Online parameter identification method for battery submodule of energy storage battery

Abstract

The present invention proposes an online parameter identification method for a battery submodule of an energy storage battery. Firstly, online parameter identification is performed on the entire battery submodule, real-time parameter identification is performed on the battery during an operation cycle, parameter identification of a switch device is added in a bypass state, and in the process of identifying the SOC, the capacity change is taken into account to perform real-time correction of the capacity; the on-line identification of the switch device of the battery submodule is completed by using the Kalman filter algorithm to estimate the on-voltage drop model of the IGBT; for the battery model parameter identification, a second-order RC model is used to estimate the battery parameters by recursive least squares RLS, and a SOC-OCV model is used to estimate the SOC value by an adaptive extended Kalman filter estimator AKEF.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage power stations, and in particular to an online parameter identification method for a battery submodule of an energy storage battery. Background Art

[0002] Energy storage technology is a revolutionary technology that can change the way energy and electricity are produced. It is also an important means to ensure the reliable power supply of the power system. Energy storage batteries are the most important component of energy storage power stations. Real-time status monitoring of energy storage battery groups is a key factor in the safe and stable operation of energy storage power stations. In view of this, it is very important to study the real-time monitoring technology of battery submodules in energy storage power stations.

[0003] The prior art has the following defects and deficiencies:

[0004] (1) Existing methods only identify the parameters of battery cells, without considering the impact of switching devices in the bypass state of the battery submodule.

[0005] (2) Existing methods analyze battery health from two perspectives: capacity and internal resistance. However, the capacity perspective method only considers the battery health during the discharge phase and ignores the battery health during the charging phase. The resistance perspective only considers the internal resistance and does not consider the capacity factor.

[0006] (3) Existing methods only use the rated capacity for calculation during the SOC identification process, and do not take into account the change in capacity as the operating time increases. Summary of the invention

[0007] In view of the defects and shortcomings of the prior art, the present invention proposes an online parameter identification method for a battery submodule of an energy storage battery, the design key points of which include:

[0008] (1) Perform online parameter identification on the entire battery submodule, perform real-time parameter identification on the battery during the operation cycle, and add parameter identification on the switching device in the bypass state.

[0009] (2) A SOH calculation method is proposed, which can reflect both the charging capacity and charging health status, as well as the discharging capacity and discharging health status. It also takes into account the influence of the battery charging capacity and internal resistance. This method does not require a known initial value and is also applicable to situations where the initial capacity is unknown.

[0010] (3) During the SOC identification process, the capacity change is taken into account and the capacity is corrected in real time.

[0011] The present invention specifically adopts the following technical solutions:

[0012] An online parameter identification method for a battery submodule of an energy storage battery is characterized in that: first, online parameter identification is performed on the entire battery submodule, real-time parameter identification is performed on the battery during the operation cycle, parameter identification of a switch device is added in the bypass state, and in the process of identifying the SOC, the capacity change is taken into account and the capacity is corrected in real time;

[0013] The on-state voltage drop model of IGBT is used to estimate the switching devices of the battery submodule using the Kalman filter algorithm. For the identification of battery model parameters, the second-order RC model is used to estimate the battery parameters using the recursive least squares RLS, and the SOC-OCV model is used to estimate the SOC value using the adaptive extended Kalman filter estimator AKEF.

[0014] Furthermore, the on-line identification of the battery submodule switch device is completed by using the Kalman filter algorithm using the IGBT on-voltage drop model, which specifically includes the following steps:

[0015] The on-state voltage drop of the IGBT is approximately expressed linearly as:

[0016] U T =U T0 +i T ·r T (1)

[0017] Among them, U T is the terminal voltage when the IGBT is turned on, with the switching device conduction direction as the voltage reference direction; i T is the IGBT collector current, U T0 is the on-state voltage bias, r T is the on-state resistance;

[0018] The on-state voltage bias, on-state resistance, terminal voltage and collector current when the IGBT is turned on are linearly related. T0 、r T Kalman filtering is used for parameter identification:

