar to the current situation, obtain an empirical model corresponding to the operating condition type, estimate the energy consumption of the battery according to the empirical model, update the SOE distribution at future times according to the estimated energy consumption, and obtain the SOE prediction value; if the power parameters of the battery are completely missing, based on the known current SOC, the future energy consumption is approximately deduced according to previous operating condition data and the SOC_SOE association model to obtain the SOE prediction value.

[0012] According to one embodiment of the present invention, the state vector is constructed according to the following formula: X(k)=[SOC(k), SOE(k), U p (k),R 0 (k),R 1 (k),…] T , where X(k) is the state vector at time k, k is the discrete time step, and k = 0, 1, 2, ..., SOC(k) and SOE(k) are the state of charge and energy at time k, respectively, and U p (k) is the polarization voltage at time k, R 0 (k) is the ohmic internal resistance at time k, R 1 (k) is the polarization resistance at time k, and T is the transposed matrix.

[0013] According to one embodiment of the present invention, a discrete time observation equation of a state vector is established, specifically including: obtaining a theoretical value V of the terminal voltage according to the equivalent circuit model and the measured battery current I(k) t (k), according to the theoretical value of terminal voltage V t (k) and the measured value of the terminal voltage V t,meas (k) Obtain the input-to-output mapping matrix H(k); Establish a discrete-time observation equation according to Y(k)=C(k)X(k)+H(k)U(k)+V(k), where Y(k) is the observation vector, U(k) is the input vector, C(k) is the observation matrix, and V(k) is the measurement noise vector.

[0014] According to one embodiment of the present invention, a discrete-time state equation of a state vector is established, specifically comprising: obtaining the dynamic law of SOE changes with current, temperature and aging factor according to the energy consumption model and the equivalent circuit model, and obtaining the dynamic law of SOC changes with current, temperature and aging factor according to the energy consumption model and the SOC_SOE association model; establishing a state transfer matrix A(k) reflecting SOC and SOE changes with current, temperature and aging factor according to the dynamic law of SOE change and the dynamic law of SOC change; establishing a discrete-time state equation of the state vector according to X(k+1)=A(k)X(k)+G(k)W(k), wherein X(k+1) is the state vector at time (k+1), W(k) is the process disturbance vector, and G(k) is the noise input matrix.

[0015] According to an embodiment of the present invention, an H∞ filter is constructed according to a discrete-time state equation and an observation equation, specifically comprising: Define the state estimation error, is the estimated value of the state by the H∞ filter, according to Define the cost function of the H∞ filter, γ is the H∞ performance index γ; according to Construct an H∞ augmented system, where Z(k) is the H∞ performance output, L, M, G w To construct the matrix of auxiliary signal; Under the condition of known γ, the H∞ filter gain matrix K(k) is obtained by solving the augmented system; The H∞ performance index γ is adjusted to obtain the optimal filtering effect of the H∞ filter.

[0016] According to one embodiment of the present invention, the SOC and SOE of the current battery are jointly estimated and predicted using the filtering result of the H∞ filter, specifically including: at each time k, the measured Y(k) is input into the H∞ filter to obtain the estimated value of the state of the filter at the current time Thus, the current SOC and SOE are obtained; according to Repeat the iteration in the future time domain to obtain the estimated value at the future (k+j) time Thus, the estimated values ​​of SOE and SOC at the future (k+j) time are obtained, where j is a positive integer.

[0017] Beneficial effects of the present invention:

[0018] The present invention improves the adaptability and accuracy of the model at different aging stages by introducing aging factors to dynamically correct model parameters.

[0019] Through the clear SOC_SOE nonlinear mapping, the two variables can be jointly estimated, enriching the representation of the battery state. This joint estimation enables the estimated value to more comprehensively reflect the battery energy status and provide a more comprehensive reference for energy management strategies.

