Wave-shaped lithium battery cold plate multi-objective optimization method based on deep neural network

Abstract

The invention belongs to but is not limited to the technical field of lithium battery thermal management, and discloses a multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network, and the method comprises the steps: building a parameterized three-dimensional geometric model of a sine wave channel cold plate; performing experimental design on the channel number, amplitude, wavelength, pipe diameter and flow by using an orthogonal method to obtain a parameter combination; solving the flow heat transfer process under different parameter combination conditions by utilizing flow heat transfer numerical simulation calculation, and analyzing the influence of each factor on the cooling effect and flow power consumption of the cold plate; utilizing a numerical simulation result to train a deep neural network to predict the cooling performance of the cold plate; and carrying out multi-objective optimization design on the cold plate by combining a deep neural network with a genetic optimization algorithm to obtain a Pareto leading edge and a TOPSIS decision point. According to the method, orthogonal experimental design, numerical simulation, a neural network and a genetic optimization algorithm are comprehensively used, so that the calculation cost is reduced, and the efficiency of cold plate heat dissipation flow heat transfer characteristic analysis and optimization design is improved.

Description

Technical Field

[0001] The present invention belongs to, but is not limited to, the technical field of lithium battery thermal management, and particularly relates to a multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network. Background Art

[0002] Lithium batteries have excellent performance such as high energy density, long cycle life, and low self-discharge rate, and have been widely used in fields such as transportation and energy storage, and have no CO 2 emissions, which helps to achieve the "dual carbon" goal. Since lithium batteries are sensitive to temperature and have a narrow operating temperature range, thermal management is particularly crucial. Especially under high energy density and fast charging requirements, the thermal management challenges in variable application scenarios are exacerbated. Currently, the main methods of lithium battery thermal management include air cooling, liquid cooling, phase change material cooling, thermoelectric refrigeration, and heat pipe cooling. Among them, the liquid cooling system has gradually become a research hotspot due to its high specific heat and high density, and excellent heat dissipation effect.

[0003] The cold plate is the core component of the liquid cooling system, and the channel structure is the key to determining the heat transfer performance of the coolant, which has a significant impact on the heat dissipation efficiency and flow energy consumption. Research shows that different cold plate structures significantly affect the temperature distribution and pressure drop. Ran et al. proposed a low-flow-resistance tree-shaped cold plate, which can reduce the battery temperature but increase the temperature difference; Salimi et al. designed a microchannel cold plate with variable wave amplitude, effectively improving the temperature uniformity. In recent years, various cold plate structures such as diamond-shaped channels, wavy structures, and vein-shaped channels have emerged continuously, and the temperature rise is reduced and the temperature uniformity is improved by optimizing the cold plate channel design. Especially in large-capacity battery systems such as ships, optimizing the cold plate structure helps to balance the cooling effect and cost.

[0004] For the optimization of different cold plate structures, researchers have gradually applied methods such as topology optimization, multi-objective functions, and surrogate models, significantly improving the heat dissipation ability of the cold plate. The multi-objective optimization strategy (such as the NSGA-II algorithm) has achieved good results in balancing temperature control and flow pressure drop. In recent years, the application of deep learning and neural networks in heat transfer problems has increased. Through the neural network model, complex modeling can be avoided, the non-linear relationship between multi-dimensional data can be accurately predicted, and it effectively helps the cold plate design.

[0005] In summary, there is an urgent need for a cold plate design method that combines advanced neural network and topology optimization technologies to meet the complex lithium battery thermal management requirements, especially while improving the heat dissipation performance, reducing energy consumption and flow resistance, and meeting the multi-objective requirements in engineering applications. Summary of the Invention

[0006] Aiming at the problems existing in the prior art, the present invention provides a multi-objective optimization method and system for a wavy lithium battery cold plate based on a deep neural network.