[0019] For a system, first use the following linear differential equation to describe it:

[0020]

[0021] Among them, X k , Z k , Q k , R k are the state quantity, quantity measurement, process noise and measurement noise of the system at time k; A is the transfer matrix of the state quantity from time k-1 to time k, and H is the measurement matrix, both of which are determined by the system itself;

[0022] Kalman filter algorithm for linear non-stationary process estimation Xk The steps are:

[0023] Step A1: Initialize parameters X0, P0, Q0, R0, where P is the system estimation error covariance matrix;

[0024] Step A2: Forecast error covariance matrix:

[0025] P[k] - =AP[k-1]A T +Q (3)

[0026] Step A3: Error update:

[0027] e[k]=Z[k]-H[k]X[k] - (4)

[0028] Step A4: Calculate the Kalman gain:

[0029] K[k]=P[k] - H[k] T (H[k]P[k] - H[k] T +R[k]) -1 (5)

[0030] Step A5: Update the system state at time k:

[0031] X[k] + =X[k] - +K[k]e[k] (6)

[0032] Step A6: Update the error covariance matrix at time k:

[0033] P[k] + =(IK[k]H[k])P[k] - (7)

[0034] Step A7: Repeat the above steps until the loop is exited;

[0035] When the IGBT is turned on, the following relationship is obtained:

[0036] U T =Ei T R (8)

[0037] Further we get:

[0038] E T R=U T0 +i T r T (9)

[0039] The Kalman filter state quantity, quantity measurement, and measurement array are expressed as:

[0040]

[0041] By continuously iterating equation (10) according to the Kalman filter algorithm, the state parameters can be continuously updated to achieve real-time estimation.

[0042] Furthermore, the second-order RC model is used to estimate the battery parameters by recursive least squares RLS as follows:

[0043] The second-order RC battery model includes: oc (SOC) represents the open circuit voltage OCV, the resistance R0, and two series-connected parallel RC networks, including: R1, C1 and R2, C2; the resistance R0 is the ohmic resistance caused by the accumulation and dissipation of charge in the double layer; R1 and C1 are the polarization resistance and capacitance respectively; R2 and C2 are the concentration polarization resistance and capacitance respectively; U1 and U2 are the polarization voltages on C1 and C2 respectively; I(t) is the load current, where discharge is positive and charge is negative; U t is the terminal voltage of the battery;

[0044] The discretized equation of the second-order RC battery model is expressed as:

[0045]

[0046] The mathematical relationship involving continuous time domain SOC is written as follows:

[0047]

[0048] Using the rectangular approximation of the integral and a suitably small sampling period, the discrete-time approximation of equation (12) can be recursively written as:

[0049]

[0050] The parameters of the second-order RC model, including R0, R1, R2, C1, and C2, are identified in real time using the recursive least squares estimation battery parameter algorithm. The steps are as follows:

[0051] Step B1: Given the initial value of the parameter estimate And its covariance matrix P[k0]:

[0052]

[0053] in, According to experience, P[k0] is taken as αI, where I represents the unit matrix; the size of the scalar α is changed to adjust the convergence speed of the identification parameters;

[0054] Step B2: Calculate the gain matrix:

[0055]

[0056] Step B3: Calculate the covariance matrix:

[0057]

[0058] Step B4: Calculate the estimated error:

[0059]

[0060] Step B5: Calculate the least squares estimate at time k:

[0061]

[0062] Furthermore, the SOC value is estimated by using the SOC-OCV model with the adaptive extended Kalman filter estimator AKEF as follows:

[0063] The SOC-OCV model is shown in Equation (14), which includes: a logarithmic function with real powers, a linear function, and an exponential function; the logarithmic function plays a dominant role in the low state of charge, within which charge accumulation occurs on the surface of the active material in the lithium-ion battery; the linear function dominates in the medium SOC range, within which primary phase transitions occur in the active material; the exponential function represents the high SOC behavior, that is, the occurrence of partial redox reactions and charge accumulation:

[0064] U OC (SOC(t)) = a + b × (-ln(SOC(t))) m +c×SOC(t)+d×exp n(SOC(t)-1) (14)

[0065] Adopting an adaptive extended Kalman filter estimator AKEF, which is updated online adaptively through a dedicated SOC estimator; the control input of the SOC estimator is the current representing the behavior of the battery during discharge or charge, and the output of the SOC estimator is the SOC value estimated by the AEKF algorithm;

[0066] The innovation is set to the difference between the voltage value estimated by the filter and the actual measured voltage value:

[0067]

[0068] Using the window estimation method, the innovation covariance matrix is ​​obtained:

[0069]

[0070] Based on the filter innovation covariance matrix H[k], an adaptive estimate of the covariance of the system and observation noise is obtained:

[0071]

[0072] The state space representation is:

[0073]

[0074] In formula (23), D d =R0;

[0075] The output equation is the same as equation (11); the SOC state vector is x[k] = [U1[k]U2[k]SOC[k]] T ; w[k] is the unmeasured process noise that affects the system state, v[k] is the measurement noise that does not affect the system state but is reflected in the system output estimate y[k]; Let Q[k] be Gaussian white noise with zero mean and covariance; Set to Gaussian white noise R[k] with zero mean and covariance; AEKF uses the filter's new information sequence to provide further new information, which allows the parameters Q and R to be estimated and updated iteratively;

[0076] The matrix C used in the AEKF filter is the derivative of the output equation with respect to the state vector before the estimation time, that is:

[0077]

[0078] in,

[0079] The battery capacity is calculated by the current predicted SOC and the previous SOC according to the discrete equation of the ampere-hour integration method. The calculation formula is:

[0080]

[0081] The adaptive extended Kalman filter algorithm and capacity correction estimate the SOC through a recursive iterative process. The specific calculation steps are as follows:

[0082] Input: u[k] = I[k], k∈{k0,...N-1}

[0083] Output:

[0084] data: R0[k],R1[k],C1[k],R2[k],C2[k]

[0085] Step C1: Initialize state variables and covariance matrix

[0086]

[0087] Step C2: Update state variables and error covariance

[0088]

[0089] Step C3: Error update

[0090]

[0091] Step C4: Adaptive process

[0092]

[0093] Step C5: Calculate Kalman gain

[0094] K[k]=P[k] - C[k] T (C[k]P[k] - C[k] T +R[k]) -1 (30)

[0095] Step C6: State variable update

[0096]

[0097] Step C7: Error covariance measurement update

[0098] P[k] + =(IK[k]C[k])P[k] - (32)

[0099] Step C8: Noise covariance update

[0100]

[0101] Step C9: Capacity Correction

[0102]

[0103] Step C10: Repeat the above steps until the loop is exited.

[0104] Further, the following steps are included:

[0105] Step 1: Initialize SOC, battery terminal voltage, battery terminal current and IGBT switch state;

[0106] Step 2: Determine whether the battery is bypassed. If yes, use the IGBT on-state voltage drop model and Kalman filter algorithm to estimate and complete the online identification of the battery submodule switch device;

[0107] If not, a second-order RC model is used to estimate the battery parameters using recursive least squares RLS, and a SOC-OCV model is used to estimate the SOC value using an adaptive extended Kalman filter estimator AKEF.

[0108] The present invention and its preferred embodiment disclose an online parameter identification method for a battery submodule of an energy storage battery. First, online parameter identification is performed on the entire battery submodule, real-time parameter identification is performed on the battery during the operation cycle, and parameter identification of the switching device is added in the bypass state. Then, a SOH calculation method is proposed, which can reflect both the charging capacity and the charging health status, as well as the discharging capacity and the discharging health status, while also considering the influence of the battery charging capacity and internal resistance. This method does not require a known initial value, and is also applicable to situations where the initial capacity is unknown. Finally, in the process of identifying the SOC, the capacity change is taken into account, and the capacity is corrected in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0110] Figure 1 Schematic diagram of an equivalent model of a battery submodule according to an embodiment of the present invention.