[0020] For charging pile load conditions with different data completeness, the present invention can flexibly use direct calculation, empirical model or correlation model to estimate energy loss, thereby improving the applicability in complex application scenarios;

[0021] The H∞ filtering algorithm takes into account model errors and measurement noise, so that SOC and SOE estimation and prediction can maintain high accuracy and robustness under uncertain conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of a method for jointly estimating and predicting SOC and SOE of a battery according to an embodiment of the present invention;

[0023] Figure 2 is a schematic diagram of an equivalent circuit model of a battery according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] Figure 1 : is a flow chart of a method for jointly estimating and predicting SOC and SOE of a battery according to an embodiment of the present invention. Figure 1 As shown, the battery SOC and SOE joint estimation and prediction method includes the following steps:

[0026] S1, based on the model parameters and the aging factor α, an improved equivalent circuit model is established. The model parameters include: voltage source V oc 、Ohmic internal resistance R 0 , polarization resistance R 1 and polarization capacitance C 1 , a SOC_SOE correlation model of the battery is established based on the SOC_SOE correlation parameters.

[0027] Specifically, the equivalent circuit model may be a Nernst equivalent circuit model, such as Figure 2 As shown, the model includes a voltage source V oc 、Ohmic internal resistance R 0 , polarization resistance R 1 and polarization capacitance C 1 And other key components.

[0028] Voltage source V ocRepresents the open circuit voltage of the circuit, and its value has a nonlinear corresponding relationship with the state of charge (SOC) of the battery; ohmic internal resistance R 0 Indicates the instantaneous response characteristics of the battery, the instantaneous change in voltage when the reaction current changes, R 0 It is used to describe the transient response characteristics of the battery, that is, the rapid change of the terminal voltage when the current changes. As the battery is used and the aging process accumulates, the internal microstructure changes and the loss of active materials will cause the ohmic internal resistance to gradually increase. In order to accurately reflect this change, the present invention introduces the aging factor α to R 0 Make dynamic corrections.

[0029] The aging factor α takes the cycle number n, usage time, operating temperature and discharge rate as inputs, and establishes the battery aging model according to the following formula (1):

[0030]

[0031] Among them, R 0 is the ohmic internal resistance of the battery, n is the number of cycles of the battery, and β are the aging coefficients obtained by fitting the experimental data, The ohmic internal resistance is R 0 The aging factor α corresponding to the battery.

[0032] Dynamically adjust R 0 The value of makes:

[0033]

[0034] Among them, R 0,init is the initial internal resistance of the battery.

[0035] RC parallel network (R 1 and C 1 ) is used to describe the polarization characteristics of a battery.

[0036] Terminal voltage V t is the actual measurable battery terminal voltage. Based on this equivalent circuit model, the battery terminal voltage V t The expression is:

[0037] V t =V oc -I·R 0 -U p (3)

[0038] Where, I is the load current (discharging is positive, charging is negative), U p is the polarization voltage. The change law of the polarization voltage Up follows the following first-order differential equation:

[0039]

[0040] Formulas (3) and (4) are combined with the aging factor α to form an improved equivalent circuit model.

[0041] Based on the Nernst equation, the present invention establishes a SOC_SOE correlation model of the mathematical relationship between the state of charge (SOC) and the state of energy (SOE), and the model expression is as follows:

[0042] SOE = k 1 ·SOC+k 2 ·SOC 2 +k 3 ·SOC 3 (5)

[0043] Among them, k 1 , k 2 , k 3 is the SOC_SOE associated parameter to be identified.

[0044] S2, obtain multiple groups of HPPC test data of the battery, identify model parameters and aging factor α according to the HPPC test data to generate an equivalent circuit model of the battery, obtain (SOC, SOE) data points covering the full SOC range of the battery, identify SOC_SOE correlation parameters according to the data points, and generate a SOC_SOE correlation model of the battery.

[0045] Specifically, in order to achieve accurate parameter identification of the equivalent circuit model and the SOC_SOE correlation model, a series of standardized experiments are required for the test battery. First, a series of targeted experiments are conducted to obtain the response data and energy characteristic data of the battery under different operating conditions. The experiments mainly include the following two categories:

[0046] (1) HPPC (Hybrid Pulse Power Characterization) test data: A specific pulse charge and discharge sequence is applied to the battery at different SOC levels, temperature conditions, and aging conditions, and the dynamic response data of terminal voltage, current, and temperature are recorded. By repeating the HPPC test in a new battery and in an aged state after multiple cycles (e.g., hundreds to thousands of times), a comprehensive data set spanning multiple aging stages can be obtained.

