Automatic construction method of lithium ion battery electrothermal coupling model

Abstract

The invention relates to a lithium ion battery electrothermal coupling model automatic construction method comprising the following steps: carrying out charging and discharging tests on a lithium ion battery to obtain electrical characteristic parameters and thermal characteristic parameters; based on the electrical characteristic parameters and the thermal characteristic parameters, constructing an ECM electrothermal coupling model based on a neural network; performing ECM electrothermal coupling model parameter identification and synchronous optimization of a model structure by optimizing a loss function; and building an electrothermal coupling model based on the neural network based on the simulation platform. Compared with the prior art, the invention provides an electrothermal coupling model automatic construction strategy simultaneously considering temperature and SOC effects, and uses the artificial neural network to quantitatively capture the relationship between circuit parameters and temperature / SOC, thereby realizing mutual coupling of electrical characteristics and thermal characteristics of the lithium ion battery. And the model complexity can be reduced through automatic model selection optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium ion batteries, and in particular to a method for automatically constructing an electric-thermal coupling model of a lithium ion battery. Background Art

[0002] As one of the most promising energy storage technologies, lithium-ion batteries have been increasingly used in portable electronic devices, electric vehicles, hybrid vehicles, and smart grids. However, lithium-ion batteries also have high safety risks. Therefore, it is necessary to monitor, detect, estimate and control lithium-ion batteries in real time. Existing technologies mainly rely on battery management systems (BMS) to ensure the safe and reliable operation of lithium-ion batteries. Battery models are the basis for BMS to estimate battery status, warn of risks, and control management. However, due to the complex chemical reactions inside lithium-ion batteries and the mutual coupling of multiple physical fields such as electricity and heat, it is extremely challenging to establish a battery model with strong nonlinearity, accuracy, and low complexity.

[0003] Lithium-ion battery models include mechanism models, data-driven models, and equivalent circuit models (ECM). Among them, the mechanism model studies the microstructure and thermodynamic kinetics of the battery, and uses partial differential equations to describe various mass transfer and transport processes inside the battery. However, due to its high computational complexity, its application in BMS is limited; the data-driven model uses modern machine learning methods and has a high generalization ability for nonlinear lithium-ion batteries, but this method relies on a large amount of test data for analysis, requires a large number of samples, requires high computing equipment, requires test results under different working conditions, has a high initial training time cost, and is difficult to cover all usage conditions. There are certain limitations in practical applications; the electrothermal coupling model based on ECM has high computational accuracy and relatively low computational complexity, and is widely used.

[0004] The key to the coupling process of the electrical model and the thermal model is to explicitly embed nonlinear factors such as temperature and battery state of charge (SOC) into the circuit parameters to improve the prediction accuracy of electrical and thermal characteristics. However, the currently developed electrical and thermal coupling models not only lack a general modeling method to simply embed various effects such as temperature, SOC, health status and current into circuit parameters, but also lack an automatic model selection method to optimize and reduce model complexity to meet the low computational requirements in actual BMS. Summary of the invention

[0005] The purpose of the present invention is to provide a method for automatically constructing an electrothermal coupling model of a lithium-ion battery in order to overcome the defects of the above-mentioned prior art, which can automatically construct an electrothermal coupling model of a lithium-ion battery based on simultaneous consideration of temperature and SOC effects.

[0006] The purpose of the present invention can be achieved by the following technical solution: A method for automatically constructing an electric-thermal coupling model of a lithium-ion battery comprises the following steps:

[0007] S1. Perform charge and discharge tests on lithium-ion batteries to obtain electrical and thermal characteristic parameters;

[0008] S2, based on the electrical characteristic parameters and thermal characteristic parameters, construct an ECM electrothermal coupling model based on a neural network;

[0009] S3, by optimizing the loss function, the ECM electrothermal coupling model parameters are identified and the model structure is optimized simultaneously;

[0010] S4. Build a neural network-based electrothermal coupling model based on the simulation platform.

[0011] Furthermore, the electrical characteristic parameters and thermal characteristic parameters in step S1 include the change pattern of the open circuit voltage OCV (Open circuit voltage) of the lithium-ion battery with SOC, the entropy thermal coefficient, and the change pattern of the current, voltage and temperature with time under the set working conditions.