[0111] Figure 2 Schematic diagram of an IGBT parameter identification method according to an embodiment of the present invention.

[0112] Figure 3 Schematic diagram of a second-order RC model according to an embodiment of the present invention.

[0113] Figure 4 Schematic diagram of the SOC estimation method according to an embodiment of the present invention.

[0114] Figure 5 The figure is a schematic diagram of the method flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0115] Hereinafter, the specific embodiments of the present application will be described in detail with reference to the accompanying drawings. According to these detailed descriptions, those skilled in the art can clearly understand the present application and can implement the present application. Without violating the principles of the present application, the features in the various embodiments can be combined to obtain new implementations, or certain features in certain embodiments can be replaced to obtain other preferred implementations.

[0116] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail with the accompanying drawings as follows:

[0117] This embodiment provides an online parameter identification method for a battery submodule of an energy storage battery. First, online parameter identification is performed on the entire battery submodule, real-time parameter identification is performed on the battery during the operation cycle, and parameter identification of the switching device is added in the bypass state. Then, a SOH calculation method is proposed, which can reflect both the charging capacity and the charging health status, as well as the discharging capacity and the discharging health status, while also considering the influence of the battery charging capacity and internal resistance. This method does not require a known initial value, and is also applicable to situations where the initial capacity is unknown. Finally, in the process of identifying the SOC, the capacity change is taken into account, and the capacity is corrected in real time.

[0118] The specific design includes the following:

[0119] 1. Online identification of battery submodule switch devices

[0120] The battery submodule is the basic component unit of the battery cluster. The IGBT element is repeatedly turned on and off in a passive balancing state, and will age and fail under the repeated action of thermal shock. Its service life is directly related to the normal operation of the battery system.

[0121] The equivalent model of the battery submodule is as follows: Figure 1 As shown in the figure, when the IGBT gate is given an on signal, the battery will be in bypass state, and the battery output voltage is 0.

[0122] The on-state voltage drop of IGBT can be approximately expressed linearly:

[0123] U T =U T0 +i T ·r T (1)

[0124] Among them, U T is the terminal voltage when the IGBT is turned on, with the switching device conduction direction as the voltage reference direction; i T is the IGBT collector current, U T0 is the on-state voltage bias, r T is the on-state resistance.

[0125] The on-state voltage bias, on-state resistance, terminal voltage and collector current when the IGBT is turned on are linearly related. T0 、r T Kalman filtering can be used for parameter identification.

[0126] For a system, it should first be described by the following linear differential equation:

[0127]

[0128] Among them, X k , Z k , Q k , R k is the state quantity of the system at time k, the quantity measurement, process noise and measurement noise. A is the transfer matrix of the state quantity from time k-1 to time k, and H is the measurement matrix, both of which are determined by the system itself.

[0129] Kalman filter algorithm for linear non-stationary process estimation X k The steps are:

[0130] ① Initialize parameters X0, P0, Q0, R0, where P is the system estimation error covariance matrix.

[0131] ②Forecast error covariance matrix:

[0132] P[k] - =AP[k-1]A T +Q (3)

[0133] ③Error update:

[0134] e[k]=Z[k]-H[k]X[k] - (4)

[0135] ④Calculate Kalman gain:

[0136] K[k]=P[k] - H[k] T (H[k]P[k] - H[k] T +R[k]) -1 (5)

[0137] ⑤Update the system state at time k:

[0138] X[k] + =X[k] - +K[k]e[k] (6)

[0139] ⑥ Update the error covariance matrix at time k:

[0140] P[k] + =(IK[k]H[k])P[k] - (7)

[0141] ⑦ Repeat the above steps until the loop is exited.

[0142] When the IGBT is turned on, the following relationship can be obtained:

[0143] UT =Ei T R (8)

[0144] Further we get:

[0145] E T R=U T0 +i T r T (9)

[0146] The Kalman filter state quantity, quantity measurement, and measurement array can be expressed as:

[0147]

[0148] By iterating equation (10) according to the Kalman filter algorithm, the state parameters can be continuously updated to achieve real-time estimation. The IGBT parameter estimation method is as follows: Figure 2 shown.

