A joint estimation method for branch current, state of charge and power state of parallel battery pack

Abstract

The present invention discloses a method for jointly estimating the branch current, state of charge, and power state of a parallel battery pack. The method first obtains the branch current of the parallel battery pack and establishes an equivalent circuit model of the battery. The estimated branch current and terminal voltage are then used as inputs. Model parameters are identified using a recursive least squares method with a forgetting factor. An adaptive forgetting factor that varies with the residual is added to an extended Kalman filter to estimate the state of charge. The method also estimates the power state under the constraints of the state of charge and terminal voltage. The method takes into account the impact of inconsistencies between lithium-ion battery cells on branch current. The adaptive forgetting factor that varies with the residual also improves the adaptability of the extended Kalman filter to different environments, thereby improving the accuracy of branch current, state of charge, and power state estimation.

Description

Technical Field

[0001] The present invention relates to the field of battery technology, in particular to the technical field of electric vehicle power battery systems, and specifically to a method for jointly estimating branch current, state of charge, and power state of a parallel battery pack. Background Art

[0002] To meet energy and power requirements, electric vehicle battery systems typically consist of hundreds of battery cells connected in series or parallel. However, due to manufacturing tolerances and varying operating conditions, inconsistencies between battery cells are unavoidable. Inconsistencies in impedance and capacity between parallel cells can lead to uneven current distribution across each branch. Currently, for large parallel battery packs, battery management systems typically only monitor the total current and fail to obtain the current in each branch. If the average branch current, obtained by dividing the total current by the number of branches, is used as the actual branch current, cell inconsistencies are ignored, resulting in significant errors in the estimated state of charge (SOC) and state of power (SOP). Excessively low peak power can lead to insufficient battery energy supply, affecting vehicle performance; excessively high peak power can severely impact battery life. Therefore, accurate estimation of branch current, SOC, and SOP is particularly important for parallel battery packs.

[0003] At the same time, model-based SOC estimation mostly uses the Extended Kalman Filter (EKF). Due to the error between the equivalent circuit model and the actual physical model of the battery, as the number of filtering increases, the error accumulates, resulting in reduced accuracy or even divergence. It is necessary to add an appropriate forgetting factor to correct the state estimation error covariance. However, a constant forgetting factor cannot be adjusted as the system changes. Therefore, to better monitor and estimate the state of each cell in the parallel battery pack and improve battery life, it is necessary to jointly estimate the branch current, state of charge, and power state of the parallel battery pack. This is a technical problem that needs to be solved urgently. Summary of the Invention

[0004] In order to address the deficiencies in the prior art, the present invention provides a method for jointly estimating branch current, state of charge and power state of a parallel battery pack to solve the technical problems mentioned in the background technology.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack, comprising the following steps:

[0006] S1, obtain the voltage U of the two branches t1 、U t2 , and an estimated value of branch current I;

[0007] S2. Establish a first-order RC equivalent circuit model of the battery cell and fit the OCV-SOC curve based on the relationship between the open circuit voltage and state of charge of the battery;

[0008] S3, inputting the estimated branch current I and branch voltage obtained in S1 into the recursive least squares method with a forgetting factor to identify the parameters of the first-order RC equivalent circuit model;

[0009] S4, prior estimation of SOC and polarization voltage through EKF;

[0010] S5, perform a posteriori estimation of SOC and polarization voltage, add adaptive forgetting factor to update state estimation error covariance P in real time k ;

[0011] S6. Using the SOC estimated in S5, calculate the continuous charging peak current under SOC constraint and continuous discharge peak current

[0012] S7. Calculate the corresponding continuous discharge peak current under the terminal voltage constraint based on the equivalent circuit model parameters identified in S3. and continuous charging peak current

[0013] S8, based on the continuous charge and discharge peak current obtained in S6 and S7, and considering the maximum discharge current constraint value and minimum charge current constraint value of the power battery factory design, obtain the battery continuous charge peak current and continuous discharge peak current The resulting battery continuous charging peak current and continuous discharge peak current Substitute the first-order RC equivalent circuit model of the battery described in S3 to calculate the terminal voltage corresponding to the continuous peak current, thereby calculating the continuous charge and discharge peak power under multiple constraints.