[0007] The present invention is implemented as follows. A multi-objective optimization method for a wavy lithium-ion battery cold plate based on a deep neural network includes the following steps:

[0008] Step 1: Establish a parametric three-dimensional geometric model of a sinusoidal wavy channel cold plate;

[0009] Step 2: Use the orthogonal method to conduct experimental design on five variable parameters, namely the number of channels, wave amplitude, wavelength, pipe diameter, and flow rate, to obtain parameter combinations;

[0010] Step 3: Use numerical simulation calculations of fluid flow and heat transfer to solve the fluid flow and heat transfer processes under different parameter combinations, and analyze the influence of each factor on the heat dissipation effect and flow power consumption of the cold plate;

[0011] Step 4: Use the numerical simulation results to train a deep neural network to predict the heat dissipation performance of the cold plate;

[0012] Step 5: Use the deep neural network combined with the genetic optimization algorithm to conduct multi-objective optimization design on the cold plate to obtain the Pareto front and TOPSIS decision points.

[0013] Furthermore, the width of the cold plate is 150 mm, the height is 200 mm, and the thickness is 3.5 mm; the number of flow channels in the cold plate is N, the cross-section is a circle with a radius of D, the channel shape is constructed based on the sine function, the wave amplitude is A, and the wavelength is λ; all channels are connected through the top and bottom square cavities; the inlet and outlet are rectangular cross-sections of 12 mm × 2 mm, and the channel is 30 mm high.

[0014] Furthermore, a fully connected neural network is established based on Keras. The input layer has six parameters, namely the discharge rate, the number of channels, the pipe diameter, the wave amplitude, the wavelength, and the flow rate, and the output layer is T max 、PEC and Three performance indicators, and set the number of hidden layer network nodes and the activation function.

[0015] Furthermore, an input-output data set is sorted out from the simulation calculation results, and then the data set is normalized. 10% of the data set is divided into the test set, and the remaining part is the training set; during the training stage, the mean square error is used as the loss function, and the training set is used to train the established neural network to update the weights of each node of the neural network.

[0016] Furthermore, the comprehensive performance index PEC of fluid flow and heat transfer is calculated by the following formula:

[0017]

[0018] Entropy production rate Is calculated by the following formula:

[0019]

[0020] Furthermore, the multi-objective optimization design method includes:

[0021] (1) Initialize the population

[0022] (2) Non-dominated sorting and crowding degree calculation

[0023] (3) Set Gen = 1

[0024] (4) Selection, crossover, and mutation

[0025] (5) Population merging (2N)

[0026] (6) Non-dominated sorting and crowding degree calculation

[0027] (7) Generate a new population

[0028] (8) Determine whether Gen is less than the set value. If it is less, return to step (4). If it is greater, output the Pareto front;

[0029] Among them, in steps (2) and (6), the trained neural network is used for non-dominated sorting and crowding degree calculation.

[0030] Another object of the present invention is to provide a multi-objective optimization system for a wavy lithium-ion battery cold plate based on a deep neural network, including:

[0031] A parametric three-dimensional geometry module that establishes a parametric three-dimensional geometry model of a sinusoidal wavy channel cold plate;

[0032] An orthogonal experimental design module that uses the orthogonal method to conduct experimental design on five variable parameters, namely the number of channels, wave amplitude, wavelength, pipe diameter, and flow rate, to obtain parameter combinations;

[0033] A simulation calculation and analysis module that uses numerical simulation of fluid flow and heat transfer to solve the fluid flow and heat transfer processes under different parameter combinations, and analyzes the influence of each factor on the heat dissipation effect and flow power consumption of the cold plate;

[0034] A neural network training module that uses the numerical simulation results to train a deep neural network to predict the heat dissipation performance of the cold plate;

[0035] A multi-objective optimization module that uses a deep neural network combined with a genetic optimization algorithm to conduct multi-objective optimization design on the cold plate to obtain the Pareto front and TOPSIS decision points.

[0036] Another object of the present invention is to provide a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the wavy lithium battery cold plate multi-objective optimization method based on a deep neural network.

[0037] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the wavy lithium battery cold plate multi-objective optimization method based on a deep neural network.

[0038] Another object of the present invention is to provide an information data processing terminal, which includes the wavy lithium battery cold plate multi-objective optimization system based on a deep neural network.