[0149] 2. Battery model parameter identification

[0150] 2.1 Battery Second-Order Model

[0151] The second-order RC model has higher accuracy and is more suitable for parameter estimation of battery packs. Figure 3 As shown, the second-order RC battery model consists of U oc (SOC) represents the open circuit voltage (OCV), the resistor R0, and two parallel RC networks (R1, C1 and R2, C2) connected in series. The resistor R0 is the ohmic resistance caused by the accumulation and dissipation of charge in the double layer; R1 and C1 are the polarization resistance and capacitance respectively; R2 and C2 are the concentration polarization resistance and capacitance respectively; U1 and U2 are the polarization voltages on C1 and C2 respectively; I(t) is the load current (assuming discharge is positive and charge is negative); U t is the terminal voltage of the battery.

[0152] The discretized equation of the second-order RC battery model can be expressed as:

[0153]

[0154] The mathematical relationship involving continuous time domain SOC can be written as follows:

[0155]

[0156] Using the rectangular approximation of the integral and a suitably small sampling period, the discrete-time approximation recursion of equation (12) can be written as:

[0157]

[0158] 2.2 SOC-OCV model

[0159] An accurate SOC-OCV model can better identify the battery SOC. Currently, most methods use high-order polynomials for fitting, which ensures accuracy but does not have the physical and chemical significance of the battery. This embodiment uses a functional relationship that can both ensure fitting accuracy and reflect the actual significance of the battery in the selection of the SOC-OVC model.

[0160] The SOC-OCV model is shown in Equation (14), where logarithmic, linear, and exponential functions with real powers can be clearly seen. The logarithmic function plays a dominant role at low states of charge, within which charge accumulation on the surface of active materials in lithium-ion batteries occurs. The linear function dominates in the medium SOC range, within which primary phase transitions of active materials occur. The last exponential function contributes to high SOC behavior, where both partial redox reactions and charge accumulation occur.

[0161] U OC (SOC(t)) = a + b × (-ln(SOC(t))) m +c×SOC(t)+d×exp n(SOC(t)-1) (14)

[0162] 2.3 Online parameter identification based on recursive least squares (RLS)

[0163] The parameters in the second-order RC model mainly include R0, R1, R2, C1, and C2. The ohmic internal resistance and polarization internal resistance are affected by operating conditions such as SOC, temperature, and charge and discharge rate, and need to be identified in real time to improve accuracy.

[0164] The steps of the recursive least squares algorithm for estimating battery parameters are as follows:

[0165] ① Given the initial value of the parameter estimate and its covariance matrix P[k0]

[0166]

[0167] Generally speaking, An approximate value can be selected based on experience, and P[k0] can be taken as αI, where I represents the unit matrix. Changing the size of the scalar α can adjust the convergence speed of the identification parameters.

[0168] ②Calculate the gain matrix

[0169]

[0170] ③Calculate the covariance matrix

[0171]

[0172] ④Calculate the estimated error

[0173]

[0174] ⑤ Calculate the least squares estimate at time k

[0175]

[0176] 2.4 SOC estimation method based on capacity correction and adaptive extended Kalman filter

[0177] In order to better estimate the SOC, this embodiment uses an adaptive extended Kalman filter estimator (AKEF). The covariance parameter in the AEKF method is not a constant, but is updated online adaptively through a dedicated SOC estimator, thereby improving the estimation performance of the EKF. The control input of the SOC estimator is the current representing the behavior of the battery during discharge or charging, and the output of the SOC estimator is the SOC value estimated by the AEKF algorithm.

[0178] AKEF improves the SOC estimation accuracy by updating the covariance of state noise and observation noise in the system equation in real time with new information. The new information set is the difference between the voltage value estimated by the filter and the actual measured voltage value:

[0179]

[0180] Using the window estimation method, the innovation covariance matrix is ​​obtained:

[0181]

[0182] Based on the filter innovation covariance matrix H[k], an adaptive estimate of the covariance of the system and observation noise is obtained:

[0183]

[0184] The performance of the extended Kalman filter depends heavily on the accuracy of the pre-determined noise matrix. The state space representation is:

[0185]

[0186] In formula (23), D d =R0.