[0012] Furthermore, the step S1 specifically includes the following steps:

[0013] S11, performing an open circuit voltage curve test on the lithium-ion battery to obtain a SOC-OCV curve of the lithium-ion battery at a set temperature;

[0014] S12, according to the SOC-OCV curves at different set temperatures, obtain the OCV=f(SOC, T) three-dimensional surface by fitting or interpolation to obtain the OCV model;

[0015] Under the set SOC condition, the corresponding entropy thermal coefficient is determined as:

[0016]

[0017] Where T is the average surface temperature of the battery;

[0018] S13. Determine the response curve of the battery terminal voltage and temperature over time under the set working conditions of the lithium-ion battery.

[0019] Furthermore, the specific process of step S11 is as follows:

[0020] Place the lithium-ion battery in a constant temperature box and wait for 5 hours after the temperature in the constant temperature box reaches the target temperature to ensure that the battery reaches a thermodynamic equilibrium state;

[0021] The battery was charged at a constant current and constant voltage (CC-CV) rate of 1C to the cut-off voltage and cut-off current, and the battery was discharged at a constant current rate of 1C to a certain SOC. After standing for 2 hours, the open circuit voltage close to equilibrium was measured;

[0022] Repeat the above steps to obtain the SOC-OCV curve of the lithium-ion battery at a specific temperature.

[0023] Furthermore, the step S2 specifically includes the following steps:

[0024] S21, respectively establishing a first-order equivalent circuit model for describing the electrical characteristics of lithium ions and a thermal characteristic model of lithium ion batteries;

[0025] S22. Establish a single-layer feedforward network model with multiple inputs and outputs, couple the first-order equivalent circuit model and the thermal characteristic model, and obtain an electrothermal coupling model based on a neural network.

[0026] Furthermore, the first-order equivalent circuit model in step S21 is specifically:

[0027]

[0028] Among them, V 1，k is the voltage across the RC network of the battery at time k, V 0，k is the voltage across the resistor R0 at time k, Δt is the time interval from time k-1 to time k, OCV k is the open circuit voltage of the battery at time k, I k is the current flowing through the external circuit at time k, R 0，k , R 1，k , C 1，k represents the values ​​of the resistors R0, R1 and capacitor C1 at time k, τ is the time constant, τ k Indicates the value of the time constant at time k, which stipulates that the battery charging current is positive and the discharge current is negative;

[0029] The thermal characteristic model of the lithium-ion battery is specifically a lumped thermal model, and the mathematical formula describing the heat generation rate of the battery is:

[0030]

[0031] Where Q is the heat generation rate, I is the current flowing through the battery, OCV is the open circuit voltage of the battery, V tis the battery terminal voltage, T is the average surface temperature of the battery, the first term on the right side of the equation is the irreversible heat during the operation of the battery cell, and the second term is the reversible heat during the operation of the battery cell;

[0032] Assuming the battery surface temperature is uniform, then:

[0033]

[0034] Where, m is the mass of the battery cell, C p is the total heat capacity of the cell, A is the surface area of ​​the cell, h is the heat transfer coefficient, and the thermal parameter C p and h are both identified using cell measurement data;

[0035] After discretization, we get:

[0036]

[0037] Among them, T amb is the ambient temperature.

[0038] Furthermore, the step S22 specifically includes the following steps:

[0039] Through a single-layer feedforward network with multiple inputs and multiple outputs, a single-layer feedforward network model of three circuit parameters R0, R1, and C1 relative to temperature and SOC is constructed;

[0040] The electrical characteristic model calculates the SOC and the battery terminal voltage V based on the applied current I and circuit parameters. t , where the circuit parameters include OCV, R0, R1, and C1. OCV is calculated by the open circuit voltage model obtained by experimental measurement, and R0, R1, and C1 are calculated by the neural network model based on SOC and temperature T;

[0041] Set OCV, I, V t Used to calculate the irreversible heat generation rate; the entropy heat coefficient T and I are used to calculate the reversible heating rate;

[0042] The reversible heat generation rate and the irreversible heat generation rate are then added together and used in the thermal model to calculate the temperature at the next moment.