[0039] Combined with the above technical solutions and solved technical problems, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0040] First, the present invention establishes a three-dimensional model of a sinusoidal wavy channel cold plate structure for lithium-ion battery heat dissipation, conducts a numerical simulation study on its flow and heat transfer process, constructs a surrogate model using a deep neural network, and conducts a multi-objective optimization design in combination with the genetic algorithm NSGA-II. The advantages are as follows:

[0041] (1) The trained deep neural network can quickly and accurately predict T according to the input parameters. max 、PEC and The errors compared with the simulation results are within 5.0%, 5.0% and 10% respectively, indicating that the surrogate model established using the neural network can accurately predict the heat dissipation performance of the cold plate.

[0042] (2) The Pareto fronts with PEC maximization minimization, T max minimization and minimization as the objectives are obtained, and the specific parameters corresponding to the TOPSIS decision points are given, which can provide theoretical guidance for engineering design.

[0043] (3) The comprehensive use of orthogonal experimental design, numerical simulation, neural network and genetic optimization algorithm reduces the calculation cost and improves the efficiency of cold plate heat dissipation flow and heat transfer characteristic analysis and optimization design.

[0044] The present invention provides a new design method and multi-objective optimization design results for lithium battery thermal management and cold plate engineering design, which can provide theoretical guidance for engineering applications. Description of the Drawings

[0045] Figure 1It is the flowchart of the multi-objective optimization method for the wavy lithium-ion battery cold plate based on the deep neural network provided by the embodiment of the present invention;

[0046] Figure 2 It is the structural diagram of the multi-objective optimization system for the wavy lithium-ion battery cold plate based on the deep neural network provided by the embodiment of the present invention;

[0047] Figure 3 It is the schematic diagram of the geometric model provided by the embodiment of the present invention; (a) Schematic diagram of the battery pack structure; (b) Schematic diagram of the calculation area;

[0048] Figure 4 It is the result diagram of grid independence verification provided by the embodiment of the present invention;

[0049] Figure 5 It is the result diagram of time step independence verification provided by the embodiment of the present invention;

[0050] Figure 6 It is the comparison diagram of the simulation results and the contrast experiment results provided by the embodiment of the present invention;

[0051] Figure 7 It is the schematic diagram of the neural network provided by the embodiment of the present invention;

[0052] Figure 8 It is the T provided by the embodiment of the present invention max Comparison diagram of the predicted value and the simulation calculation value;

[0053] Figure 9 It is the comparison diagram of the PEC predicted value and the simulation calculation value provided by the embodiment of the present invention;

[0054] Figure 10 It is provided by the embodiment of the present invention Comparison diagram of the predicted value and the simulation calculation value;

[0055] Figure 11 It is the optimization design flowchart provided by the embodiment of the present invention;

[0056] Figure 12 It is provided by the embodiment of the present invention with PEC and As the target optimization result diagram;

[0057] Figure 13 It is provided by the embodiment of the present invention with T max And As the target optimization result diagram. Detailed implementation manners

[0058] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0059] As Figure 1 shown, a multi-objective optimization method for a wavy lithium-ion battery cold plate based on a deep neural network provided by an embodiment of the present invention includes the following steps:

[0060] Step 1, establish a parametric three-dimensional geometric model of a sinusoidal wavy channel cold plate;

[0061] Step 2, use the orthogonal method to conduct experimental design on 5 variable parameters including the number of channels, wave amplitude, wavelength, pipe diameter, and flow rate to obtain parameter combinations;

[0062] Step 3, use the numerical simulation of fluid flow and heat transfer to solve the fluid flow and heat transfer process under different parameter combinations, and analyze the influence of each factor on the heat dissipation effect and flow power consumption of the cold plate;

[0063] Step 4, use the numerical simulation results to train a deep neural network to predict the heat dissipation performance of the cold plate;

[0064] Step 5, use the deep neural network combined with the genetic optimization algorithm to conduct multi-objective optimization design on the cold plate to obtain the Pareto front and the TOPSIS decision point.