[0187] The output equation is the same as equation (11). The SOC state vector is x[k] = [U1[k]U2[k]SOC[k]] T w[k] is the unmeasured process noise that affects the system state, and v[k] is the measurement noise that does not affect the system state but can be reflected in the system output estimate y[k]. Assume Gaussian white noise Q[k] with zero mean and covariance; R[k] is assumed to be Gaussian white noise with zero mean and covariance. The AEKF uses the filter’s innovation sequence to provide further innovations that allow the parameters Q and R to be estimated and updated iteratively.

[0188] The matrix C used in the AEKF filter is the derivative of the output equation with respect to the state vector before the estimation time, that is:

[0189]

[0190] in,

[0191] During repeated charging and discharging, the capacity of the battery is constantly decreasing. Currently, the impact of battery capacity changes is not taken into account during battery parameter identification. Online correction of battery capacity for parameter identification can increase the accuracy of SOC estimation. The battery capacity can be calculated based on the discrete equation of the ampere-hour integration method by using the current predicted SOC and the previous SOC. The calculation formula is:

[0192]

[0193] The adaptive extended Kalman filter algorithm and capacity correction estimate the SOC through a recursive iterative process. The specific calculation steps are as follows:

[0194] Input: u[k] = I[k], k∈{k0,…N-1}

[0195] Output:

[0196] data: R0[k],R1[k],C1[k],R2[k],C2[k]

[0197] ① Initialize state variables and covariance matrix

[0198]

[0199] ② Update state variables and error covariance

[0200]

[0201] ③Error update

[0202]

[0203] ④ Adaptive process

[0204]

[0205] ⑤Calculate Kalman gain

[0206] K[k]=P[k] - C[k] T (C[k]P[k] - C[k] T +R[k]) -1 (30)

[0207] ⑥State variable update

[0208]

[0209] ⑦ Error covariance measurement update

[0210] P[k] + =(IK[k]C[k])P[k] - (32)

[0211] ⑧Noise covariance update

[0212]

[0213] ⑨Capacity correction

[0214]

[0215] ⑩Repeat the above steps until the loop exits.

[0216] The SOC estimation method based on data-driven adaptive extended Kalman filter is as follows: Figure 4 shown.

[0217] 2.5 Energy Storage Battery State of Health (SOH) Assessment

[0218] The health status of a battery SOH is defined as a comprehensive reflection of the charging health status and the discharging health status, that is, it can reflect not only the discharging health status, but also the charging health status:

[0219]

[0220] Among them SOH charge For charging health status, SOH discharge The discharge health status. Since the self-discharge of lithium-ion batteries can be ignored in a short period of time, the actual storage capacity at the end of charging and the actual storage capacity before the start of discharge can be considered equal, so they are unified as the actual storage capacity W nominal .W charge The input power of the battery during charging, W discharge The amount of electricity output from the battery output terminal during discharge.

[0221] From formula (35), we can get:

[0222]

[0223] In this way, SOH can reflect both the charging capacity and charging health as well as the discharging capacity and discharging health.

[0224]

[0225] Among them U c (t) is the battery terminal voltage at time t during charging, I c (t) is the charging current at time t.

[0226]

[0227] Among them U d (t) is the battery terminal voltage at time t during discharge, I d (t) is the discharge current at time t.

[0228]

[0229] Where U OC (SOC(t)) is the open circuit voltage of the battery at time t, and R0(t) is the internal resistance of the battery at time t. The parameters on the right side of the equal sign in the above health status estimation formula can be obtained through the data acquisition system. charge and W nominal The difference reflects the energy loss caused by the internal resistance of the battery during the charging process. nominal and W discharge The difference reflects the energy loss caused by the internal resistance of the battery during discharge.

