Power battery module health state estimation method based on multi-parameter coupling

Abstract

The invention provides a power battery module health state estimation method based on multi-parameter coupling. The method comprises the following steps: firstly, establishing a second-order RC equivalent model and a space state equation; obtaining a time sequence model; establishing an associated battery health state model of the equivalent parameters; training a battery equivalent parameter data generation model; constructing a power battery module data set, and establishing a battery module internal resistance, polarization resistance and capacitance multi-parameter equivalent model; then, establishing a mathematical model of current, voltage and capacity states of the battery module; constructing a battery module health state characterization model; normalizing time sequence data of voltage, current and capacity of the battery module so as to process the time sequence data into curve space vectors; finally, establishing a battery module health state estimation deep learning model and training the battery module health state estimation deep learning model; the evaluation model is established and evaluated, and the health state of the battery module is estimated based on the model meeting the requirement, so that the evaluation time is effectively shortened, the cost is reduced, and the screening efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health state estimation, and in particular to a method for estimating the health state of a power battery module based on multi-parameter coupling. Background Art

[0002] The use of new energy vehicles in my country is constantly increasing, and the content of energy storage devices in the power system is also increasing, but the treatment of a large number of retired batteries has become an increasingly prominent issue. The health status estimation of large-scale retired power batteries has the problems of long evaluation time, high cost and low screening efficiency. Reasonable evaluation of the health status of retired power battery modules is of great significance to ensure battery safety, improve resource utilization efficiency, reduce environmental pollution and reduce economic costs.

[0003] During the use of power battery modules, they will undergo long-term charge and discharge cycles and gradually age, leading to problems such as capacity decline, increased internal resistance, and decreased thermal management performance. If a health status assessment is not performed, retired batteries may still carry safety hazards, such as short circuits, overheating, leakage, and even fire or explosion hazards. Through health status assessment, potential safety risks of batteries, such as internal short circuits and thermal runaway of battery cells, can be discovered in a timely manner to prevent them from entering the secondary use or recycling link, ensuring the safety of the entire system. In addition, the health status assessment of retired battery modules can help determine whether the battery is suitable for secondary use. Although the power battery of an electric vehicle will decline during use, its remaining energy and capacity can still be used in some low-power or less stringent scenarios, such as home energy storage, commercial energy storage systems, backup power supplies, etc.

[0004] Based on the above premise of the necessity of estimating the health status of battery modules, there are two issues that need further research: ① Considering the large demand for battery data in deep learning model training, establish a battery module data generation model; ② In order to reduce the time of battery module health status assessment and improve efficiency, establish a deep learning model for battery module health status assessment. Therefore, based on the above two issues, we urgently need to develop a new method to estimate the health status of battery modules. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art or related art.

[0006] To this end, the purpose of the present invention is to propose a method for estimating the health status of a power battery module based on multi-parameter coupling.

[0007] In order to achieve the above-mentioned purpose, the technical solution of the present invention provides a method for estimating the health status of a power battery module based on multi-parameter coupling. The method for estimating the health status of a power battery module based on multi-parameter coupling includes: step S1: establishing a second-order RC equivalent model of the power battery module and a spatial state equation of a second-order RC equivalent model of a power battery cell; step S2: performing differential discretization on the spatial state equation established in step S1 to obtain the timing models corresponding to the three parameters of battery voltage, battery state of charge, and battery capacity; step S3: based on the timing model obtained in step S2, establishing a battery health status model associated with the internal resistance, polarization resistance, and capacitance equivalent parameters of the battery equivalent model; step S4: based on the associated battery health status model established in step S3 Healthy state model, establish a battery equivalent parameter data generation model, and then train the battery equivalent parameter data generation model based on the battery equivalent parameter set, and generate battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters; step S5: randomly sample the battery monomers that generate equivalent parameters to construct a power battery module data set, and establish a multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance; step S6: based on the second-order RC equivalent model established in step S1 and the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance established in step S5, establish the battery module current, voltage and capacity Step S7: construct a battery module health status characterization model based on the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance established in step S5, and the battery module current, voltage and capacity status mathematical model established in step S6; Step S8: based on the battery module current, voltage and capacity status mathematical model established in step S6, and the timing model obtained in step S2, obtain the voltage, current and capacity timing data of the battery module under dynamic working conditions through simulation, and normalize the obtained voltage, current and capacity timing data of the battery module to obtain the voltage, current and capacity timing data of the battery module. The voltage, current and capacity time series data are processed into curve space vectors; step S9: based on the characteristics of the curve space vectors in step S8, a deep learning model for estimating the health status of the battery module is established, and then the deep learning model for estimating the health status of the battery module is trained through the curve space vectors of the voltage, current and capacity of the battery module and the battery module health status characterization model in step S7; step S10: establishing an evaluation model, and evaluating the deep learning model established in step S9 to obtain a deep learning model that meets the requirements, and then estimating the health status of the power battery module based on the deep learning model that meets the requirements.