[0043] Furthermore, the single-layer feedforward network contains l hidden layer nodes, specifically:

[0044]

[0045] g(a,b,x)=1 / (1+exp(-(a·x+b)))

[0046]

[0047] Where x∈Rn , y∈R p are input and output vectors respectively, the input vector x is [SOC, T] T , the output vector is [R0,R1,C1] T , R 0,k , R 1,k , C 1,k is the circuit parameter at time k, a j ∈R n , b j ∈R are the input weight and bias of the jth neuron, β ij is the output weight of the jth neuron relative to the i-th component of the output vector, g is the activation function, the weight a and bias b are randomly generated, and the output weight β is obtained by parameter identification.

[0048] Furthermore, the specific process of step S3 is as follows:

[0049] The weighted normalized sum of squares error is defined as:

[0050]

[0051] Among them, y k and Represent the measured and predicted values ​​of the terminal voltage and temperature respectively.

[0052] Since the circuit parameters R0, R1, and C1 have clear physical meanings, their non-negativity needs to be guaranteed. The Latin hypercube sampling method is used to generate random samples within the SOC and temperature range, and the nonlinear semi-infinite programming problem is transformed into a nonlinear programming problem with finite constraints. The following constraints are added in the parameter identification:

[0053]

[0054] Among them, SOC p , T p are the pth SOC and temperature samples, respectively, and P is the specified number of samples generated by the Latin hypercube sampling method;

[0055] Taking the weighted normalized sum of squares as the optimization target, the optimization problem is solved and the parameter identification process is completed. Considering that with the increase of hidden neurons in the neural network, the fitting ability can be improved, but the prediction performance may deteriorate and the computational cost will also become higher, therefore, in order to study the influence of the number of hidden neurons on the fitting and prediction performance of the electrothermal coupling model based on artificial neural network, a penalty objective function containing the l1 norm is defined to achieve the optimal pruning of hidden neurons:

[0056]

[0057] Therefore, the loss function containing the l1 penalty norm is taken as the optimization target to solve the above optimization problem and realize the parameter identification and automatic model construction process.

[0058] Furthermore, the step S4 specifically builds a lithium-ion battery electric-thermal coupling simulation model in MATLAB or PYTHON, and performs corresponding setting working condition simulation.

[0059] Compared with the prior art, the present invention has the following advantages:

[0060] The present invention performs charge and discharge tests on lithium-ion batteries to obtain electrical characteristic parameters and thermal characteristic parameters (including the change law of the open circuit voltage OCV of the lithium-ion battery with SOC, the entropy thermal coefficient, and the change law of the current, voltage and temperature with time under characteristic working conditions), and then constructs an ECM electrothermal coupling model based on a neural network, and realizes the simultaneous optimization of model parameter identification and model structure by optimizing the loss function, thereby realizing an automatic construction strategy of the electrothermal coupling model that simultaneously considers the temperature and SOC effects, and uses an artificial neural network to quantitatively capture the relationship between circuit parameters and temperature / SOC, thereby realizing the mutual coupling of the electrical and thermal characteristics of the lithium-ion battery, and being able to reduce the model complexity through automatic model selection optimization.

[0061] The present invention proposes a unified model construction framework for the strong nonlinear dynamic characteristics of lithium-ion batteries in electricity and heat, and uses neural networks to model temperature / SOC related circuit parameters, which can provide accurate model support for battery state estimation, energy management, and safety warning.

[0062] When performing simultaneous optimization of the ECM electrothermal coupling model parameter identification and model structure, the present invention adopts a Latin hypercube sampling method to generate random samples within the SOC and temperature range, and converts the nonlinear semi-infinite programming problem that is difficult to solve effectively into a nonlinear programming problem with finite constraints, which can greatly simplify the electrothermal coupling model solution process. In addition, in the parameter identification process, a loss function containing an l1 penalty norm is used as the optimization target, which can achieve the optimal pruning of hidden neurons and ensure the optimal parameter identification and model optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0064] Figure 2 It is a schematic diagram of the first-order equivalent circuit model;

[0065] Figure 3 Schematic diagram of the coupling strategy of the neural network, electrical characteristic model and thermal model;