[0230] like Figure 5 As shown, according to the above design, the method flow provided in this embodiment includes the following steps:

[0231] Step 1: Initialize SOC, battery terminal voltage, battery terminal current and IGBT switch state;

[0232] Step 2: Determine whether the battery is bypassed. If yes, use the IGBT on-state voltage drop model and Kalman filter algorithm to estimate and complete the online identification of the battery submodule switch device;

[0233] If not, a second-order RC model is used to estimate the battery parameters using recursive least squares RLS, and a SOC-OCV model is used to estimate the SOC value using an adaptive extended Kalman filter estimator AKEF.

[0234] The above is a preferred embodiment of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

[0235] This patent is not limited to the above-mentioned best implementation mode. Anyone can derive other forms of online parameter identification methods for energy storage battery submodules under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should be covered by this patent.

Claims

1. A method for online parameter identification of a battery submodule of an energy storage battery, characterized in that: First, the entire battery submodule is identified online, and the battery is identified in real time during the operation cycle. The switch device is identified in bypass mode, and the capacity change is taken into account during the SOC identification process to correct the capacity in real time. It uses the on-state voltage drop model of IGBT to estimate the switch devices of the battery submodule online using the Kalman filter algorithm. For the identification of battery model parameters, the second-order RC model is used to estimate the battery parameters using the recursive least squares RLS, and the SOC-OCV model is used to estimate the SOC value using the adaptive extended Kalman filter estimator AKEF. The following steps are involved: Step 1: Initialize SOC, battery terminal voltage, battery terminal current and IGBT switch state; Step 2: Determine whether the battery is bypassed. If yes, use the IGBT on-state voltage drop model and Kalman filter algorithm to estimate and complete the online identification of the battery submodule switch device; If not, the second-order RC model is used to estimate the battery parameters with recursive least squares RLS, and the SOC-OCV model is used to estimate the SOC value with adaptive extended Kalman filter estimator AKEF; Using the IGBT on-state voltage drop model and the Kalman filter algorithm to estimate the battery submodule switch device online identification specifically includes the following steps: The on-state voltage drop of the IGBT is approximately expressed linearly as: U T =U T0 +i T ·r T (1) Among them, U T is the terminal voltage when the IGBT is turned on, with the switching device conduction direction as the voltage reference direction; i T is the IGBT collector current, U T0 is the on-state voltage bias, r T is the on-state resistance; The on-state voltage bias, on-state resistance, terminal voltage and collector current when the IGBT is turned on are linearly related. T0 、r T Kalman filtering is used for parameter identification: For a system, first use the following linear differential equation to describe it: Among them, X k , Z k , Q k , R k are the state quantity, quantity measurement, process noise and measurement noise of the system at time k; A is the transfer matrix of the state quantity from time k-1 to time k, and H is the measurement matrix, both of which are determined by the system itself; Kalman filter algorithm for linear non-stationary process estimation X k The steps are: Step A1: Initialize parameters X0, P0, Q0, R0, where P is the system estimation error covariance matrix; Step A2: Forecast error covariance matrix: P[k] - =AP[k-1]A T +Q (3) Step A3: Error update: e[k]=Z[k]-H[k]X[k] - (4) Step A4: Calculate the Kalman gain: K[k]=P[k] - H[k] T (H[k]P[k] - H[k] T +R[k]) -1 (5) Step A5: Update the system state at time k: X[k] + =X[k] - +K[k]e[k] (6) Step A6: Update the error covariance matrix at time k: P[k] + =(I-K[k]H[k])P[k] - (7) Step A7: Repeat the above steps until the loop is exited; When the IGBT is turned on, the following relationship is obtained: U T =E-i T R (8) Further we get: E-i T R=U T0 +i T r T (9) The Kalman filter state quantity, quantity measurement, and measurement array are expressed as: By continuously iterating equation (10) according to the Kalman filter algorithm, the state parameters can be continuously updated to achieve real-time estimation; The second-order RC model is used to estimate the battery parameters by recursive least squares RLS as follows: The second-order RC model