[0008] Preferably, in step S1, the second-order RC equivalent model of the power battery module is composed of l series modules connected in parallel; each series module is composed of n battery cells connected in series; each battery cell includes: an open circuit voltage, a first RC network, a second RC network, and an ohmic internal resistance connected in series; the first RC network includes: an electrochemical polarization resistor and an electrochemical polarization capacitor connected in parallel; the second RC network includes: a concentration polarization resistor and a concentration polarization capacitor connected in parallel; wherein the open circuit voltage is used to reflect the static potential of the power battery; the electrochemical The loop of the first RC network composed of the polarization resistor and the electrochemical polarization capacitor is used to reflect the electrochemical polarization effect of the power battery, wherein the electrochemical polarization internal resistance is used to reflect the resistance of the electrochemical reaction, and the polarization capacitor is used to reflect the double-layer capacitance; the loop of the second RC network composed of the concentration polarization resistor and the concentration polarization capacitor is used to reflect the concentration polarization effect of the power battery, wherein the concentration polarization resistor is used to reflect the resistance of ion diffusion, and the concentration polarization capacitor is used to reflect the capacitance effect caused by the change of ion concentration; the ohmic internal resistance is used to reflect the inherent internal resistance of the power battery;

[0009] The mathematical expression of the space state equation of the second-order RC equivalent model of the power battery cell is:

[0010]

[0011] In formula (1) and formula (2), is the battery terminal voltage; is the open circuit voltage; is the ohmic internal resistance; are the electrochemical polarization resistance and electrochemical polarization capacitance; is the electrochemical polarization voltage; are the concentration polarization resistance and concentration polarization capacitance; is the concentration polarization voltage; I sk The ohmic internal resistance of current.

[0012] Preferably, in step S2, the mathematical expressions of the timing models corresponding to the three parameters of the battery voltage, the battery state of charge and the battery capacity are respectively:

[0013]

[0014] In formula (3), is the battery voltage at time t; I sk (t) is the ohmic internal resistance at time t The current; is the electrochemical polarization voltage at time t; is the concentration polarization voltage at time t; is the battery state of charge at time t; is the initial state of charge of the battery; C sk is the battery capacity; τ is the integral variable based on time t; Q sk (t) is the battery capacity at time t, is the maximum capacity of the battery;

[0015] In step S3, the mathematical expression of the associated battery health state model of the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters is:

[0016]

[0017] In formula (4), The health status of the battery is pre-generated parameters; is the initial reference value of the battery internal resistance; For the generator are the initial reference values ​​of concentration polarization resistance and concentration polarization capacitance respectively; are the initial reference values ​​of electrochemical polarization resistance and electrochemical polarization capacitance, respectively; For the generator and is the weight coefficient of each parameter.

[0018] Preferably, the step S4 specifically includes: step S4.1: establishing a conditional generative adversarial network model for generating equivalent parameters based on the health status of the power battery module according to the associated battery health status model; the mathematical expression of the objective function of the conditional generative adversarial network model is:

[0019]

[0020] In formula (5), G and D are the generator and discriminator respectively; x and z are real data and random noise respectively; and The corresponding parameters are and The input noise of the parameter generator; x~p data With z~p z (z) are the probability distributions of real data and noise input respectively; is the data distribution x~p data Expectations of real data; is the data distribution z~p z (z) The expectation of random noise input;

[0021] Step S4.2: training the conditional generative adversarial network model based on the battery equivalent parameter set, and generating battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters; wherein the mathematical expression of the loss function of the conditional generative adversarial network model is:

[0022]

[0023] In formula (6), L D The loss function of the discriminator of the conditional adversarial network model is generated; L G The loss function for the generator of the conditional generative adversarial network model.

[0024] Preferably, the step S5 specifically includes: step S5.1: constructing a power battery module data set by performing Monte Carlo random sampling on battery cells that generate equivalent parameters, and establishing a multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance; wherein the mathematical expression of the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance is:

[0025]

[0026] In equations (7) and (8), M is the power battery module array, which consists of l×n battery cells connected in series and in parallel. The same row represents battery cells connected in series, and the column represents row battery packs connected in parallel. s,k It is the parameter information of the battery cell in the sth row and the kth column.

[0027] Preferably, in step S6, the mathematical expression of the mathematical model of the power battery module current, voltage and capacity state is:

[0028]

[0029] In formula (9), U s is the sum of the voltages of the individual cells in the sth row of series-connected batteries; U s,n is the voltage of the single cell in the sth row and the nth column; I is the sum of the currents of the battery packs in series in each row; C is the capacity of the power battery module, which is the sum of the capacities of the battery packs in series in each row; C minl The capacity of the smallest battery cell in the battery pack connected in series in the lth row.

[0030] Preferably, in step S7, the mathematical expression of the battery module health status characterization model is:

[0031]

[0032] In formula (10), S OH is the health status of the power battery module; C nom It is the initial reference value of the capacity of the power battery module.