[0014] Furthermore, the two branch voltages U recorded in step S1 t1 、U t2 , and a branch current estimate I are obtained by the following method:

[0015] S1.1 records the total current of the parallel battery pack under the three working conditions of FUDS, UDDS, and HPPC, the voltage of the two branches, the current of one branch, and the total current I under the DST working condition. L The voltage U of the two branches t,1 and U t,2 ;

[0016] S1.2 sets the parameters of the RBF neural network, integrates the total current under the three operating conditions of FUDS, UDDS, and HPPC, the voltages of the two branches, and the current of one branch, and normalizes the integrated data set to serve as the training set for training the RBF neural network;

[0017] S1.3, the total current of the test set I L , the voltage U of the two branches t1 、U t2 The input is input into the trained RBF neural network and denormalized to obtain a branch current I output by the RBF neural network.

[0018] Furthermore, the OCV-SOC curve fitting method described in S2 is as follows:

[0019] Perform OCV test to obtain the corresponding relationship between SOC and OCV, and fit it using formula (1):

[0020] U ocv (z) = α0 + α1z + α2z 2 +α3z 3 +α4 / z+α5ln(z)+α6ln(1-z) (1)

[0021] In formula (1), U ocv (z) represents the function of open circuit voltage OCV, which is a function of SOC; α0, α1, …, α6 represent the fitting coefficients of the OCV-SOC curve; z represents the SOC of the battery, calculated by the ampere-hour integration method, where the calculation formula of the ampere-hour integration method is formula (2)

[0022]

[0023] In formula (2), η represents the coulombic efficiency of the battery; C max The maximum available capacity in the current state.

[0024] Furthermore, the specific method for identifying the parameters of the first-order RC equivalent circuit model in S3 is as follows:

[0025] The covariance matrix at time k is Pk, and the initial value of the ohmic internal resistance is set to R ohm,0 , initial value of polarization internal resistance R p,0 , initial value of polarized capacitance C p,0 , initial value of open circuit voltage U OCV,0 , the initial value of the covariance matrix is P0, the parameter vector at time k and the data vector of the system are divided into formula (3) and formula (4):

[0026] θ k =[M k a1 a2 a3] T(3)

[0027]

[0028] In formula (3), M k =U OCV,k -a1U OCV,k-1 ;

[0029] The specific formula for parameter identification is formula (5):

[0030]

[0031] In formula (5), y is the output variable of the system; ω is the stationary zero-mean white noise; is the parameter θ k The estimated value of K k is the algorithm gain; λ is the forgetting factor, which ranges from (0,1]. When the forgetting factor is 1, the algorithm becomes a recursive least squares method without forgetting effect.

[0032] The model parameters at time k can be obtained as formula (6):

[0033]

[0034] Furthermore, in S5, the SOC at the kth sampling moment is set to z k , the polarization voltage is U p,k , the state vector is X k =[z k U p,k ] T , the state estimation error covariance is P X,k , set the initial values of the above parameters to z0, U p,0 、X0=[z0 U p,0 ] T , P0, system noise covariance is Q and observation error covariance is R;

[0035] When updating for the first time, the initial moment X0 is regarded as the posterior estimate of the previous moment Use formula (7) to make a priori estimation of EKF:

[0036]

[0037] In formula (7), f is the state update function, which can be written as formula (8) in discrete state:

[0038]

[0039] In formula (8), A is the state transfer matrix, at time k In discrete state

[0040]

[0041] The state estimation error covariance P0 at the initial moment is regarded as the The prior value of the state estimation error covariance is calculated using formula (9):

[0042]

[0043] According to the prior estimate of the state and the open circuit voltage value obtained from the SOC-OCV curve, the new information d is obtained by subtracting it from the measured value Y of the voltage sensor. Therefore, the new information d at time k is k The specific calculation formula is formula (10):

[0044]

[0045] In formula (10), h is the observation equation for calculating the open circuit voltage value. The observation equation in the discrete state is formula (11)

[0046]

[0047] Calculate the gain matrix K at time k X,k and through K X,k And the innovation matrix d k Calculate the posterior estimate of the state The specific calculation formula of the gain matrix and the state posterior estimate is formula (12):