[0047] (2) SOC_SOE data acquisition: Under known conditions, the battery SOC is gradually changed (such as from full charge to low charge or vice versa), and the corresponding SOE value is measured and calculated at each SOC point. This process usually uses precise capacity measurement, energy metering and control methods to ensure that the acquired SOC_SOE data has high reliability for model fitting.

[0048] Through the above experiments, we can obtain: terminal voltage and current data at different aging stages, different SOCs and temperatures, which are used to identify the equivalent circuit model parameters and the aging factor α; several groups of (SOC, SOE) data points covering the entire SOC range are used to identify the SOC_SOE associated model parameters. The following introduces the identification method of model parameters, aging factor α and SOC_SOE associated parameters.

[0049] According to an embodiment of the present invention, the model parameters and the aging factor α are identified based on the HPPC test data, specifically including: using the HPPC test data to preliminarily estimate R 0 ,R 1 ,C 1 The initial value of R 0 ,R 1 ,C 1 Adjustments are made to optimize the terminal voltage predicted by the equivalent circuit model using a nonlinear least squares optimization method The terminal voltage V measured by HPPC test data t The error between them is used to obtain the optimal model parameter R 0 ,R 1 ,c 1 ; The R corresponding to each aging stage was obtained by repeatedly performing nonlinear least squares fitting on the HPPC test data of different cycle numbers. 0 ,R 1 ,c 1 , the obtained parameter variation trend with the number of cycles N is regressed with the aging factor function to determine the aging factor α; according to R 0 ,R 1 ,C 1 and HPPC test data, calculate the voltage source V oc , according to the voltage source V oc Get V oc Corresponding curve with SOC.

[0050] Specifically, the nonlinear least squares optimization method is to minimize the terminal voltage predicted by the equivalent circuit model The terminal voltage V measured by HPPC test data t The error between ECM , for example, the error J can be obtained according to the formula ECM :

[0051]

[0052] In one embodiment of the present invention, the SOC_SOE correlation parameter is identified according to the data point to generate the SOC_SOE correlation model of the battery, specifically including: according to J=Σ(SOE i -(k 1 ·SOCi +k 2 ·SOC i 2 +k 3 ·SOC i 3 )) 2 Establish objective function J, SOE i is the energy state value of the ith data point, SOC i is the state of charge value of the ith data point, k 1 , k 2 and k 3 is the first association parameter, the second association parameter and the third association parameter; is k 1 , k 2 , k 3 Set a reasonable initial value and use an iterative optimization algorithm to minimize the objective function J. In each iteration, the algorithm calculates the gradient of the objective function relative to each parameter, adjusts the parameter value, and gradually reduces the objective function value until the preset convergence standard is reached and k is obtained. 1 , k 2 and k 3 The optimal solution for k 1 , k 2 and k 3 Substitute the optimal solution into SOC=k 1 ·SOC+k 2 ·SOC 2 +k 3 ·SOC 3 , generate the SOC_SOE correlation model of the battery.

[0053] In actual operation, first k 1 , k 2 , k 3 Set reasonable initial values, which can be determined based on known characteristics of the battery or prior knowledge. Then, use an iterative optimization algorithm (such as gradient descent, Newton's method, etc.) to minimize the objective function. In each iteration, the algorithm calculates the gradient of the objective function relative to each parameter, adjusts the parameter value, and gradually reduces the objective function value until the preset convergence standard is reached and the optimal solution is obtained.

[0054] S3, establishing a battery energy consumption model according to the completeness of the power parameters.

[0055] In practical applications, batteries not only need to meet instantaneous power output, but also need to cope with long-term energy allocation strategies. Dynamic load conditions represented by charging piles have significant time-varying characteristics: power demand may change rapidly in a short period of time, and in a long period of time it manifests itself as periodic, staged or random characteristics. Based on the improved Nernst equivalent circuit model and SOC_SOE correlation model parameters obtained, the present invention establishes a multi-time scale energy consumption model, and formulates a hierarchical prediction strategy based on the differences in data integrity to meet the needs of different application scenarios.