[0066] Figure 4a , 4bSchematic diagram of neural network parameter identification results for a hidden layer node number range of 1-10 in an embodiment;

[0067] Figure 5a , 5b Schematic diagram of fitting results and prediction results when the number of hidden layer nodes is 3, 6, and 9 respectively in the embodiment;

[0068] Figure 6a , 6b Schematic diagram of the optimal pruning results of parameter identification and model in the embodiment;

[0069] Figure 7a , 7b It is a schematic diagram comparing the model fitting value and the experimental results in the embodiment;

[0070] Figure 8a , 8b Schematic diagram for comparing the model simulation results and the experimental test results in the embodiment. DETAILED DESCRIPTION

[0071] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0072] Example

[0073] like Figure 1 As shown, a method for automatically constructing an electrothermal coupling model of a lithium-ion battery includes the following steps:

[0074] S1. Perform charge and discharge tests on lithium-ion batteries to obtain electrical and thermal characteristic parameters;

[0075] S2, based on the electrical characteristic parameters and thermal characteristic parameters, construct an ECM electrothermal coupling model based on a neural network;

[0076] S3, by optimizing the loss function, the ECM electrothermal coupling model parameters are identified and the model structure is optimized simultaneously;

[0077] S4. Build a neural network-based electrothermal coupling model based on the simulation platform.

[0078] This embodiment applies the above technical solution and performs model accuracy verification. The main process includes:

[0079] 1. By performing charge and discharge tests on lithium-ion batteries, the electrical characteristic model parameters and thermal characteristic model parameters are obtained: including the change law of the open circuit voltage OCV of the lithium-ion battery with SOC, the entropy thermal coefficient, and the current, voltage and temperature change curves with time under characteristic working conditions.

[0080] 2. Construct an electrothermal coupling model based on neural network.

[0081] 3. Simultaneous selection of model parameter identification and model structure is achieved by optimizing the loss function.

[0082] 4. Build a neural network-based electrothermal coupling model based on the simulation platform and verify the model accuracy under dynamic conditions.

[0083] The battery charge and discharge test performed in step 1 of this embodiment takes a 20Ah lithium iron phosphate soft-pack battery as an example, and the specific process is as follows:

[0084] 1.1. Lithium-ion battery open circuit voltage curve test

[0085] Place the lithium-ion battery in a constant temperature box and wait for 5 hours after the temperature in the constant temperature box reaches the target temperature to ensure that the battery reaches a thermodynamic equilibrium state;

[0086] The battery was charged at a constant current and constant voltage (CC-CV) rate of 1C to the cut-off voltage and cut-off current, and the battery was discharged at a constant current rate of 1C to a certain SOC. After standing for 2 hours, the open circuit voltage close to equilibrium was measured;

[0087] Repeat the above steps to obtain the SOC-OCV curve of the lithium-ion battery at a specific temperature;

[0088] 1.2. Lithium-ion battery entropy thermal coefficient test

[0089] Change the temperature and repeat step 1.1 to measure the SOC-OCV curve of lithium-ion batteries at different temperatures. Obtain the OCV model by fitting or interpolating the three-dimensional surface of OCV=f(SOC, T). At the same time, at a specific SOC, That is the entropy thermal coefficient.

[0090] 1.3. Characteristic working condition test

[0091] Determine the response curve of the battery's terminal voltage and temperature over time under PCD (pulse current discharge), CCD (constant current discharge), and HPPC (Hybrid Pulse Power Characterization) conditions.

[0092] Step 2: Construct an ECM electrothermal coupling model based on a neural network. The specific process is as follows:

[0093] 2.1 Electrical characteristics model

[0094] The first-order equivalent circuit model describing the electrical characteristics of lithium ions is as follows: Figure 2 As shown;

[0095] In the first-order equivalent circuit model, the mathematical formula describing the quantitative relationship between circuit characteristic parameters is:

[0096]

[0097] Among them, V 1，k is the voltage across the RC network of the battery at time k, V 0，k is the voltage across the resistor R0 at time k, Δt is the time interval from time k-1 to time k, OCV k is the open circuit voltage of the battery at time k, I k is the current flowing through the external circuit at time k, R 0，k , R 1，k , C 1，k represents the values ​​of the resistors R0, R1 and capacitor C1 at time k, τ is the time constant, τ k It represents the value of the time constant at time k. Assuming that the battery charging current is positive and the discharging current is negative, the above formula is valid for both the charging and discharging processes.