[0033] Preferably, in step S8, the voltage, current and capacity time series data of the power battery module under dynamic working conditions are obtained by simulation, and the obtained voltage, current and capacity time series data of the battery module are normalized to process the obtained voltage, current and capacity time series data of the battery module into a mathematical expression corresponding to a space vector:

[0034]

[0035] In formula (11), I(t) is the charge and discharge current under dynamic conditions; f(t) is the current data of the power battery module under working conditions; h(t) is the preprocessing data; μ h , σ h are the mean and standard deviation of the data respectively; X icv is a three-dimensional vector of voltage, current and capacity data. I′(t), V′(t) and C′(V) are respectively the normalized time series data of the power battery module voltage, current and capacity data.

[0036] Preferably, the step S9 specifically includes: step S9.1: based on the characteristics of the curve space vector in step S8, establishing a deep learning model based on the residual network-gated recurrent unit-attention mechanism; wherein the mathematical expression corresponding to the deep learning model based on the residual network-gated recurrent unit-attention mechanism is:

[0037]

[0038] In formula (12), ReLU is a nonlinear activation function; BN is batch normalization; W oh (l) and b oh (l) are the weight and bias of the convolution kernel of the lth layer respectively; X icv (l) Input 3D curve space vector for layer l; z t is the update gate; h t-1 is the hidden state of the previous time step t-1; is the candidate hidden state; h t is the hidden state at the current time step t; e t is the attention score; α t is the attention weight; e K is the attention score of the Kth input; exp is the exponential operation; T is the input sequence length; c is the context vector;

[0039] Step S9.2: Input the context vector c into the fully connected layer to complete the classification or regression task. The corresponding mathematical expression for the SOH estimation of the power battery module is:

[0040]

[0041] In formula (13), W oh and b oh are the regression layer weights and biases, respectively. is the estimated health status;

[0042] Step S9.3: Use the mean square error as the loss function of the regression task, calculate the gradient through back propagation, and update the model parameters to continuously optimize the model; wherein the mathematical expression corresponding to the loss function of the regression task is:

[0043]

[0044] In formula (14), Γ is the loss function value; N is the total number of power battery module samples for health state estimation; is the predicted value of the health status of the power battery module; SOH,i It is the actual value of the health status of the power battery module.

[0045] Preferably, in step S10, the mathematical expression corresponding to the health status estimation of the power battery module based on the deep learning model that meets the requirements is:

[0046]

[0047] In formula (15), is the correlation between the predicted value and the true value of the health status of the power battery module. The closer it is to 1, the higher the estimation accuracy of the deep learning model; It is the average value of the true value of the power battery module health status samples.

[0048] Beneficial effects of the present invention:

[0049] The method for estimating the health state of a power battery module based on multi-parameter coupling provided by the present invention takes the power battery module with multi-parameter coupling as the research object, and proposes a strategy for estimating the health state of a battery module, which has a certain complexity. Specifically, the method for estimating the health state of a power battery module based on multi-parameter coupling provided by the present invention first characterizes the health state characteristics by using the internal resistance, polarization resistance and capacitance parameters of the battery equivalent model, and establishes a nonlinear correlation model between the equivalent parameters of the battery module and the health state. Secondly, the internal resistance, polarization resistance and capacitance equivalent parameters of the battery monomer equivalent model are obtained by generating the associated health state through data, and the data is processed by a random sampling method to construct a battery "monomer-module" data set, and the dynamic charge and discharge curve of the battery module under multiple working conditions is obtained. Finally, by establishing a deep learning model for estimating the health state of the battery module, the multi-working condition operation data of the battery module is comprehensively considered to estimate the health state of the battery module, so as to achieve the purpose of "estimating the health state of the battery module by generating multi-working condition data of the battery module, establishing and training a deep learning model, and shortening the evaluation time, reducing costs, and improving the screening efficiency.

[0050] Additional aspects and advantages of the invention will become apparent from the following description, or may be learned by practice of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A schematic flow chart of a method for estimating the health status of a power battery module based on multi-parameter coupling according to an embodiment of the present invention is shown;

[0052] Figure 2 A schematic flow chart showing a method for estimating the health status of a power battery module based on multi-parameter coupling according to another embodiment of the present invention is shown;

[0053] Figure 3 A second-order RC equivalent model diagram of a power battery module according to an embodiment of the present invention is shown;

[0054] Figure 4 A simulation result diagram showing the number of charge and discharge cycles of a battery module and the estimated and actual values ​​of the SOH of the battery module according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0055] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, Figures 1 to 4 As shown, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementations. It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0056] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0057] Figure 1 FIG. 1 is a schematic flow chart of a multi-time scale control method for electric heating gas hydrogen taking into account battery cascade utilization according to an embodiment of the present invention. Figure 1 As shown, the power battery module health status estimation method based on multi-parameter coupling includes:

[0058] Step S1: establishing a spatial state equation of a second-order RC equivalent model of a power battery module and a second-order RC equivalent model of a power battery cell;