[0065] I. Simulation Model and Verification

[0066] 1 Parametric Geometric Model

[0067] The structure of the lithium-ion battery pack is as Figure 3 (a) shown, consisting of square lithium-ion batteries, a cold plate, and liquid inlet and outlet pipelines. The cold plate is arranged on the side of the battery. This structure has good symmetry. To reduce the simulation calculation cost, the inlets and outlets are simplified, and a half-thickness battery and a half-thickness cold plate structure as shown in Figure 3 (b) are selected for parametric modeling. The battery is a lithium iron phosphate square battery with geometric dimensions of 150 mm (width) × 200 mm (height) × 30 mm (thickness). The outer dimensions of the cold plate are the same as those of the battery, with a width of 150 mm, a height of 200 mm, and a thickness of 3.5 mm. The number of flow channels in the cold plate is N, the cross-section is a circle with a radius of D, the channel shape is constructed based on the sine function, the wave amplitude is A, and the wavelength is λ. All channels are connected through the top and bottom square cavities. The inlets and outlets are rectangular cross-sections of 12 mm × 2 mm and channels with a height of 30 mm.

[0068] 2 Control Equations and Definite Solution Conditions

[0069] To establish the control equations for the heat dissipation process of the wavy cold plate of the lithium-ion battery, the following simplifications are made:

[0070] (1) The internal structure of the battery is complex, but the heat transfer mainly occurs through heat conduction and has obvious anisotropic characteristics. The battery is simplified as a solid with anisotropic thermal conductivity and isotropic other physical properties.

[0071] (2) The cooling working fluid is in forced flow, and the influence of gravity on the heat transfer process of the working fluid in the cold plate is ignored.

[0072] (3) The flow of the cooling working fluid is laminar, and the heat generated by viscous dissipation is ignored.

[0073] (4) In the temperature range where the battery is located, the physical properties of the cooling working fluid and the cold plate metal material are less affected by temperature changes. It is assumed that the physical property parameters of the cooling working fluid and the cold plate metal material are constants.

[0074] Heat is generated during the charge and discharge process of the battery. The heat transfer is mainly through heat conduction, and the temperature field satisfies the three-dimensional heat conduction differential equation with an internal heat source, that is

[0075]

[0076] In Equation (1), the subscript b represents the battery, represents the heat generation rate per unit volume inside the battery. The battery heat generation rate includes the reversible heat of reaction and the irreversible Joule heat, which is expressed as

[0077]

[0078] In Equation (2), I / A represents the charging or discharging current, R / Ω represents the internal resistance, E / V represents the open-circuit voltage, V / V represents the battery potential, and dE / dT / V·K -1 represents the temperature coefficient coefficient depending on the state of charge (SOC) of the battery. All the above parameters change with time. For a specific battery, under a given charge and discharge rate, each parameter can be obtained through experimental measurement. Li et al.

[57] established an empirical expression based on the experimental measurement results and substituted it into Equation (2) to obtain the calculation formula (3) for the change of the battery heat generation rate with time. The specific numerical values of A 1 to A 7 are shown in Table 1.

[0079]

[0080] Table 1 Coefficients of the polynomial for the internal heat generation rate of lithium-ion batteries

[0081]

[0082] Heat transfer in the cold plate substrate relies on heat conduction and no heat is generated. The temperature field satisfies the three-dimensional heat conduction differential equation without an internal heat source, that is

[0083]

[0084] In formula (4), ρ w / kg·m -3 represents the density of the aluminum plate, represents the specific heat of the aluminum plate, k w / W·m -1 ·K -1 represents the thermal conductivity of the aluminum plate.

[0085] The flow and heat transfer process of the cooling working fluid in the channel satisfies mass conservation, momentum conservation, and energy conservation, as shown in formulas (5-7),

[0086]

[0087] In formulas (5-7), ρ / kg·m -3 represents density, μ / kg·m -1 ·s -1 represents the dynamic viscosity, c p / J·kg -1 ·K -1 represents the specific heat capacity, k / W·m -1 ·K -1 represents the thermal conductivity. represents the velocity vector of the coolant, T / K represents temperature, p / Pa represents pressure, t / s represents time, and the following table l represents the coolant.

[0088] The battery is equivalent to a solid with uniform composition and anisotropic thermal conductivity. The cold plate is made of aluminum, and the cooling working fluid is a 40% ethylene glycol aqueous solution. The main physical property parameters are shown in Table 2.