[0048]

[0049] In formula (12), C X,k is the observation matrix,

[0050] Furthermore, the residual ε is calculated from the posterior estimates of SOC and polarization voltage in S5. k , which is formula (13):

[0051]

[0052] Variable forgetting factor s k The update formula is (14):

[0053]

[0054] σ k =exp[-floor(ρ·|ε k |)] (14)

[0055] In formula (14), σ kis the adjustment index of the adaptive forgetting factor; floor is not greater than ρ·|ε k Function of the largest integer |s min is the minimum value of the forgetting factor; ρ is a sensitive gain coefficient, controlling s k The rate tends to 1;

[0056] Adaptive forgetting factor Update, the specific update formula is formula (15):

[0057]

[0058] Furthermore, the SOC posterior estimate value estimated in step 6 is constrained, and L sampling periods are selected to perform continuous SOP estimation on the battery. The model parameters of the power battery within the L sampling periods are regarded as constants, and the current is regarded as a constant current. According to the ampere-hour integration method formula, the continuous charge and discharge peak current expression of the battery under the SOC constraint is derived as formula (16):

[0059]

[0060] In formula (16), and are the peak charge current and peak discharge current of the battery under SOC constraint; max 、z min The maximum and minimum SOC values allowed during battery charging and discharging are set to 90% and 10% respectively; k is the SOC value at time k estimated in step 6.

[0061] Furthermore, S7 specifically includes: the model parameter R at time k obtained by S3 ohm,k 、R p,k and C p,k The terminal voltage at the moment k+L is obtained as:

[0062] U t,k+L =U OCV,k+L -U p,k+L -I k+L R ohm,k

[0063] Then the open circuit voltage and polarization voltage of the battery at time k+L are:

[0064]

[0065]

[0066] Substitute the open circuit voltage and polarization voltage into the terminal voltage expression, then U t1 At the time k+L, it is formula (17):

[0067]

[0068] Based on formula (17), the continuous charge and discharge peak currents of the battery are:

[0069]

[0070]

[0071] In formula (18), and They represent the continuous discharge peak current and continuous charge peak current of L sampling cycles under voltage constraint; U max and U min Set the upper and lower cut-off voltages according to the specifications of the selected battery.

[0072] Furthermore, the continuous charge and discharge peak current under multiple constraints is calculated in S8, as shown in formula (19):

[0073]

[0074] In formula (19), I max and I min are the maximum discharge current and minimum charge current designed for the battery at the factory; combined with the battery terminal voltage, the continuous charge and discharge peak power is further obtained as formula (20):

[0075]

[0076] In formula (20), P max and P min These are the discharge power limit and charge power limit designed for the battery before leaving the factory.

[0077] Furthermore, the first-order RC equivalent circuit model described in S2 specifically adopts formula (21):

[0078]

[0079] In formula (21), the subscript k represents the kth sampling moment; Δt is the sampling period; R ohm represents the ohmic internal resistance; I represents the current; τ is the time constant and τ=R p C p , R p and C p are the polarization internal resistance and polarization capacitance of the battery respectively; U t1 is the terminal voltage; model parameter R ohm 、R p and C p Obtained by online identification using the recursive least squares method with forgetting factor; UOCV Indicates the battery open circuit voltage OCV.

[0080] Compared with the prior art, the beneficial effects of the present invention are:

[0081] 1. The present invention proposes a method for jointly estimating branch current, SOC and SOP of a parallel lithium-ion battery pack. Due to the inconsistency between lithium-ion battery cells, the RBF neural network can improve the estimation accuracy of branch current compared to the method of directly averaging the main current to calculate the branch current, thereby making the SOC and SOP with the branch current estimated by RBF as the input current more accurate.

[0082] 2. The present invention adds an adaptive forgetting factor that can change with the residual to the EKF. Compared with the forgetting factor that only changes with the number of filters, the present invention can adjust the forgetting factor value in real time as the system changes, and has better adaptability to different environments; the present invention adjusts the size of the forgetting factor through the residual, and can more promptly track the situation where the gain matrix tends to 0 due to the accumulation of model errors, and adjust the state estimation error covariance.