[0056] To this end, according to an embodiment of the present invention, an energy consumption model of the battery is established according to the completeness of the power parameters, specifically including: if the complete power parameters of the battery can be obtained in real time, the energy consumption and SOE prediction value of the battery are directly calculated using the known power parameters; if some power parameters of the battery are missing, the currently known partial data are matched with the typical operating condition library to find the operating condition type most similar to the current situation, obtain the empirical model corresponding to the operating condition type, estimate the energy consumption of the battery according to the empirical model, update the SOE distribution at future times according to the estimated energy consumption, and obtain the SOE prediction value; if the power parameters of the battery are completely missing, based on the known current SOC, the future energy consumption is approximately deduced according to the previous operating condition data and the SOC_SOE association model to obtain the SOE prediction value.

[0057] Specifically, (1) when the system can obtain complete power parameters (including current, voltage, power demand curve, etc.) in real time, the energy consumption and SOE prediction value can be directly calculated using the known parameters. The energy consumption of the battery mainly includes: internal resistance energy loss and load energy consumption. Internal resistance energy loss calculation: In the equivalent circuit model, the ohmic internal resistance R 0 and polarization branch (R 1 ,C 1 ) can be calculated based on the instantaneous current I and real-time parameters. When considering the polarization effect, the instantaneous internal resistance power loss includes two parts: ohmic internal resistance loss and polarization loss. The total internal resistance power loss P loss (t) is:

[0058] P loss (t) = I 2 (t)R 0 +I(t)U p (t) (7)

[0059] To obtain the total energy loss within a certain time interval, the above power can be integrated over time.

[0060] Load energy consumption calculation: If the load places a constant or time-varying power demand P on the battery load(t), then according to the current terminal voltage V t The output power is obtained by summing the current I, and the power accumulated within a given prediction time range is the energy consumption. In other words, if the current SOC, SOE, internal resistance parameters and load power curve of the battery are known, the future SOE value can be estimated by predicting the energy consumption within a period of time. If it is assumed that the load power curve is known within the prediction time window, the SOE distribution at future moments can be updated based on the relationship between energy consumption and the original SOE to achieve energy state prediction.

[0061] (2) If some of the battery's power parameters are missing, match the currently known partial data with the typical operating condition library to find the operating condition type that is most similar to the current situation, obtain the empirical model corresponding to the operating condition type, estimate the battery's energy consumption based on the empirical model, update the SOE distribution at future times based on the estimated energy consumption, and obtain the SOE prediction value.

[0062] In actual operation, there may be situations where some power parameters are missing due to sensor failure, communication interruption or data delay. For example, the credibility of real-time current and voltage data decreases, or there is a lack of accurate prediction information on the future power demand of the load. In order to obtain relatively reliable energy state prediction in such cases, the present invention introduces a typical working condition matching and empirical model estimation mechanism. According to historical operation data in advance, the load characteristics are classified and clustered to form several "typical working condition" libraries. For example, the peak load of the charging pile during the day, the low load at night, and the special mode on weekends are summarized into several types of typical energy consumption curves through statistical analysis. When some power parameters are missing, the system matches the currently known partial data (such as the current SOC, the current current trend, the weather and time period information) with the typical working condition library to find the working condition type most similar to the current situation. After matching the corresponding working condition, the future energy consumption is estimated using the pre-established empirical model. The empirical model can use linear regression, polynomial fitting or an average power distribution diagram based on prior data. Although this method may be less accurate than the direct calculation method, it can still provide a certain degree of credibility in the data missing scenario. Energy consumption prediction.

[0063] After obtaining the estimated energy consumption data, substituting the estimated energy consumption and performing an approximate SOE prediction update can provide a trend reference for the battery management system and avoid falling into a blind spot when the data is incomplete.

[0064] (3) If the battery power parameters are completely missing, based on the current SOC, the future energy consumption is approximately deduced according to the previous operating condition data and the SOC_SOE association model to obtain the SOE prediction value.

[0065] In extreme cases, the load power parameters are completely missing, such as a complete communication interruption or a serious sensor failure. At this point, the present invention uses the following strategy to perform approximate energy consumption derivation and SOE prediction. Default operating condition assumption: In the absence of available power parameters, according to the default operating mode or preset rules of the device, it is assumed that the load power is a fixed value, linearly decreasing or maintaining a steady state. For example, if historical data statistics show that the charging pile load is extremely low or close to idle during the night time when no one is operating, the load can be defaulted to a minimum value state. On the basis of the known current SOE, the future energy consumption is approximately derived by reversely utilizing the SOC_SOE association model and previous operating condition data. Assuming that in the short term, the battery energy consumption mainly comes from internal resistance and tiny self-discharge (if self-discharge can be ignored, the internal resistance loss is the main one), the internal resistance loss is calculated based on the current current approximation (the assumption can be maintained from the data of the previous moment or a small value assumption can be used), and this is used as the basic energy consumption indicator. When there is no external power consumption data, the extremely small power assumption is adopted, so that the future SOE changes can be approximately dominated by internal resistance losses. This method can provide a minimum SOE trend reference in the absence of data, which helps to ensure that the battery management system does not completely lose its judgment of the battery status trend under abnormal circumstances.