[0059] Step S2: Differentiate and discretize the spatial state equation established in step S1 to obtain time series models corresponding to the three parameters of battery voltage, battery state of charge, and battery capacity;

[0060] Step S3: Based on the timing model obtained in step S2, a battery health status model associated with the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters is established;

[0061] Step S4: Based on the associated battery health status model established in step S3, a battery equivalent parameter data generation model is established, and then the battery equivalent parameter data generation model is trained based on the battery equivalent parameter set, and the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters are generated;

[0062] Step S5: constructing a power battery module data set by randomly sampling battery cells that generate equivalent parameters, and establishing a multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance;

[0063] Step S6: Based on the second-order RC equivalent model established in step S1 and the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance established in step S5, a mathematical model of battery module current, voltage and capacity state is established;

[0064] Step S7: constructing a battery module health status characterization model based on the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance established in step S5, and the mathematical model of battery module current, voltage and capacity state established in step S6;

[0065] Step S8: Based on the mathematical model of the battery module current, voltage and capacity state established in step S6 and the timing model obtained in step S2, the voltage, current and capacity timing data of the battery module under dynamic working conditions are obtained through simulation, and the obtained voltage, current and capacity timing data of the battery module are normalized to process the obtained voltage, current and capacity timing data of the battery module into a curve space vector;

[0066] Step S9: Based on the characteristics of the curve space vector of step S8, a deep learning model for estimating the health state of the battery module is established, and then the deep learning model for estimating the health state of the battery module is trained through the curve space vectors of the voltage, current, and capacity of the battery module and the battery module health state characterization model of step S7;

[0067] Step S10: Establish an evaluation model and evaluate the deep learning model established in step S9 to obtain a deep learning model that meets the requirements, and then estimate the health status of the power battery module based on the deep learning model that meets the requirements.

[0068] In this embodiment, the method for estimating the health state of a power battery module based on multi-parameter coupling provided by the present invention takes the power battery module with multi-parameter coupling as the research object, and proposes a strategy for estimating the health state of a battery module, which has a certain complexity. Specifically, the method for estimating the health state of a power battery module based on multi-parameter coupling provided by the present invention first characterizes the health state characteristics by using the internal resistance, polarization resistance and capacitance parameters of the battery equivalent model, and establishes a nonlinear correlation model between the equivalent parameters of the battery module and the health state. Secondly, the internal resistance, polarization resistance and capacitance equivalent parameters of the battery monomer equivalent model are obtained by generating the associated health state through data, and the data is processed by a random sampling method to construct a battery "monomer-module" data set, and the dynamic charge and discharge curve of the battery module under multiple working conditions is obtained. Finally, by establishing a deep learning model for estimating the health state of a battery module, the multi-working condition operation data of the battery module is comprehensively considered to estimate the health state of the battery module, so as to achieve the purpose of "estimating the health state of the battery module by generating multi-working condition data of the battery module, establishing and training a deep learning model, and shortening the evaluation time, reducing costs, and improving the screening efficiency.

[0069] In this embodiment, specifically, in step S4, a battery equivalent parameter data generation model may be established based on the associated battery health status model established in step S3, and the mathematical expression of the battery equivalent parameter data generation model is:

[0070]

[0071] Where M s,k The health status is The battery cell parameters generate an equivalent parameter data model; f is the battery equivalent parameter data generation model.

[0072] Further, specifically, in step S9, based on the characteristics of the curve space vector in step S8, the mathematical expression corresponding to the process of establishing the battery module health state estimation deep learning model and its training model is specifically:

[0073]

[0074] In the formula, g(X; φ X ) is the battery module data curve feature extraction module; φ X is the module parameter; It is a temporal modeling module that processes the temporal dependency of features and captures the law of battery changes over time; p(h; ω h ) is the prediction output module, which is mapped to the health state and generates the final regression value or classification result; Θ is the model parameter set; Θ * is the optimal parameter after optimization; argmin Θ is the minimization loss function about the parameter set Θ; l is the error function; Q(X i ; Θ) is the model prediction function; R(Θ) is the regularization term; λ is the weight of the regularization strength, and N is the number of training samples.