[0083] 3. Because the residual error in lithium-ion battery SOC estimation is of a small magnitude (far less than 1), squaring it requires adjusting the sensitive gain coefficient to a large value to achieve a significant change in the forgetting factor. The present invention converts the residual error to a positive value by taking its absolute value, significantly reducing the adjustment range of the sensitive gain coefficient. Therefore, the adaptive forgetting factor of the present invention, which can change with the residual error, has the advantages of high sensitivity to system changes, rapid response, good adaptability, and simple parameter adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Attachment Figure 1 It is a flowchart of the present invention;

[0085] Attachment Figure 2 Schematic diagram of a first-order RC equivalent circuit model used in the method of the present invention;

[0086] Attachment Figure 3 It is a schematic diagram of the RBF neural network structure adopted in the method of the present invention. DETAILED DESCRIPTION

[0087] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0088] The present invention provides a method for jointly estimating branch current, SOC and SOP of a parallel lithium-ion battery pack, comprising the following steps:

[0089] S1, obtain the voltage U of the two branches t1 、U t2 , and a branch current estimated value I. In this embodiment, the branch voltage and the branch current estimated value are obtained by:

[0090] The parallel battery pack experiment consisted of one G16 lithium-ion battery with a nominal capacity of 30.244Ah and one G31 lithium-ion battery with a nominal capacity of 29.927Ah. Current sensors were connected in series to the main line and one branch line, and voltage sensors were connected in parallel to both branches. The Arbin sampling time was fixed at 1s, and the dspace sampling time was fixed at 50ms. Under DST and FUDS operating conditions, both batteries had an initial SOC of 100%. Under UDDS operating conditions, the experiment was conducted after discharging 6.4Ah.

[0091] S1.1 records the total current of the parallel battery pack under the three working conditions of FUDS, UDDS, and HPPC, the voltage of the two branches, the current of one branch, and the total current I under the DST working condition. L The voltage U of the two branches t,1 and U t,2 ;

[0092] S1.2 sets the parameters of the RBF neural network, integrates the total current under the three operating conditions of FUDS, UDDS, and HPPC, the voltages of the two branches, and the current of one branch, and normalizes the integrated data set. This data set is used as the training set to train the RBF neural network. The RBF network has high approximation accuracy and can almost achieve complete approximation. It is also extremely convenient to design, and the network can automatically add neurons until the accuracy requirements are met. The RBF neural network is an efficient feedforward network with the best approximation performance and global optimal characteristics that other forward networks do not have. It also has a simple structure and fast training speed.

[0093] The structure of the RBF neural network is divided into three parts: input layer, hidden layer, and output layer. The number of input layer nodes is selected as 3, the number of hidden layer nodes is 10, and the number of output layer nodes is 1. The activation function between the input layer and the hidden layer is the abnormal S function, which is expressed as follows:

[0094]

[0095] Where, j represents the jth neuron in the hidden layer of the network; h j Represents the output of the Gaussian basis function; ||xc j || is the Euclidean distance; x represents the network input; b j represents the normalization constant of the jth hidden node; c jRepresents the center vector of the Gaussian function of the jth hidden node.

[0096] The activation function between the hidden layer and the output layer is a linear function, which is expressed as follows: O = Wh

[0097] Where O is the expected output value of the output neuron; h = [h j ] T , represents the output of Gaussian basis function; W=[w j ], representing the adjustment weight between the output layer neuron and the j-th neuron in the hidden layer.

[0098] The training algorithm is gradient descent, and the loss function is:

[0099]

[0100]

[0101] Where, e i is the error signal when the i-th sample is input; M is the number of neurons in the hidden layer, which is 2 here; p is the number of units in the input layer, which is 3 here.

[0102] The number of network iterations was set to 1000, the learning rate was set to 0.01, and the network training accuracy was set to 0.00002. Training ended when the sample error did not decrease after 20 consecutive iterations. The RBF neural network was trained using the main current dataset and two branch voltage datasets as input and the branch current dataset as output to obtain a trained RBF neural network.