[0066] Through the coping methods of the above three types of scenarios, the present invention has the ability to predict energy states under different data completeness. Whether it is short-term rapid response prediction (relying on real-time complete data) or medium- and long-term trend analysis (using typical working conditions and empirical models), or approximation and default strategies when data is completely missing, the multi-scenario, multi-time scale processing framework proposed by the present invention enables the prediction results to have better robustness and applicability in complex real environments.

[0067] S4, based on the model parameters and SOC_SOE associated parameters, a state vector including SOC, SOE, polarization voltage, ohmic internal resistance, and polarization resistance is constructed through the H∞ filtering algorithm. According to the equivalent circuit model, SOC_SOE associated model, and energy consumption model, the discrete-time state equation and observation equation of the state vector are established. According to the discrete-time state equation and observation equation, an H∞ filter is constructed. The filtering result of the H∞ filter is used to jointly estimate and predict the SOC and SOE of the current battery.

[0068] Specifically, based on the known model parameters and the SOC_SOE relationship, the state vector containing SOC, SOE, polarization voltage, and internal resistance parameters is constructed through the H∞ filtering algorithm. The state equation and observation equation are established using the improved Nernst equivalent circuit model. By adjusting the H∞ performance index γ, a set of robust gain matrices is obtained under the condition of considering the system model error and measurement noise, and the current SOC and SOE are jointly estimated with high robustness and high accuracy, and the future SOE and SOC are further predicted.

[0069] The following describes how to use the filtering result of the H∞ filter to jointly estimate and predict the SOC and SOE of the current battery in conjunction with a specific embodiment.

[0070] First, define the state vector containing SOC, SOE and other key parameters. Construct the state vector X(k) in units of discrete time step k (k = 0, 1, 2, ...):

[0071] X(k)=[SOC(k),SOE(k),U p (k),R 0 (k),R 1 (k),…] T (8)

[0072] Among them, SOC(k) and SOE(k) are the state of charge and energy state at time k, respectively. p (k) is the polarization voltage, R 0 (k) is the ohmic internal resistance at time k, R 1 (k) is the polarization resistance at time k (corrected by the aging factor), T is the transposed matrix, and the state vector can be expanded to include more parameters as needed.

[0073] According to the improved Nernst equivalent circuit model and SOC_SOE association model, the following discrete time state equation and observation equation can be established:

[0074] The observation equation is constructed based on the measured values ​​of terminal voltage and current. t The theoretical value of (k) is given by the Nernst equivalent circuit model:

[0075] V t (k) = V oc (SOC(k))-I(k)R 0 (k)-U p (k) (9)

[0076] The measured terminal voltage data is recorded as V t,meas (k), read by the sensor, the current I(k) is also a measurable input. According to the theoretical value of the terminal voltage V t (k) and the measured value of the terminal voltage V t,meas (k) Obtain the input-to-output mapping matrix H(k).

[0077] Taking into account measurement noise and uncertainty, the observation equation can be written as:

[0078] Y(k)=C(k)X(k)+H(k)U(k)+V(k) (10)

[0079] Among them, k(k) is the observation vector, including the measured terminal voltage and auxiliary data such as measurable current and temperature; C(k) is the observation matrix, which maps the state vector to the measurement quantity and describes the relationship between the model state and the actual measurement value; U(k) is the input vector (known environmental parameters, etc.); H(k) is the mapping matrix from input to output; V(k) is the measurement noise vector.

[0080] Typically, the terminal voltage is the main observed quantity, and the observation equation can be written as:

[0081] V t,meas (k) = V oc (SOC(k))-I(k)R 0 (k)-U p (k)+v(k) (11)

[0082] Where v(k) is the observation noise.