[0089] Table 2 Main physical property parameters

[0090]

[0091] At the initial moment, the temperatures of the battery and the cold plate are both 25°C. The mass flow rate and temperature of the cooling working fluid are given at the inlet of the cold plate; the outlet of the cold plate is set as a pressure outlet; both the left and right sides are set as symmetric boundaries. In this invention, the finite volume method is used to numerically solve the heat dissipation problem of the lithium-ion battery cold plate. The iterative method is selected as the SIMPLE algorithm, and the convergence residual of the energy equation is set to 10 -8 , and the convergence residuals of other equations are set to 10 -6 .

[0092] 3 Verification of independence from grid and time step

[0093] Select the combination of a wavy channel with N = 4, D = 1.5 mm, A = 4 mm, λ = 50 mm, and q = 0.4 g / s to conduct the grid and time step independence verification. Use the Fluent-meshing software to divide the established geometric model into polyhedral grids, divide boundary layer grids at the fluid-solid contact surface, and encrypt the fluid region. Figure 4 Shows the variation of the pressure difference between the inlet and outlet and the fluid outlet temperature with the number of grids. When the number of grids reaches 1.17×10 6 After that, the result change is small. Compared with 2.31×10 6 After that, the pressure difference between the inlet and outlet changes by 1.4%, and the fluid outlet temperature changes by 0.1%. To balance the calculation accuracy and calculation consumption, 1.17×10 6 grids are selected to carry out the subsequent calculations, and the same grid division parameters are maintained for other cases.

[0094] Figure 5 Shows the variation of the pressure difference between the inlet and outlet and the fluid outlet temperature with the time step. It can be found that after the time step is 1.0 s and the time step continues to decrease, the pressure difference between the inlet and outlet and the fluid outlet temperature almost remain unchanged. Therefore, the time step in the simulation is selected as 1.0 s.

[0095] 4 Accuracy verification

[0096] To verify the numerical model developed in this study, a geometric model identical to [ZHENG Y, CHE Y, HU X, et al. Thermal state monitoring of lithium-ion batteries: Progress, challenges, and opportunities [J]. Progress in Energy and Combustion Science, 2024, 100: 101120.] was established, and numerical simulation was carried out. Figure 6 Shows the comparison between the simulated results of the battery surface temperature and the measured results of the comparative experiment under different discharge rates. The results are in good agreement, and the maximum deviation is 0.45%, verifying the reliability of the numerical calculation method of the present invention.

[0097] II. Orthogonal experiment design

[0098] The wavy channel cold plate includes the number of channels N, wave amplitude A, wavelength λ, and pipe diameter D, plus the inlet mass flow rate qm, for a total of 5 factors. Each parameter takes 4 levels. The orthogonal experiment design can reasonably select representative combinations to achieve the analysis and research of multiple factors at a lower cost. Combining with the research problems of the present invention, the L16(45) orthogonal table is selected to carry out the experimental design, with a total of 16 combinations, as shown in Table 3.

[0099] Table 3 Orthogonal experimental design table

[0100]

[0101]

[0102] 3. Deep Neural Network Prediction Model

[0103] Based on Keras, we established Figure 7 The fully connected neural network shown in the figure has six parameters: discharge rate, number of channels, tube diameter, amplitude, wavelength and flow rate as input layer, and T as output layer. max , PEC and Three performance indicators are set, and the number of hidden layer network nodes and activation function are set. In the training phase, the mean square error is used as the loss function, and the training set is used to train the established neural network and update the weights of each node in the neural network.

[0104] 4. Genetic Algorithm Construction

[0105] Genetic optimization algorithms simulate biological evolution mechanisms and can achieve global random search. The second-generation non-dominated solution sorting genetic algorithm (NSGA-II) selects suitable individuals through crowding calculation and elite retention strategy, which can reduce the algorithm calculation complexity and effectively avoid falling into local convergence during the optimization process. It is widely used in multi-objective optimization problems.

[0106] 5. Data Processing

[0107] The cooling medium flows in the cold plate channel and takes away the heat to cool the battery. The cold plate flow heat transfer process can be analyzed by the comprehensive performance index of flow heat transfer PEC. Among them, PEC can be calculated by formula (8-11).