[0103] S1.3, the total current of the test set (under DST conditions) I L , the voltage U of the two branches t1 、U t2 The input is input into the trained RBF neural network and denormalized to obtain a branch current I output by the RBF neural network.

[0104] S2. Establish a first-order RC equivalent circuit model of the battery cell and fit the OCV-SOC curve based on the relationship between the open circuit voltage and state of charge of the battery;

[0105] The first-order RC equivalent circuit model specifically adopts the following form:

[0106]

[0107] Where, subscript k represents the kth sampling moment; Δt is the sampling period; R ohm represents the ohmic internal resistance; I represents the current; τ is the time constant and τ=R p C p , R p and Cp are the polarization internal resistance and polarization capacitance of the battery respectively; U t1 is the terminal voltage; model parameter R ohm 、R p and C p Obtained by online identification using the recursive least squares method with forgetting factor; U OCV Indicates the battery open circuit voltage OCV, which can be obtained through the OCV-SOC curve;

[0108] The OCV-SOC curve fitting method is as follows:

[0109] Perform OCV test to obtain the corresponding relationship between SOC and OCV, and fit it using formula (1):

[0110] U ocv (z) = α0 + α1z + α2z 2 +α3z 3 +α4 / z+α5ln(z)+α6ln(1-z) (1)

[0111] In formula (1), U ocv (z) represents the function of open circuit voltage OCV, which is a function of SOC; α0, α1, …, α6 represent the fitting coefficients of the OCV-SOC curve; z represents the SOC of the battery, calculated by the ampere-hour integration method, where the calculation formula of the ampere-hour integration method is formula (2)

[0112]

[0113] In formula (2), η represents the coulombic efficiency of the battery; C max The maximum available capacity in the current state.

[0114] S3, the branch current I obtained in S1 and the voltage U of the battery cell t1 Input into the recursive least squares method with forgetting factor to identify the parameters of the first-order RC equivalent circuit model;

[0115] Set the initial value of ohmic internal resistance R0 to 0.002Ω and the initial value of polarization internal resistance R p,0 is 0.002Ω, the initial value of polarized capacitance C p,0 1.5×10 4 F, initial value of open circuit voltage U OCV,0 is 4.2V, the covariance matrix P k For diag[10 5 10 5 10 5 10 5 ], the parameter vector at time k and the data vector of the system are divided into formula (3) and formula (4):

[0116] θk =[M k a1 a2 a3] T (3)

[0117]

[0118] In formula (3), M k =U OCV,k -a1U OCV,k-1 ;

[0119] The specific formula for parameter identification is formula (5):

[0120]

[0121] In formula (5), y is the output variable of the system; ω is the stationary zero-mean white noise; is the parameter θ k The estimated value of K k is the algorithm gain; λ is the forgetting factor, which ranges from (0,1]. When the forgetting factor is 1, the algorithm becomes a recursive least squares method without forgetting effect.

[0122] The model parameters at time k can be obtained as formula (6):

[0123]

[0124] S4, prior estimation of SOC and polarization voltage through EKF;

[0125] Set the initial SOC value z0 = 1 and the initial polarization voltage value U p,0 = 0, then the initial state vector is X0 = [1 0] T , initialize the state estimation error covariance P X,0 =diag[0.00025 0.00025], system noise covariance Q=diag[0.00001 0.000025] and observation error covariance R=4, SOC estimation is completed through EKF, and the adaptive forgetting factor is added to update the state estimation error covariance in real time;

[0126] When updating for the first time, the initial moment X0 is regarded as the posterior estimate of the previous moment Use formula (7) to make a priori estimation of EKF:

[0127]

[0128] In formula (7), f is the state update function, which can be written as formula (8) in discrete state:

[0129]

[0130] In formula (8), A is the state transfer matrix, at time k In discrete state

[0131] The state estimation error covariance P0 at the initial moment is regarded as the The prior value of the state estimation error covariance is calculated using formula (9):

[0132]

[0133] According to the prior estimate of the state and the open circuit voltage value obtained from the SOC-OCV curve, the new information d is obtained by subtracting it from the measured value Y of the voltage sensor. Therefore, the new information d at time k is k The specific calculation formula is formula (10):

[0134]