[0083] The state equation is based on the electrochemical model and the dynamic changes of parameters. After discretizing the continuous state model, it can be generalized as:

[0084] X(k+1)=A(k)X(k)+G(k)W(K) (12)

[0085] Among them, A(k) is the state transfer matrix, which reflects the dynamics of SOC, SOE and polarization parameters changing with current, temperature and aging factors. W(k) is the process disturbance vector, including model uncertainty and process noise, and G(k) is the noise input matrix.

[0086] Specifically, the evolution of SOE includes two parts, internal resistance energy loss and load energy consumption, which can be expressed as:

[0087] SOE(k+1)=SOE(k)-ΔE loss (k) (13)

[0088] Where, ΔE loss (k) is the energy consumption, which is determined by the internal resistance loss (including polarization effect) and the load power and can be obtained through the aforementioned energy consumption model. SOE(k+1) is obtained according to the energy consumption model. The dynamic law of SOE changing with current, temperature and aging factor is obtained according to the energy consumption model and the equivalent circuit model. The dynamic law of SOC changing with current, temperature and aging factor can be obtained according to the energy consumption model and the SOC_SOE association model. According to the dynamic law of SOE change and the dynamic law of SOC change, a state transfer matrix A(k) reflecting SOC and SOE changing with current, temperature and aging factor is established, and it is substituted into formula (13) to establish the discrete time state equation of the state vector.

[0089] H∞ filtering aims to find a filter in the presence of uncertainty (noise, model error) so that the transfer function gain from disturbance to estimation error is bounded by a certain H∞ performance index γ. In one embodiment of the present invention, an H∞ filter is constructed based on a discrete-time state equation and an observation equation, specifically including:

[0090] Define the state estimation error and the cost function of the H∞ filter:

[0091]

[0092] in, is the filter's estimate of the state.

[0093] H∞ filtering requires that under the most unfavorable disturbances W(k), V(k), the energy of the estimated error does not exceed a certain limit. That is, the cost function of the H∞ filter is that for a given γ>0,

[0094]

[0095] γ is the H∞ performance indicator.

[0096] Constructing an augmented system: To facilitate the solution of the H∞ filter gain, the system state, output and interference terms are combined to form an augmented system description. This augmented system can usually be expressed in the form of a standard H∞ control problem, which transforms the filter design into a process of solving a set of algebraic Riccati equations or linear matrix inequalities. The general form of the augmented system is:

[0097]

[0098] Where Z(k) is the H∞ performance output, L, M, G w To construct the matrix of auxiliary signal, the H∞ filtering problem can be transformed into standard form.

[0099] Under the condition of known γ, the H∞ filter gain matrix K(k) is obtained by solving the augmented system: Under the condition of known γ, the H∞ filter gain matrix K(k) can be obtained by solving a set of discrete time algebraic Riccati equations or using LMI optimization tools. The recursive form of the filter state estimation is:

[0100]

[0101] Adjust the H∞ performance index γ: In actual implementation, in order to adapt to nonlinear and time-varying characteristics, the system can be linearized at each sampling moment, or a time-varying H∞ filtering method can be used to update the gain K(k) online in real time. By selecting a smaller γ, the filter's ability to suppress interference can be improved, making the estimate more robust; but too small γ may cause the filter to be difficult to converge or increase the estimation deviation. Conversely, a larger γ improves the flexibility of the balance point between convergence and robustness. Adjust γ according to the actual operating conditions and the requirements for estimation accuracy and robustness to obtain the optimal filtering effect.

[0102] In one embodiment of the present invention, the SOC and SOE of the current battery are jointly estimated and predicted using the filtering result of the H∞ filter, specifically including:

[0103] After completing the H∞ filter gain design, the measured voltage, current and other known data are used to perform online iterative estimation of the state. At each time k, the measured Y(k) is input into the H∞ filter update equation to obtain the current time Thus we get and Since the filter is robust, it can maintain high estimation accuracy in the presence of modeling errors and measurement noise.

[0104] Using the identified parameter model and filtering results, the future load power P is predicted load (k+j) The change in internal resistance (determined by the aging factor) is used to extrapolate the future trend of SOE. The prediction process is as follows:

[0105]

[0106] Repeat the iteration in the future time domain to obtain the estimated value at the future (k+j) time Thus, the estimated values ​​of SOE and SOC at the future (k+j) time are obtained, where j is a positive integer.