[0098] 2.2 Thermal characteristics model

[0099] The thermal characteristic model is a lumped thermal model, and the mathematical formula describing the heat generation rate of the battery is:

[0100]

[0101] Where Q is the heat generation rate, I is the current flowing through the battery, OCV is the open circuit voltage of the battery, V t is the battery terminal voltage, and T is the average surface temperature of the battery. The first term on the right side of the equation is the irreversible heat during the operation of the battery cell, and the second term is the reversible heat during the operation of the battery cell.

[0102] Assuming the battery surface temperature is uniform, then:

[0103]

[0104] Where, m is the mass of the battery cell, C p is the total heat capacity of the cell, A is the surface area of ​​the cell, and h is the heat transfer coefficient. Thermal parameters C p and h need to be identified using cell measurement data.

[0105] Discretize the above formula to get:

[0106]

[0107] 2.3. Electrical and thermal characteristics coupling strategy based on neural network

[0108] The construction process of the electrothermal coupling model is as follows: Figure 3 As shown, the specific steps are:

[0109] (1) A single-layer feedforward network model of three circuit parameters R0, R1, and C1 relative to temperature and SOC is constructed through a single-layer feedforward network with multiple inputs and multiple outputs.

[0110] (2) The electrical characteristic model calculates the SOC and the battery terminal voltage V based on the applied current I and circuit parameters (including OCV, R0, R1, C1). t , where OCV is calculated by the open circuit voltage model obtained from experimental measurements, and R0, R1, and C1 are calculated by the neural network model based on SOC and temperature T.

[0111] (3) OCV, I, V calculated by the electrical characteristics model t Used to calculate irreversible heat generation rate; entropy thermal coefficient T and I are used to calculate the reversible heat generation rate, and then the reversible heat generation rate is added to the irreversible heat generation rate to calculate the temperature at the next moment in the thermal model.

[0112] Specifically, a single-layer feedforward network with l hidden layer nodes is:

[0113]

[0114] Where x∈R n , y∈R p are the input and output vectors respectively, and a j ∈R n , b j ∈R are the input weight and bias of the jth neuron, β ij is the output weight of the jth neuron relative to the i-th component of the output vector. g is the activation function, which is the sigmoid function in this scheme:

[0115] g(a,b,x)=1 / (1+exp(-(a·x+b)))

[0116] The weights a and biases b are randomly generated, while the output weights β are obtained by parameter identification. The input vector x is [SOC, T] T , the output vector is [R0,R1,C1] T According to the above definition, the circuit parameters at time k can be written as:

[0117]

[0118] Step 3 is to perform parameter identification and automatic model construction. The specific process is as follows:

[0119] The weighted normalized sum of squares error is defined as:

[0120]

[0121] Among them, y k and Represent the measured and predicted values ​​of the terminal voltage and temperature respectively.

[0122] Since the circuit parameters R0, R1, and C1 have clear physical meanings, their non-negativity needs to be guaranteed. This solution uses Latin hypercube sampling to generate random samples within the SOC and temperature range, transforming the nonlinear semi-infinite programming problem into a nonlinear programming problem with finite constraints. The following constraints are added in parameter identification:

[0123]

[0124] Among them, SOC p , T p are the pth SOC and temperature samples respectively, and P is the specified number of samples generated by the Latin hypercube sampling method.

[0125] Taking the weighted normalized sum of squares as the optimization target, the parameters of the neural network with the number of hidden layer nodes ranging from 1 to 10 were identified under the PCD condition. Figure 4a , 4b As shown in the figure, the fitting results and prediction results of the hidden layer with 3, 6, and 9 nodes are shown in Figure 5a , 5b shown.