[0075] In one embodiment of the present invention, in step S1, the second-order RC equivalent model of the power battery module is composed of 1 series modules connected in parallel; each series module is composed of n battery cells connected in series; each battery cell includes: an open circuit voltage, a first RC network, a second RC network, and an ohmic internal resistance connected in series; the first RC network includes: an electrochemical polarization resistor and an electrochemical polarization capacitor connected in parallel; the second RC network includes: a concentration polarization resistor and a concentration polarization capacitor connected in parallel; wherein the open circuit voltage is used to reflect the static potential of the power battery; The loop of the first RC network composed of the electrochemical polarization resistor and the electrochemical polarization capacitor is used to reflect the electrochemical polarization effect of the power battery, wherein the electrochemical polarization internal resistance is used to reflect the resistance of the electrochemical reaction, and the polarization capacitor is used to reflect the double-layer capacitance; the loop of the second RC network composed of the concentration polarization resistor and the concentration polarization capacitor is used to reflect the concentration polarization effect of the power battery, wherein the concentration polarization resistor is used to reflect the resistance of ion diffusion, and the concentration polarization capacitor is used to reflect the capacitance effect caused by the change of ion concentration; the ohmic internal resistance is used to reflect the inherent internal resistance of the power battery. In a specific embodiment, Figure 3 As shown, the battery cells in row 1 and column 1 include: open circuit voltage The first RC network, the second RC network, and the ohmic internal resistance Composition: The first RC network includes: electrochemical polarization resistors connected in parallel and electrochemical polarization capacitance The second RC network includes: concentration polarization resistors connected in parallel and concentration polarization capacitance The battery cells in row 1 and column n include: The first RC network, the second RC network, and the ohmic internal resistance Composition: The first RC network includes: electrochemical polarization resistors connected in parallel and electrochemical polarization capacitance The second RC network includes: concentration polarization resistors connected in parallel and concentration polarization capacitance The battery cells in row 2 and column 1 include: The first RC network, the second RC network, and the ohmic internal resistance Composition: The first RC network includes: electrochemical polarization resistors connected in parallel and electrochemical polarization capacitance The second RC network includes: concentration polarization resistors connected in parallel and concentration polarization capacitance

[0076] The mathematical expression of the space state equation of the second-order RC equivalent model of the power battery cell is:

[0077]

[0078]

[0079] In formula (1) and formula (2), is the battery terminal voltage; is the open circuit voltage; is the ohmic internal resistance; are the electrochemical polarization resistance and electrochemical polarization capacitance; is the electrochemical polarization voltage; are the concentration polarization resistance and concentration polarization capacitance; is the concentration polarization voltage; I sk The ohmic internal resistance of current.

[0080] In one embodiment of the present invention, in step S2, the mathematical expressions of the timing models corresponding to the three parameters of the battery voltage, the battery state of charge, and the battery capacity are respectively:

[0081]

[0082] In formula (3), is the battery voltage at time t; I sk (t) is the ohmic internal resistance at time t The current; is the electrochemical polarization voltage at time t; is the concentration polarization voltage at time t; is the battery state of charge at time t; is the initial state of charge of the battery; C sk is the battery capacity; τ is the integral variable based on time t; Q sk (t) is the battery capacity at time t, is the maximum capacity of the battery;

[0083] In step S3, the mathematical expression of the associated battery health state model of the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters is:

[0084]

[0085] In formula (4), The health status of the battery is pre-generated parameters; is the initial reference value of the battery internal resistance; For the generator are the initial reference values ​​of concentration polarization resistance and concentration polarization capacitance respectively; are the initial reference values ​​of electrochemical polarization resistance and electrochemical polarization capacitance, respectively; For the generator and is the weight coefficient of each parameter.

[0086] In one embodiment of the present invention, the step S4 specifically includes: step S4.1: according to the associated battery health state model, establishing a conditional generative adversarial network model for generating equivalent parameters based on the health state of the power battery module; the mathematical expression of the objective function of the conditional generative adversarial network model is:

[0087]

[0088] In formula (5), G and D are the generator and discriminator respectively; x and z are real data and random noise respectively; and The corresponding parameters are and The input noise of the parameter generator; x~p data With z~p z (z) are the probability distributions of real data and noise input respectively; is the data distribution x~pdata Expectations of real data; is the data distribution z~p z (z) The expectation of random noise input;

[0089] Step S4.2: training the conditional generative adversarial network model based on the battery equivalent parameter set, and generating battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters; wherein the mathematical expression of the loss function of the conditional generative adversarial network model is:

[0090]

[0091] In formula (6), L D The loss function of the discriminator of the conditional adversarial network model is generated; L G The loss function for the generator of the conditional generative adversarial network model.

[0092] In one embodiment of the present invention, the step S5 specifically includes: step S5.1: constructing a power battery module data set by performing Monte Carlo random sampling on battery cells that generate equivalent parameters, and establishing a multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance; wherein the mathematical expression of the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance is:

[0093]

[0094]

[0095] In equations (7) and (8), M is the power battery module array, which consists of l×n battery cells connected in series and in parallel. The same row represents battery cells connected in series, and the column represents row battery packs connected in parallel. s,k It is the parameter information of the battery cell in the sth row and the kth column.

[0096] In one embodiment of the present invention, in step S6, the mathematical expression of the mathematical model of the current, voltage and capacity state of the power battery module is:

[0097]

[0098] In formula (9), U s is the sum of the voltages of the individual cells in the sth row of series-connected batteries; U s,n is the voltage of the single cell in the sth row and the nth column; I is the sum of the currents of the battery packs in series in each row; C is the capacity of the power battery module, which is the sum of the capacities of the battery packs in series in each row; C minl The capacity of the smallest battery cell in the battery pack connected in series in the lth row.