[0107] In summary, the method for jointly estimating and predicting the SOC and SOE of a battery according to an embodiment of the present invention improves the adaptability and accuracy of the model at different aging stages by introducing an aging factor to dynamically correct the model parameters; through a clear SOC_SOE nonlinear mapping, it allows the two variables to be jointly estimated, enriching the characterization of the battery state. This joint estimation enables the estimated value to more comprehensively reflect the battery energy status and provide a more comprehensive reference for the energy management strategy; for charging pile load conditions with different data completeness, the present invention can flexibly use direct calculation, empirical models or correlation models to estimate energy loss, thereby improving its applicability in complex application scenarios; the H∞ filtering algorithm is used to consider model errors and measurement noise, so that SOC and SOE estimation and prediction still maintain high accuracy and robustness under uncertain conditions.

[0108] In the description of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily for the same embodiment or example. Any process or method description in a flow chart or otherwise described herein may be understood to represent a module, fragment or portion of a code including one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, in which functions may be performed in a non-sequential order as shown or discussed, including in a substantially simultaneous manner or in reverse order according to the functions involved, which should be understood by a technician in the technical field to which the embodiments of the present invention belong.

[0109] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0110] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for jointly estimating and predicting SOC and SOE of a battery, characterized in that: The following steps are involved: An improved equivalent circuit model is established based on the model parameters and the aging factor α. The model parameters include: voltage source V oc , ohmic internal resistance R0, polarization resistance R1 and polarization capacitance C1, and establish a SOC_SOE correlation model of the battery according to the SOC_SOE correlation parameters; Acquire multiple groups of HPPC test data of the battery, identify model parameters and aging factor α according to the HPPC test data to generate an equivalent circuit model of the battery, acquire (SOC, SOE) data points covering the full SOC range of the battery, identify SOC_SOE correlation parameters according to the data points, and generate a SOC_SOE correlation model of the battery; Establish the energy consumption model of the battery according to the completeness of the power parameters; Based on the model parameters and SOC_SOE associated parameters, a state vector including SOC, SOE, polarization voltage, ohmic internal resistance and polarization resistance is constructed by H∞ filtering algorithm. According to the equivalent circuit model, the SOC_SOE associated model and the energy consumption model, a discrete-time state equation and an observation equation of the state vector are established. According to the discrete-time state equation and the observation equation, an H∞ filter is constructed. The filtering result of the H∞ filter is used to jointly estimate and predict the SOC and SOE of the current battery.

2. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 1, characterized in that: Aging factor α is identified as follows: according to Establish a battery aging model, where R0 is the ohmic internal resistance of the battery, n is the number of battery cycles, and β are the aging coefficients obtained by fitting the experimental data, Indicates the aging factor α corresponding to the battery with an ohmic internal resistance of R0; Dynamically adjust the value of R0 so that Among them, R 0,init is the initial internal resistance of the battery.

3. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 1, characterized in that: Identifying model parameters and aging factor α according to the HPPC test data specifically includes: The initial values ​​of R0, R1, and C1 are preliminarily estimated using the HPPC test data, R0, R1, and C1 are adjusted, and the terminal voltage predicted by the equivalent circuit model is optimized using a nonlinear least squares optimization method. The terminal voltage V measured by HPPC test data t The error between them is calculated to obtain the optimal model parameters R0, R1, C1; By repeatedly performing nonlinear least squares fitting on the HPPC test data of different cycle numbers, the R0, R1, and C1 corresponding to each aging stage are obtained. The obtained parameter variation trend with the cycle number N is regressed with the aging factor function form to determine the aging factor α. Calculate the voltage source V based on R0, R1, C1 and HPPC test data oc , according to the voltage source V oc Get V oc Corresponding curve with SOC.

4. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 1, characterized in that: Identifying SOC_SOE correlation parameters according to the data points to generate a SOC_SOE correlation model of the battery, specifically including: According to J = ∑ (SOE i -(k1·SOC i +k2·SOC i 2 +k3·SOC i 3 )) 2 Establish objective function J, SOE i is the energy state value of the ith data point, SOC i is the state of charge value of the i-th data point, k1, k2 and k3 are the first associated parameter, the second associated parameter and the third associated parameter; Set reasonable initial values ​​for k1, k2, and k3, and use an iterative optimization algorithm to minimize the objective function J. In each iteration, the algorithm calculates the gradient of the objective function relative to each parameter, adjusts the parameter value, and gradually reduces the objective function value until the preset convergence standard is reached and the optimal solution of k1, k2, and k3 is obtained; Substitute the optimal solution of k1, k2 and k3 into SOE = k1·SOC + k2·SOC 2 +k3·SOC 3 , generate the SOC_SOE correlation model of the battery.

5. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 1, characterized in that: The energy consumption model of the battery is established according to the completeness of the power parameters, including: If the complete power parameters of the battery can be obtained in real time, the energy consumption and SOE prediction value of the battery can be directly calculated using the known power parameters; If some power parameters of the battery are missing, match the currently known partial data with the typical operating condition library to find the operating condition type most similar to the current situation, obtain the empirical model corresponding to the operating condition type, estimate the energy consumption of the battery according to the empirical model, update the SOE distribution at future times according to the estimated energy consumption, and obtain the SOE prediction value; If the battery power parameters are completely missing, based on the known current SOC, the future energy consumption is approximately deduced according to past operating data and the SOC_SOE association model to obtain the SOE prediction value.

6. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 1, characterized in that: The state vector is constructed according to the following formula: X(k)=[SOC(k),SOE(k),U p (k),R0(k),R1(k),…] T , Where X(k) is the state vector at time k, k is the discrete time step, and k = 0, 1, 2, ..., SOC(k) and SOE(k) are the state of charge and energy at time k, respectively, and U p (k) is the polarization voltage at time k, R0(k) is the ohmic internal resistance at time k, R1(k) is the polarization resistance at time k, and T is the transposed matrix.

7. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 6, characterized in that: Establish the discrete time observation equation of the state vector, including: According to the equivalent circuit model and the measured battery current I(k), the theoretical value of the terminal voltage V is obtained. t (k), according to the theoretical value of terminal voltage V t (k) and the measured value of the terminal voltage V t,meas (k) Obtain the input-to-output mapping matrix H(k); The discrete-time observation equation is established according to Y(k)=C(k)X(k)+H(k)U(k)+V(k), where Y(k) is the observation vector, U(k) is the input vector, C(k) is the observation matrix, and V(k) is the measurement noise vector.

8. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 7, characterized in that: Establish the discrete-time state equation of the state vector, including: According to the energy consumption model and the equivalent circuit model, the dynamic law of SOE changing with current, temperature and aging factor is obtained; according to the energy consumption model and the SOC_SOE association model, the dynamic law of SOC changing with current, temperature and aging factor is obtained; According to the dynamic law of SOE change and the dynamic law of SOC change, a state transfer matrix A(k) reflecting the SOC and SOE with current, temperature and aging factor is established; The discrete-time state equation of the state vector is established according to X(k+1)=A(k)X(k)+G(k)W(k), where X(k+1) is the state vector at time (k+1), W(k) is the process disturbance vector, and G(k) is the noise input matrix.

9. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 8, characterized in that: The H∞ filter is constructed based on the discrete time state equation and observation equation, which includes: according to Define the state estimation error, is the estimated value of the state by the H∞ filter, according to Define the cost function of the H∞ filter, γ is the H∞ performance index γ; according to Construct an H∞ augmented system, where Z(k) is the H∞ performance output, L, M, G w To construct the matrix of auxiliary signal; Under the condition of known γ, the H∞ filter gain matrix K(k) is obtained by solving the augmented system; Adjust the H∞ performance index γ to obtain the optimal filtering effect of the H∞ filter.

10. The method for jointly estimating and predicting the SOC and SOE of a battery according to claim 9, characterized in that: The filtering result of the H∞ filter is used to jointly estimate and predict the SOC and SOE of the current battery, including: At each time k, the measured Y(k) is input into the H∞ filter to obtain the estimated value of the state of the filter at the current time Thus, the current SOC and SOE are obtained; according to Repeat the iteration in the future time domain to obtain the estimated value at the future (k+j) time Thus, the estimated values ​​of SOE and SOC at the future (k+j) time are obtained, where j is a positive integer.

Images