[0135] In formula (10), h is the observation equation for calculating the open circuit voltage value. The observation equation in the discrete state is formula (11)

[0136]

[0137] Calculate the gain matrix K at time k X,k and through K X,k And the innovation matrix d k Calculate the posterior estimate of the state The specific calculation formula of the gain matrix and the state posterior estimate is formula (12):

[0138]

[0139] In formula (12), C X,k is the observation matrix,

[0140] S5, perform a posteriori estimation of SOC and polarization voltage, add adaptive forgetting factor to update state estimation error covariance P in real time k ; Calculate the residual ε based on the posterior estimate of the state k , which is formula (13):

[0141]

[0142] Variable forgetting factor s k The update formula is (14):

[0143]

[0144] In formula (14), σ kis the adjustment index of the adaptive forgetting factor; floor is not greater than ρ·|ε k Function of the largest integer |s min is the minimum value of the forgetting factor; ρ is a sensitive gain coefficient, controlling s k The rate tends to 1;

[0145] Adaptive forgetting factor Update, the specific update formula is formula (15):

[0146]

[0147] S6. Using the SOC estimated in S5, calculate the continuous charging peak current under SOC constraint and continuous discharge peak current The SOC posterior estimate value estimated in S5 is constrained, and 30 sampling periods are selected to estimate the battery's continuous SOP. The model parameters of the power battery within the 30 sampling periods are regarded as constants, and the current is regarded as a constant current. According to the ampere-hour integration method, the continuous charge and discharge peak current expression of the battery under the SOC constraint is derived as formula (16):

[0148]

[0149] In formula (16), and are the peak charge current and peak discharge current of the battery under SOC constraint; max 、z min The maximum and minimum SOC values allowed during battery charging and discharging are set to 90% and 10% respectively; k is the SOC value at time k estimated in step 6.

[0150] S7. Calculate the corresponding continuous discharge peak current under the terminal voltage constraint based on the equivalent circuit model parameters identified in S3. and continuous charging peak current

[0151] The model parameter R at time k obtained by S3 ohm,k 、R p,k and C p,k The terminal voltage at the moment k+L is obtained as:

[0152] U t,k+L =U OCV,k+L -U p,k+L -I k+L R ohm,k

[0153] Then the open circuit voltage and polarization voltage of the battery at time k+L are:

[0154]

[0155]

[0156] Substitute the open circuit voltage and polarization voltage into the terminal voltage expression, then U t1 At the time k+L, it is formula (17):

[0157]

[0158] Based on formula (17), the continuous charge and discharge peak currents of the battery are:

[0159]

[0160] In formula (18), and They represent the continuous discharge peak current and continuous charge peak current of L sampling cycles under voltage constraint; U max and U min Set the upper and lower cut-off voltages according to the specifications of the selected battery.

[0161] S8, based on the continuous charge and discharge peak current obtained in S6 and S7, and considering the maximum discharge current constraint value and minimum charge current constraint value of the power battery factory design, obtain the battery continuous charge peak current and continuous discharge peak current The resulting battery continuous charging peak current and continuous discharge peak current Substitute the first-order RC equivalent circuit model of the battery described in S3 to calculate the terminal voltage corresponding to the continuous peak current, and thus calculate the continuous charge and discharge peak power under multiple constraints, as shown in formula (19):

[0162]

[0163] In formula (19), I max and I min are the maximum discharge current and minimum charge current designed for the battery at the factory; combined with the battery terminal voltage, the continuous charge and discharge peak power is further obtained as formula (20):

[0164]

[0165] In formula (20), P max and P min These are the discharge power limit and charge power limit designed for the battery before leaving the factory.