includes: oc (SOC) represents the open circuit voltage OCV, the resistance R0, and two series-connected parallel RC networks, including: R1, C1 and R2, C2; the resistance R0 is the ohmic resistance caused by the accumulation and dissipation of charge in the double layer; R1 and C1 are the polarization resistance and capacitance respectively; R2 and C2 are the concentration polarization resistance and capacitance respectively; U1 and U2 are the polarization voltages on C1 and C2 respectively; I(t) is the load current, where discharge is positive and charge is negative; U t is the terminal voltage of the battery; The discretized equation of the second-order RC battery model is expressed as: The mathematical relationship involving continuous time domain SOC is written as follows: Using the rectangular approximation of the integral and a suitably small sampling period, the discrete-time approximation of equation (12) can be recursively written as: The parameters of the second-order RC model, including R0, R1, R2, C1, and C2, are identified in real time using the recursive least squares estimation battery parameter algorithm. The steps are as follows: Step B1: Given the initial value of the parameter estimate And its covariance matrix P[k0]: in, According to experience, P[k0] is taken as αI, where I represents the unit matrix; the size of the scalar α is changed to adjust the convergence speed of the identification parameters; Step B2: Calculate the gain matrix: Step B3: Calculate the covariance matrix: Step B4: Calculate the estimated error: Step B5: Calculate the least squares estimate at time k: The SOC-OCV model is used to estimate the SOC value using the adaptive extended Kalman filter estimator AKEF as follows: The SOC-OCV model is shown in Equation (14), which includes: a logarithmic function with real powers, a linear function, and an exponential function; the logarithmic function plays a dominant role in the low state of charge, within which charge accumulation occurs on the surface of the active material in the lithium-ion battery; the linear function dominates in the medium SOC range, within which primary phase transitions occur in the active material; the exponential function represents the high SOC behavior, that is, the occurrence of partial redox reactions and charge accumulation: U OC (SOC(t))=a+b×(-ln(SOC(t))) m +c×SOC(t)+d×exp n(SOC(t)-1) (14) Adopting an adaptive extended Kalman filter estimator AKEF, which is updated online adaptively through a dedicated SOC estimator; the control input of the SOC estimator is the current representing the behavior of the battery during discharge or charge, and the output of the SOC estimator is the SOC value estimated by the AEKF algorithm; The innovation is set to the difference between the voltage value estimated by the filter and the actual measured voltage value: Using the window estimation method, the innovation covariance matrix is ​​obtained: Based on the filter innovation covariance matrix H[k], an adaptive estimate of the covariance of the system and observation noise is obtained: The state space representation is: In formula (23), D d =R0; The output equation is the same as equation (11); the SOC state vector is x[k] = [U1[k]U2[k]SOC[k]] T ; w[k] is the unmeasured process noise that affects the system state, v[k] is the measurement noise that does not affect the system state but is reflected in the system output estimate y[k]; Let Q[k] be Gaussian white noise with zero mean and covariance; Set to Gaussian white noise R[k] with zero mean and covariance; AEKF uses the filter's new information sequence to provide further new information, which allows the parameters Q and R to be estimated and updated iteratively; The matrix C used in the AEKF filter is the derivative of the output equation with respect to the state vector before the estimation time, that is: in, The battery capacity is calculated by the current predicted SOC and the previous SOC according to the discrete equation of the ampere-hour integration method. The calculation formula is: The adaptive extended Kalman filter algorithm and capacity correction estimate the SOC through a recursive iterative process. The specific calculation steps are as follows: Input: u[k] = I[k], k∈{k0,...N-1} Output: data: Step C1: Initialize state variables and covariance matrix Step C2: Update state variables and error covariance Step C3: Error update Step C4: Adaptive process Step C5: Calculate Kalman gain K[k]=P[k] - C[k] T (C[k]P[k] - C[k] T +R[k]) -1 (30) Step C6: State variable update Step C7: Error covariance measurement update P[k] + =(I-K[k]C[k])P[k] - (32) Step C8: Noise covariance update Step C9: Capacity Correction Step C10: Repeat the above steps until the loop is exited.

Images