[0099] In one embodiment of the present invention, in step S7, the mathematical expression of the battery module health status characterization model is:

[0100]

[0101] In formula (10), S OH is the health status of the power battery module; C nom It is the initial reference value of the capacity of the power battery module.

[0102] In one embodiment of the present invention, in step S8, the voltage, current, and capacity time series data of the power battery module under dynamic conditions are obtained by simulation, and the obtained voltage, current, and capacity time series data of the battery module are normalized to process the obtained voltage, current, and capacity time series data of the battery module into a mathematical expression corresponding to a space vector:

[0103]

[0104] In formula (11), I(t) is the charge and discharge current under dynamic conditions; f(t) is the current data of the power battery module under working conditions; h(t) is the preprocessing data; μ h , σ h are the mean and standard deviation of the data respectively; X icv is a three-dimensional vector of voltage, current and capacity data. I′(t), V′(t) and C′(V) are respectively the normalized time series data of the power battery module voltage, current and capacity data.

[0105] In one embodiment of the present invention, the step S9 specifically includes: step S9.1: based on the characteristics of the curve space vector in step S8, establishing a deep learning model based on the residual network-gated recurrent unit-attention mechanism; wherein the mathematical expression corresponding to the deep learning model based on the residual network-gated recurrent unit-attention mechanism is:

[0106]

[0107] In formula (12), ReLU is a nonlinear activation function; BN is batch normalization; W oh (l) and b oh (l) are the weight and bias of the convolution kernel of the lth layer respectively; X icv (l) Input 3D curve space vector for layer l; z t is the update gate; h t-1 is the hidden state of the previous time step t-1; is the candidate hidden state; h t is the hidden state at the current time step t; et is the attention score; α t is the attention weight; e K is the attention score of the Kth input; exp is the exponential operation; T is the input sequence length; c is the context vector;

[0108] Step S9.2: Input the context vector c into the fully connected layer to complete the classification or regression task. The corresponding mathematical expression for the SOH estimation of the power battery module is:

[0109]

[0110] In formula (13), W oh and b oh are the regression layer weights and biases, respectively. is the estimated health status;

[0111] Step S9.3: Use the mean square error as the loss function of the regression task, calculate the gradient through back propagation, and update the model parameters to continuously optimize the model; wherein the mathematical expression corresponding to the loss function of the regression task is:

[0112]

[0113] In formula (14), Γ is the loss function value; N is the total number of power battery module samples for health state estimation; is the predicted value of the health status of the power battery module; SOH,i It is the actual value of the health status of the power battery module.

[0114] In one embodiment of the present invention, in step S10, the mathematical expression corresponding to the health status estimation of the power battery module based on the deep learning model that meets the requirements is:

[0115]

[0116] In formula (15), is the correlation between the predicted value and the true value of the health status of the power battery module. The closer it is to 1, the higher the estimation accuracy of the deep learning model; It is the average value of the true value of the power battery module health status samples.

[0117] Figure 2 FIG. 2 is a schematic flow chart of a method for estimating the health status of a power battery module based on multi-parameter coupling according to another embodiment of the present invention. Figure 2As shown, the method for estimating the health state of a power battery module based on multi-parameter coupling includes: extracting data such as voltage, current, capacity, internal resistance, etc. of the operating conditions of the battery module; establishing an equivalent circuit model of the battery module and establishing a state-space equation; characterizing a battery module health state SOH model; establishing a nonlinear correlation model between dynamic parameters of the battery module and the health state; setting a limiting current in consideration of the correlation between module battery parameters; establishing operating constraints; establishing a battery module inconsistency evaluation index; training a GAN through operating condition data; judging in real time whether the trained GAN meets the requirements; when the judgment result is no, returning to the step of training the GAN through the operating condition data until the trained GAN meets the requirements; when the trained GAN meets the requirements, generating multi-operating condition data based on the GAN; obtaining battery cell operating condition data; judging in real time whether the generated cell operating condition data meets the actual requirements; when the judgment result is yes, randomly sampling based on the generated cell data; when When the judgment result is no, return to the step of generating multi-operating condition data based on GAN until the generated multi-operating condition data meets the requirements; construct a battery module data set; perform battery module inconsistency evaluation; judge in real time whether the battery module inconsistency meets the evaluation requirements; when the judgment result is yes, normalize and discretize the battery operation data and process it into a space vector; when the judgment result is no, return to the step of constructing the battery module data set; after normalizing and discretizing the battery operation data and processing it into a space vector, perform residual network data feature extraction; perform gated recurrent unit to capture timing features; perform attention mechanism weight analysis; perform model optimization training; judge in real time whether the optimized trained model meets the requirements; when the judgment result is no, return to the step of model optimization training until the optimized trained model meets the requirements; after the optimized trained model meets the requirements, it means that the model training is completed, and the health status of the battery module is evaluated based on the model that meets the requirements.

[0118] In this embodiment, if Figure 2 As shown, the established battery equivalent data generation model can quickly generate a large amount of battery equivalent data and be applied to the training of deep models for battery module health status estimation, shortening the evaluation time, reducing costs, and improving screening efficiency.