[0108]

[0109] Entropy production rate It is an important indicator for analyzing the irreversibility of the process. The calculation of entropy production rate is shown in formula (12-14):

[0110]

[0111] like Figure 2 As shown, the embodiment of the present invention provides a multi-objective optimization method of a wavy lithium battery cold plate based on a deep neural network, a multi-objective optimization system of a wavy lithium battery cold plate based on a deep neural network, comprising:

[0112] Parametric 3D geometry module, to establish the parametric 3D geometry model of the sinusoidal wave channel cold plate;

[0113] Orthogonal experimental design module, using the orthogonal method to conduct experimental design on five variable parameters, namely the number of channels, wave amplitude, wavelength, pipe diameter, and flow rate, to obtain parameter combinations;

[0114] Simulation calculation and analysis module, using numerical simulation of fluid flow and heat transfer to solve the fluid flow and heat transfer processes under different parameter combinations, and analyzing the influence of each factor on the heat dissipation effect and flow power consumption of the cold plate;

[0115] Neural network training module, using the numerical simulation results to train a deep neural network to predict the heat dissipation performance of the cold plate;

[0116] Multi-objective optimization module, using a deep neural network combined with a genetic optimization algorithm to conduct multi-objective optimization design on the cold plate, and obtaining the Pareto front and TOPSIS decision points.

[0117] Result analysis

[0118] Deep neural network prediction

[0119] An input-output data set is sorted out from the simulation calculation results, then the data set is normalized, and 10% of the data set is divided into the test set, and the remaining part is the training set.

[0120] Use the trained neural network to predict the results, Figures 8 - 10 The predicted values and the simulation calculated values of T max , PEC and are respectively given. The comparison of the predicted values and the simulation calculated values in the training set and the test set shows that the deviation between the predicted values and the simulation calculated values of T max and PEC is within 5%, the deviation between the predicted value and the simulation calculated value is within 10%, indicating that the trained neural network can better predict the three parameters of T max , PEC and .

[0121] Multi-objective optimization design

[0122] There are many influencing factors in the transient heat conduction of the solid region and the transient fluid flow and heat transfer process in the fluid region, and the numerical simulation solution is complex and time-consuming. To improve the optimization efficiency, the trained neural network is used to predict the key parameters during the optimization iteration process. The present invention comprehensively uses a genetic optimization algorithm and a deep neural network to carry out multi-objective optimization design of the geometric structure of the wavy cold plate, and the optimization process is as Figure 11 shown, and finally the Pareto front solution set is obtained.

[0123] Based on the Geatpy toolbox, this invention uses the NASGA-II optimization algorithm to carry out multi-objective optimization design. The constraint condition is the maximum battery temperature to ensure battery safety, and finally the Pareto front is obtained. Based on the Pareto front, the TOPSIS method is used with the data entropy value as the weight, and finally the decision point is given to provide theoretical guidance for practical applications.

[0124] Figure 12 The optimization results obtained with the maximum and minimum of PEC as the objectives are given. Since the number of channels takes discrete values, the curves in the figure have segmented situations. By comparing the Pareto fronts at different discharge rates, it can be found that as the discharge rate increases, the Pareto front moves in the increasing direction. This is mainly because as the discharge rate increases, the battery temperature rises, and the non-equilibrium temperature difference in the heat transfer process increases, resulting in an increase in the entropy production rate. In the case of 1C, the front is mainly distributed in the region with a smaller PEC value. In the case of 3C, the front is mainly distributed in the region with a larger PEC value, while in the case of 2C, the PEC value range is relatively wide. This is mainly because under the constraint that the maximum battery temperature is less than 60 °C, a larger cooling working fluid flow rate is required in the case of 3C. According to the previous analysis results, PEC increases with the increase of the flow rate. In the case of 1C, a small flow rate can ensure that the maximum battery temperature is less than 60 °C. Therefore, the Pareto front in the case of 1C mainly corresponds to a small flow rate, and thus the PEC value is smaller. The parameters corresponding to the TOPSIS decision point are shown in Table 4.

[0125] Figure 13 The optimization results obtained with the minimum of T max and minimum as the objectives are given. Since the number of channels takes discrete values, the Pareto front in the figure has segmented situations. Similarly, as the discharge rate increases, the battery temperature rises, the non-equilibrium heat transfer temperature difference increases, and the entropy production rate of the Pareto front increases. Therefore, the Pareto fronts at high discharge rates are concentrated in the region where both T max and values are larger, and the Pareto fronts at low discharge rates are concentrated in the region where both T max and values are smaller. The parameters corresponding to the TOPSIS decision point are shown in Table 4.