[0126] As the number of hidden neurons in the neural network increases, the fitting ability can be improved, but the prediction performance may deteriorate and the computational cost will also increase. Therefore, this scheme studies the influence of the number of hidden neurons on the fitting and prediction performance of the electrothermal coupling model based on artificial neural network, and defines a penalty objective function containing the l1 norm to achieve the optimal pruning of hidden neurons:

[0127]

[0128] At the same time, since R0, R1, and C1 have different orders of magnitude, the regularization parameters λ2 and λ3 are equal to λ1 and 10 respectively. -6 λ1.

[0129] Taking the loss function with l1 penalty norm as the optimization target, when the number of hidden layer nodes is 10, the above optimization problem is solved in the range of logλ -6 to 1 to achieve parameter identification and optimal pruning of the model. The results are as follows: Figure 6a , 6b Furthermore, when logλ=1, logλ=-2 and logλ=-5, the model fitting value is compared with the experimental results. Figure 7a , 7b shown.

[0130] According to the parameter identification results, it can be seen that the model parameter identification method with l1 norm penalty term provides a more convenient way to adaptively determine the optimal terms of different circuit parameters, which will greatly help to build simpler models without sacrificing prediction performance.

[0131] Step 4: Build an electric-thermal coupling model based on a neural network based on the simulation platform, and verify the model accuracy under dynamic conditions. The specific process is as follows:

[0132] Build a lithium-ion battery electrothermal coupling simulation model in MATLAB or PYTHON, simulate the HPPC characteristic working conditions, compare the simulation results with the experimental test results, and verify the results. Figure 8a , 8b According to the verification results, under the condition of logλ=-5, the model has high accuracy in predicting the temperature and voltage.

Claims

1. A method for automatically constructing an electrothermal coupling model of a lithium-ion battery, characterized in that: The following steps are involved: S1. Perform charge and discharge tests on lithium-ion batteries to obtain electrical and thermal characteristic parameters; S2, based on the electrical characteristic parameters and thermal characteristic parameters, construct an ECM electrothermal coupling model based on a neural network; S3, by optimizing the loss function, the ECM electrothermal coupling model parameters are identified and the model structure is optimized simultaneously; S4. Build a neural network-based electrothermal coupling model based on the simulation platform.

2. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 1, characterized in that: The electrical characteristic parameters and thermal characteristic parameters in step S1 include the change pattern of the open circuit voltage OCV of the lithium-ion battery with SOC, the entropy thermal coefficient, and the change pattern of the current, voltage and temperature with time under the set working conditions.

3. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 2, characterized in that: The step S1 specifically includes the following steps: S11, performing an open circuit voltage curve test on the lithium-ion battery to obtain a SOC-OCV curve of the lithium-ion battery at a set temperature; S12, according to the SOC-OCV curves at different set temperatures, obtain the OCV=f(SOC, T) three-dimensional surface by fitting or interpolation to obtain the OCV model; Under the set SOC condition, the corresponding entropy thermal coefficient is determined as: Where T is the average surface temperature of the battery; S13. Determine the response curve of the battery terminal voltage and temperature over time under the set working conditions of the lithium-ion battery.

4. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 3, characterized in that: The specific process of step S11 is as follows: Place the lithium-ion battery in a constant temperature box and wait for 5 hours after the temperature in the constant temperature box reaches the target temperature to ensure that the battery reaches a thermodynamic equilibrium state; The battery was charged at a constant current and constant voltage of 1C CC-CV to the cut-off voltage and cut-off current, and the battery was discharged at a constant current of 1C to a certain SOC, and the open circuit voltage close to equilibrium was measured after being left for 2 hours. Repeat the above steps to obtain the SOC-OCV curve of the lithium-ion battery at a specific temperature.

5. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 3, characterized in that: The step S2 specifically includes the following steps: S21, respectively establishing a first-order equivalent circuit model for describing the electrical characteristics of lithium ions and a thermal characteristic model of lithium ion batteries; S22. Establish a single-layer feedforward network model with multiple inputs and outputs, couple the first-order equivalent circuit model and the thermal characteristic model, and obtain an electrothermal coupling model based on a neural network.

6. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 5, characterized in that: The first-order equivalent circuit model in step S21 is specifically: Among them, V 1，k is the voltage across the RC network of the battery at time k, V 0，k is the voltage across the resistor R0 at time k, Δt is the time interval from time k-1 to time k, OCV k is the open circuit voltage of the battery at time k, I k is the current flowing through the external circuit at time k, R 0，k , R 1，k , C 1，k represents the values ​​of the resistors R0, R1 and capacitor C1 at time k, τ is the time constant, τ k Indicates the value of the time constant at time k, which stipulates that the battery charging current is positive and the discharge current is negative; The thermal characteristic model of the lithium-ion battery is specifically a lumped thermal model, and the mathematical formula describing the heat generation rate of the battery is: Where Q is the heat generation rate, I is the current flowing through the battery, OCV is the open circuit voltage of the battery, V t is the battery terminal voltage, T is the average surface temperature of the battery, the first term on the right side of the equation is the irreversible heat during the operation of the battery cell, and the second term is the reversible heat during the operation of the battery cell; Assuming the battery surface temperature is uniform, then: Where, m is the mass of the battery cell, C p is the total heat capacity of the cell, A is the surface area of ​​the cell, h is the heat transfer coefficient, and the thermal parameter C p and h are both identified using cell measurement data; After discretization, we get: Among them, T amb is the ambient temperature.

7. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 6, characterized in that: The step S22 specifically includes the following steps: Through a single-layer feedforward network with multiple inputs and multiple outputs, a single-layer feedforward network model of three circuit parameters R0, R1, and C1 relative to temperature and SOC is constructed; The electrical characteristic model calculates the SOC and the battery terminal voltage V based on the applied current I and circuit parameters. t , where the circuit parameters include OCV, R0, R1, and C1. OCV is calculated by the open circuit voltage model obtained by experimental measurement, and R0, R1, and C1 are calculated by the neural network model based on SOC and temperature T; Set OCV, I, V t Used to calculate the irreversible heat generation rate; the entropy heat coefficient T and I are used to calculate the reversible heating rate; The reversible heat generation rate and the irreversible heat generation rate are then added together and used in the thermal model to calculate the temperature at the next moment.

8. The method for automatically constructing an electric-thermal coupling model of a lithium-ion battery according to claim 7, characterized in that: The single-layer feedforward network contains l hidden layer nodes, specifically: g(a,b,x)=1 / (1+exp(-(a·x+b))) Where x∈R n , y∈R p are input and output vectors respectively, the input vector x is [SOC, T] T , the output vector is [R0,R1,C1] T , R 0,k , R 1,1 , C 1,k is the circuit parameter at time k, a j ∈R n 、b i ∈R are the input weight and bias of the jth neuron, β ij is the output weight of the jth neuron relative to the i-th component of the output vector, g is the activation function, the weight a and bias b are randomly generated, and the output weight β is obtained by parameter identification.

9. The method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to claim 8, characterized in that: The specific process of step S3 is: The weighted normalized sum of squares error is defined as: Among them, y k and Represent the measured and predicted values ​​of the voltage and temperature at the road end, respectively; Since the circuit parameters R0, R1, and C1 have clear physical meanings, their non-negativity needs to be guaranteed. The Latin hypercube sampling method is used to generate random samples within the SOC and temperature range, and the nonlinear semi-infinite programming problem is transformed into a nonlinear programming problem with finite constraints. The following constraints are added in the parameter identification: Among them, SOC p , T p are the pth SOC and temperature samples, respectively, and P is the specified number of samples generated by the Latin hypercube sampling method; Taking the weighted normalized sum of squares as the optimization target, the optimization problem is solved and the parameter identification process is completed. Considering that with the increase of hidden neurons in the neural network, the fitting ability can be improved, but the prediction performance may deteriorate and the computational cost will also become higher, therefore, in order to study the influence of the number of hidden neurons on the fitting and prediction performance of the electrothermal coupling model based on artificial neural network, a penalty objective function containing the l1 norm is defined to achieve the optimal pruning of hidden neurons: Therefore, the loss function containing the l1 penalty norm is taken as the optimization target to solve the above optimization problem and realize the parameter identification and automatic model construction process.

10. A method for automatically constructing an electrothermal coupling model of a lithium-ion battery according to any one of claims 1 to 9, characterized in that: The step S4 specifically involves building a lithium-ion battery electric-thermal coupling simulation model in MATLAB or PYTHON, and performing corresponding setting working condition simulation.

Images