[0119] Figure 4 The simulation result diagram of the number of charge and discharge cycles of a battery module and the estimated and actual values ​​of the battery module SOH according to an embodiment of the present invention is shown. Figure 4 As shown, MATLAB is used to simulate and analyze the above power battery module health status estimation model to estimate the battery module health status. It can be seen that with the increase in the number of battery charge and discharge cycles, the accuracy of the power battery module health status estimation model will gradually increase with training, and finally the health status of the power battery module can be estimated within a smaller error range.

[0120] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for estimating the health status of a power battery module based on multi-parameter coupling, comprising: Step S1: establishing a spatial state equation of a second-order RC equivalent model of a power battery module and a second-order RC equivalent model of a power battery cell; Step S2: Differentiate and discretize the spatial state equation established in step S1 to obtain time series models corresponding to the three parameters of battery voltage, battery state of charge, and battery capacity; Step S3: Based on the timing model obtained in step S2, a battery health status model associated with the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters is established; Step S4: Based on the associated battery health status model established in step S3, a battery equivalent parameter data generation model is established, and then the battery equivalent parameter data generation model is trained based on the battery equivalent parameter set, and the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters are generated; Step S5: constructing a power battery module data set by randomly sampling battery cells that generate equivalent parameters, and establishing a multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance; Step S6: Based on the second-order RC equivalent model established in step S1 and the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance established in step S5, a mathematical model of battery module current, voltage and capacity state is established; Step S7: constructing a battery module health status characterization model based on the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance established in step S5, and the mathematical model of battery module current, voltage and capacity state established in step S6; Step S8: Based on the mathematical model of the battery module current, voltage and capacity state established in step S6 and the timing model obtained in step S2, the voltage, current and capacity timing data of the battery module under dynamic working conditions are obtained through simulation, and the obtained voltage, current and capacity timing data of the battery module are normalized to process the obtained voltage, current and capacity timing data of the battery module into a curve space vector; Step S9: Based on the characteristics of the curve space vector of step S8, a deep learning model for estimating the health state of the battery module is established, and then the deep learning model for estimating the health state of the battery module is trained through the curve space vectors of the voltage, current, and capacity of the battery module and the battery module health state characterization model of step S7; Step S10: Establish an evaluation model and evaluate the deep learning model established in step S9 to obtain a deep learning model that meets the requirements, and then estimate the health status of the power battery module based on the deep learning model that meets the requirements.

2. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 1, characterized in that: In step S1, the second-order RC equivalent model of the power battery module is composed of l series modules connected in parallel; each series module is composed of n battery cells connected in series; Each battery cell comprises: an open circuit voltage, a first RC network, a second RC network, and an ohmic internal resistance connected in series; the first RC network comprises: an electrochemical polarization resistor and an electrochemical polarization capacitor connected in parallel; the second RC network comprises: a concentration polarization resistor and a concentration polarization capacitor connected in parallel; wherein the open circuit voltage is used to reflect the static potential of the power battery; the loop of the first RC network composed of the electrochemical polarization resistor and the electrochemical polarization capacitor is used to reflect the electrochemical polarization effect of the power battery, wherein the electrochemical polarization internal resistance is used to reflect the resistance of the electrochemical reaction, and the polarization capacitor is used to reflect the double-layer capacitance; the loop of the second RC network composed of the concentration polarization resistor and the concentration polarization capacitor is used to reflect the concentration polarization effect of the power battery, wherein the concentration polarization resistor is used to reflect the resistance of ion diffusion, and the concentration polarization capacitor is used to reflect the capacitance effect caused by the change of ion concentration; the ohmic internal resistance is used to reflect the inherent internal resistance of the power battery; The mathematical expression of the space state equation of the second-order RC equivalent model of the power battery cell is: In formula (1) and formula (2), is the battery terminal voltage; is the open circuit voltage; is the ohmic internal resistance; are the electrochemical polarization resistance and electrochemical polarization capacitance; is the electrochemical polarization voltage; are the concentration polarization resistance and concentration polarization capacitance; is the concentration polarization voltage; I sk The ohmic internal resistance of current.

3. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 2, characterized in that: In step S2, the mathematical expressions of the timing models corresponding to the three parameters of the battery voltage, the battery state of charge, and the battery capacity are respectively: In formula (3), is the battery voltage at time t; I sk (t) is the ohmic internal resistance at time t The current; is the electrochemical polarization voltage at time t; is the concentration polarization voltage at time t; is the battery state of charge at time t; is the initial state of charge of the battery; C sk is the battery capacity; τ is the integral variable based on time t; Q sk (t) is the battery capacity at time t, is the maximum capacity of the battery; In step S3, the mathematical expression of the battery health state model associated with the battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters is: In formula (4), The health status of the battery is pre-generated parameters; is the initial reference value of the battery internal resistance; For the generator are the initial reference values ​​of concentration polarization resistance and concentration polarization capacitance respectively; are the initial reference values ​​of electrochemical polarization resistance and electrochemical polarization capacitance, respectively; For the generator and is the weight coefficient of each parameter.

4. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 3, characterized in that: The step S4 specifically includes: Step S4.1: According to the associated battery health state model, a conditional generative adversarial network model is established to generate equivalent parameters based on the health state of the power battery module; the mathematical expression of the objective function of the conditional generative adversarial network model is: In formula (5), G and D are the generator and discriminator respectively; x and z are real data and random noise respectively; and The corresponding parameters are and The input noise of the parameter generator; x~p data With z~p z (z) are the probability distributions of real data and noise input respectively; is the data distribution x~p data Expectations of real data; is the data distribution z~p z (z) the expectation of random noise input; Step S4.2: training the conditional generative adversarial network model based on the battery equivalent parameter set, and generating battery equivalent model internal resistance, polarization resistance and capacitance equivalent parameters; wherein the mathematical expression of the loss function of the conditional generative adversarial network model is: In formula (6), L D The loss function of the discriminator of the conditional adversarial network model is generated; L G The loss function for the generator of the conditional generative adversarial network model.

5. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 4, characterized in that: The step S5 specifically includes: Step S5.1: A power battery module data set is constructed by performing Monte Carlo random sampling on battery cells with equivalent parameters, and a multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance is established; wherein the mathematical expression of the multi-parameter equivalent model of battery module internal resistance, polarization resistance and capacitance is: In equations (7) and (8), M is the power battery module array, which consists of l×n battery cells connected in series and in parallel. The same row represents battery cells connected in series, and the column represents row battery packs connected in parallel. s,k It is the parameter information of the battery cell in the sth row and the kth column.

6. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 5, characterized in that: In step S6, the mathematical expression of the current, voltage and capacity state mathematical model of the power battery module is: In formula (9), U s is the sum of the voltages of the individual cells in the sth row of series-connected batteries; U s,n is the voltage of the single cell in the sth row and the nth column; I is the sum of the currents of the battery packs in series in each row; C is the capacity of the power battery module, which is the sum of the capacities of the battery packs in series in each row; C minl The capacity of the smallest battery cell in the battery pack connected in series in the lth row.

7. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 6, characterized in that: In step S7, the mathematical expression of the battery module health status characterization model is: In formula (10), S OH is the health status of the power battery module; C nom It is the initial reference value of the capacity of the power battery module.

8. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 7, characterized in that: In step S8, the voltage, current, and capacity time series data of the power battery module under dynamic working conditions are obtained through simulation, and the obtained voltage, current, and capacity time series data of the battery module are normalized to process the obtained voltage, current, and capacity time series data of the battery module into mathematical expressions corresponding to space vectors: In formula (11), I(t) is the charge and discharge current under dynamic conditions; f(t) is the current data of the power battery module under working conditions; h(t) is the preprocessing data; μ h , σ h are the mean and standard deviation of the data respectively; X icv is a three-dimensional vector of voltage, current and capacity data. I′(t), V′(t) and C′(V) are respectively the normalized time series data of the power battery module voltage, current and capacity data.

9. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 8, characterized in that: The step S9 specifically includes: Step S9.1: Based on the features of the curve space vector in step S8, a deep learning model based on a residual network-gated recurrent unit-attention mechanism is established; wherein the mathematical expression corresponding to the deep learning model based on a residual network-gated recurrent unit-attention mechanism is: In formula (12), ReLU is a nonlinear activation function; BN is batch normalization; W oh (l) and b oh (l) are the weight and bias of the convolution kernel of the lth layer respectively; X icv (l) Input 3D curve space vector for layer l; z t is the update gate; h t-1 is the hidden state of the previous time step t-1; is the candidate hidden state; h t is the hidden state at the current time step t; e t is the attention score; α t is the attention weight; e K is the attention score of the Kth input; exp is the exponential operation; T is the input sequence length; c is the context vector; Step S9.2: Input the context vector c into the fully connected layer to complete the classification or regression task. The corresponding mathematical expression for the SOH estimation of the power battery module is: In formula (13), W oh and b oh are the regression layer weights and biases, respectively. is the estimated health status; Step S9.3: Use the mean square error as the loss function of the regression task, calculate the gradient through back propagation, and update the model parameters to continuously optimize the model; wherein the mathematical expression corresponding to the loss function of the regression task is: In formula (14), Γ is the loss function value; N is the total number of power battery module samples for health state estimation; is the predicted value of the health status of the power battery module; SOH,i It is the actual value of the health status of the power battery module.

10. The method for estimating the health status of a power battery module based on multi-parameter coupling according to claim 9, characterized in that: In step S10, the mathematical expression corresponding to the health status estimation of the power battery module based on the deep learning model that meets the requirements is: In formula (15), is the correlation between the predicted value and the true value of the health status of the power battery module. The closer it is to 1, the higher the estimation accuracy of the deep learning model; It is the average value of the true value of the power battery module health status samples.

Images