[0126] Table 4 TOPSIS decision point parameters

[0127]

[0128] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those of ordinary skill in the art can understand that the above-mentioned devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code is provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and their modules of the present invention can be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above hardware circuits and software such as firmware.

[0129] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network, characterized in that: The following steps are involved: Step 1, establishing a parametric three-dimensional geometric model of the sinusoidal wave channel cold plate; Step 2: Use the orthogonal method to conduct experimental design on five variable parameters, namely, the number of channels, amplitude, wavelength, pipe diameter, and flow rate, to obtain a parameter combination; Step 3: Use the numerical simulation calculation of flow heat transfer to solve the flow heat transfer process under different parameter combination conditions, and analyze the influence of various factors on the heat dissipation effect of the cold plate and the flow power consumption; Step 4: Use the numerical simulation results to train a deep neural network to predict the cooling performance of the cold plate; Step five, use deep neural network combined with genetic optimization algorithm to perform multi-objective optimization design of the cold plate to obtain the Pareto frontier and TOPSIS decision points.

2. The multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network according to claim 1, characterized in that: The cold plate is 150mm wide, 200mm high and 3.5mm thick; the number of flow channels in the cold plate is N, the cross-section is a circle with a radius of D, and the channel shape is constructed based on a sine function with an amplitude of A and a wavelength of λ; all channels are connected by top and bottom square cavities; the inlet and outlet are rectangular cross-sections of 12mm×2mm and a groove height of 30mm.

3. The multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network according to claim 1, characterized in that: A fully connected neural network is established based on Keras. The input layer is composed of six parameters: discharge rate, number of channels, tube diameter, amplitude, wavelength and flow rate. The output layer is T max , PEC and Three performance indicators are used, and the number of hidden layer network nodes and activation functions are set.

4. The multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network as claimed in claim 3, characterized in that: The input-output data set is obtained from the simulation calculation results, and then the data set is normalized, and 10% of the data set is divided into a test set, and the rest is a training set; in the training stage, the mean square error is used as the loss function, and the training set is used to train the established neural network and update the weights of each node of the neural network.

5. The multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network according to claim 1, characterized in that: The comprehensive performance index of flow heat transfer PEC is calculated by the following formula: Entropy production rate Calculated by the following formula:

6. The multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network according to claim 1, characterized in that: Multi-objective optimization design methods include: (1) Initialize the population (2) Non-dominated sorting and congestion degree calculation (3) Set Gen = 1 (4) Selection, crossover, and mutation (5) Population merging (2N) (6) Non-dominated sorting and congestion degree calculation (7) Generate a new population (8) Determine whether Gen is less than the set value. If so, return to step (4); if so, output the Pareto frontier. In steps (2) and (6), the trained neural network is used to calculate the non-dominated sorting and congestion degree.

7. A multi-objective optimization system for a wavy lithium battery cold plate based on a deep neural network according to any one of claims 1 to 6, characterized in that: include: Parametric 3D geometry module, to establish the parametric 3D geometry model of the sinusoidal wave channel cold plate; Orthogonal experimental design module, using orthogonal method to conduct experimental design on five variable parameters including number of channels, amplitude, wavelength, pipe diameter and flow rate to obtain parameter combinations; The simulation calculation and analysis module uses the numerical simulation calculation of flow heat transfer to solve the flow heat transfer process under different parameter combination conditions, and analyzes the influence of various factors on the heat dissipation effect of the cold plate and the flow power consumption; Neural network training module, which uses numerical simulation results to train deep neural networks to predict the cooling performance of cold plates; The multi-objective optimization module uses a deep neural network combined with a genetic optimization algorithm to perform multi-objective optimization design on the cold plate to obtain the Pareto frontier and TOPSIS decision points.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the multi-objective optimization method of a wavy lithium battery cold plate based on a deep neural network as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the multi-objective optimization method for a wavy lithium battery cold plate based on a deep neural network as described in any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal includes the multi-objective optimization system for the wavy lithium battery cold plate based on a deep neural network as described in claim 7.

Images