Claims

1. A method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack, characterized by: The following steps are involved: S1, obtain the voltage U of the two branches t1 、U t2 , and an estimated value of branch current I; S2. Establish a first-order RC equivalent circuit model of the battery cell and fit the OCV-SOC curve based on the relationship between the open circuit voltage and state of charge of the battery; S3, inputting the estimated branch current I and branch voltage obtained in S1 into the recursive least squares method with a forgetting factor to identify the parameters of the first-order RC equivalent circuit model; S4, prior estimation of SOC and polarization voltage through EKF; S5, perform a posteriori estimation of SOC and polarization voltage, add adaptive forgetting factor to update state estimation error covariance P in real time k ; S6. Using the SOC estimated in S5, calculate the continuous charging peak current under SOC constraint and continuous discharge peak current S7. Calculate the corresponding continuous discharge peak current under the terminal voltage constraint based on the equivalent circuit model parameters identified in S3. and continuous charging peak current S8, based on the continuous charge and discharge peak current obtained in S6 and S7, and considering the maximum discharge current constraint value and minimum charge current constraint value of the power battery factory design, obtain the battery continuous charge peak current and continuous discharge peak current The resulting battery continuous charging peak current and continuous discharge peak current Substitute the first-order RC equivalent circuit model of the battery described in S3 to calculate the terminal voltage corresponding to the continuous peak current, thereby calculating the continuous charge and discharge peak power under multiple constraints; The OCV-SOC curve fitting method described in S2 is as follows: Perform OCV test to obtain the corresponding relationship between SOC and OCV, and fit it using formula (1): U ocv (z)=α0+α1z+α2z 2 +α3z 3 +α4 / z+α5 ln(z)+α6 ln(1-z) (1) In formula (1), U ocv (z) represents the function of open circuit voltage OCV, which is a function of SOC; α0, α1, …, α6 represent the fitting coefficients of the OCV-SOC curve; z represents the SOC of the battery, calculated by the ampere-hour integration method, where the calculation formula of the ampere-hour integration method is formula (2) In formula (2), η represents the coulombic efficiency of the battery; C max The maximum available capacity in the current state.

2. The method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack according to claim 1, characterized in that: The estimated branch current I in step S1 is obtained by the following method: S1.1 records the total current of the parallel battery pack under the three working conditions of FUDS, UDDS, and HPPC, the voltage of the two branches, the current of one branch, and the total current I under the DST working condition. L The voltage U of the two branches t,1 and U t,2 ; S1.2 sets the parameters of the RBF neural network, integrates the total current under the three operating conditions of FUDS, UDDS, and HPPC, the voltages of the two branches, and the current of one branch, and normalizes the integrated data set to serve as the training set for training the RBF neural network; S1.3, the total current of the test set I L , the voltage U of the two branches t1 、U t2 The input is input into the trained RBF neural network and denormalized to obtain a branch current I output by the RBF neural network.

3. The method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack according to claim 1, characterized in that: The specific method for identifying the parameters of the first-order RC equivalent circuit model in S3 is as follows: The covariance matrix at time k is Pk, and the initial value of the ohmic internal resistance is set to R ohm,0 , initial value of polarization internal resistance R p,0 , initial value of polarized capacitance C p,0 , initial value of open circuit voltage U OCV,0 , the initial value of the covariance matrix is P0, the parameter vector at time k and the data vector of the system are divided into formula (3) and formula (4): <h2 style=";text-align:left;direction:ltr">θ<h2 style=";text-align:left;direction:ltr"> k <h2 style=";text-align:left;direction:ltr"> =[M<h2 style=";text-align:left;direction:ltr"> k <h2 style=";text-align:left;direction:ltr"> a1 a2 a3<h2 style=";text-align:left;direction:ltr"> T <h2 style=";text-align:left;direction:ltr"> (3) In formula (3), M k =U OCV,k -a1U OCV,k-1 ; The specific formula for parameter identification is formula (5): In formula (5), y is the output variable of the system; ω is the stationary zero-mean white noise; is the parameter θ k The estimated value of K k is the algorithm gain; λ is the forgetting factor, which ranges from (0,1]. When the forgetting factor is 1, the algorithm becomes a recursive least squares method without forgetting effect. The model parameters at time k can be obtained as formula (6):

4. The method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack according to claim 1, characterized in that: In S4, the SOC at the kth sampling moment is set to z k , polarization voltage is U p,k , the state vector is X k =[z k U p,k ] T , the state estimation error covariance is P X,k , set the initial values of the above parameters to z0, U p,0 、X0=[z0 U p,0 ] T , P0, system noise covariance is Q and observation error covariance is R; When updating for the first time, the initial moment X0 is regarded as the posterior estimate of the previous moment Use formula (7) to make a priori estimation of EKF: In formula (7), f is the state update function, which can be written as formula (8) in discrete state: In formula (8), A is the state transfer matrix, at time k In discrete state The state estimation error covariance P0 at the initial moment is regarded as the The prior value of the state estimation error covariance is calculated using formula (9): According to the prior estimate of the state and the open circuit voltage value obtained from the SOC-OCV curve, the new information d is obtained by subtracting it from the measured value Y of the voltage sensor. Therefore, the new information d at time k is k The specific calculation formula is formula (10): In formula (10), h is the observation equation for calculating the open circuit voltage value. The observation equation in the discrete state is formula (11) Calculate the gain matrix K at time k X,k and through K X,k And the innovation matrix d k Calculate the posterior estimate of the state The specific calculation formula of the gain matrix and the state posterior estimate is formula (12): In formula (12), C X,k is the observation matrix, 5. The method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack according to claim 1, characterized in that: The residual ε is calculated from the posterior estimates of SOC and polarization voltage in S5 k , which is formula (13): Variable forgetting factor s k The update formula is (14): In formula (14), σ k is the adjustment index of the adaptive forgetting factor; floor is not greater than ρ·|ε k Function of the largest integer |s min is the minimum value of the forgetting factor; ρ is a sensitive gain coefficient, controlling s k The rate tends to 1; Adaptive forgetting factor Update, the specific update formula is formula (15):

6. The method for jointly estimating branch current, state of charge, and state of power of a parallel battery pack according to claim 1, characterized in that: The SOC posterior estimate value estimated in S5 is used as a constraint, and L sampling periods are selected to perform continuous SOP estimation on the battery. The model parameters of the power battery within L sampling periods are regarded as constants, and the current is regarded as a constant current. According to the ampere-hour integration method, the continuous charge and discharge peak current expression of the battery under the SOC constraint is derived as formula (16): In formula (16), and are the peak charge current and peak discharge current of the battery under SOC constraint; max 、z min are the maximum and minimum SOC values allowed when the battery is charging and discharging, respectively. k is the SOC value at time k estimated in step 6.

7. A method for jointly estimating branch current, state of charge and state of power of a parallel battery pack according to claim 1, characterized in that ,S7 specifically includes: The model parameter R at time k obtained by S3 ohm,k 、R p,k and C p,k The terminal voltage at the moment k+L is obtained as: IN t,k+L =U OCV,k+L -IN p,k+L -AND k+L R ohm,k Then the open circuit voltage and polarization voltage of the battery at time k+L are: Substitute the open circuit voltage and polarization voltage into the terminal voltage expression, then U t1 At the time k+L, it is formula (17): Based on formula (17), the continuous charge and discharge peak currents of the battery are: In formula (18), and They represent the continuous discharge peak current and continuous charge peak current of L sampling cycles under voltage constraint; U max and U min Set the upper and lower cut-off voltages according to the specifications of the selected battery.

8. A method for jointly estimating branch current, state of charge and power state of a parallel battery pack according to claim 1, characterized in that , S8 calculates the continuous charge and discharge peak current under multiple constraints, as shown in formula (19): In formula (19), I max and I min are the maximum discharge current and minimum charge current designed for the battery at the factory; combined with the battery terminal voltage, the continuous charge and discharge peak power is further obtained as formula (20): In formula (20), P max and P min These are the discharge power limit and charge power limit designed for the battery before leaving the factory.

9. A method for jointly estimating branch current, state of charge and state of power of a parallel battery pack according to claim 1, characterized in that ,The first-order RC equivalent circuit model described in S2 specifically adopts formula (21): In formula (21), the subscript k represents the kth sampling moment; Δt is the sampling period; R ohm represents the ohmic internal resistance; I represents the current; τ is the time constant and τ=R p C p , R p and C p are the polarization internal resistance and polarization capacitance of the battery respectively; U t1 is the terminal voltage; model parameter R ohm 、R p and C p Obtained by online identification using the recursive least squares method with forgetting factor; U OCV Indicates the battery open circuit voltage OCV.

Images