A method for optimizing the microchannel structure of a battery heat dissipation plate

Abstract

This paper discloses a method for optimizing the microchannel structure of a battery cooling plate. The plate comprises two longitudinally arranged collecting and distributing channels, multiple cooling channels transversely connected between the two channels, and coolant inlets and outlets at the ends of the channels, with the inlets and outlets arranged diagonally. The thermal performance of a parallel microchannel carbon fiber / epoxy resin composite plate was studied using CFD simulation. Furthermore, a response surface model (RSM) and genetic aggregation algorithm were used to optimize the channel structure for improved cooling efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of battery heat dissipation, and in particular to a method for optimizing the microchannel structure of a battery heat dissipation plate. Background Art

[0002] Electric vehicles (EVs) have experienced rapid growth in recent years. A major challenge in EV design is advanced battery packaging, encompassing cooling performance, crashworthiness, and lightweight design. For example, lithium-ion batteries must operate at temperatures below 40°C to maintain battery life. To this end, power batteries typically utilize air or liquid cooling. Meanwhile, crashworthiness and lightweighting are also key issues in packaging design. The former considers structural protection, such as impact safety, while the latter impacts the EV's range. Therefore, solutions are needed that simultaneously address the challenges of active cooling, structural protection, and lightweight design.

[0003] Carbon / epoxy composite materials have the characteristics of light weight and high strength. Carbon / epoxy composite materials with microchannel structure are gradually being used in battery cooling. Usually, the carbon / epoxy composite material is made into a rectangular plate-shaped heat sink and placed between rectangular batteries to dissipate heat, or the rectangular heat sink is curled into a cylindrical shape and placed on the outside of the cylindrical battery to dissipate heat from the battery. The heat sink usually has two longitudinally arranged collecting and dissipating flow channels and multiple cooling flow channels connected laterally between the two collecting and dissipating flow channels, forming a rectangular cooling flow channel network. The coolant inlet and outlet of the cooling flow channel network are respectively located at the diagonal corners of the cooling flow channel network. However, the heat dissipation effect of the heat sink will be affected by the resistance in the flow channel and the density of the flow channel. How to better plan the flow channel in the heat sink has become an urgent problem to be solved. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is: how to provide a microchannel structure optimization method for a battery heat sink that can better plan the channel structure within the heat sink and facilitate improved heat dissipation.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A method for optimizing the microchannel structure of a battery heat sink, wherein the battery heat sink includes two longitudinally arranged collecting and distributing channels and a plurality of cooling channels transversely connected between the two collecting and distributing channels. The collecting and distributing channels and the cooling channels form a rectangular area, and the connection between the two forms a node. The two collecting and distributing channels are respectively provided with a coolant inlet and a coolant outlet, and the coolant inlet and the coolant outlet are arranged along the diagonal lines of the rectangular area. The optimization steps are as follows:

[0007] S1. Build a solid model based on the dimensions of the battery heat sink and the initial channel structure, and use ANSYS to extract the channel structure model and perform meshing. Import the mesh model into Fluent to set simulation conditions and simulate and correct the battery heat sink's cooling performance.

[0008] S2. Establish an optimized mathematical model for the channel structure:

[0009]

[0010] Where: P L P is the distance between two nodes on the coolant inlet manifold facing the coolant outlet. R is the distance between two nodes on the collector where the coolant outlet is located, and k is the distance from the coolant outlet end to P on the same collector. R or P L The number of adjacent node segments between P1 and P k The order is from the end away from the coolant outlet to P on the distribution channel where the coolant outlet is located. R The distance between two adjacent nodes, P k+1 ~P n The order is from the coolant inlet to the P on the distribution channel where the coolant inlet is located. L The distance between two adjacent nodes, T max is the maximum temperature, which is used as the target output; d min is the minimum distance between two adjacent nodes, d max is the maximum distance between two adjacent nodes; leng is the maximum distance between the two outermost cooling channels in the longitudinal direction of the battery heat sink;

[0011] S3. Use response surface model to build the relationship between design variables and output responses, and use optimal space filling design as DOE model to optimize the design by randomly generating sampling points;

[0012] S4. Generate the most suitable output response T using genetic aggregation algorithm max RSM type: Genetic algorithm is used to generate populations of different response surfaces and solve them in parallel, and the best population is selected according to the fitness function of each response surface.

[0013] Furthermore, in step S4, the genetic algorithm solution step includes:

[0014] The response surface is represented as an ensemble using the weighted average of the different metamodels:

[0015]

[0016]

[0017] in, For the integrated prediction, is the prediction of the lth response surface, N m is the number of metamodels, w l is the weight factor of the lth response surface; The root mean square error (RMSE) of the DOE samples and the The root mean square error (PRESS) of the same design point of cross validation RMSE ) is used to estimate the optimal weight factor value and satisfies the following formula:

[0018]

[0019]

[0020]

[0021] Among them, x j is the jth design point, y(x j ) is x j The output parameter value at is the prediction of the lth response surface constructed without the jth design point, and N is the number of design points.

[0022] Furthermore, the maximum distance d between two adjacent nodes max Determine using the following steps:

[0023] Under the condition that multiple cooling channels are parallel to each other and arranged horizontally, the distance between two adjacent cooling channels is gradually increased or decreased, and the simulation model in step S1 is used to simulate the cooling performance of the cooling channels at different distances. The highest temperature T is selected. max The distance in the simulation model close to the set temperature and greater than the set temperature is taken as d max .

[0024] Furthermore, in step S1, when modeling, the diameter d of the microchannel with a rectangular cross section is converted into an equivalent microchannel with a circular cross section using the following formula: h :

[0025] d h =2ab / (a+b)

[0026] Where a and b are the length and width of the rectangular cross section, respectively.

[0027] In summary, the present invention has the advantages of being able to better plan the channel structure in the heat dissipation plate, thereby facilitating improved heat dissipation. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a schematic diagram of the structure of the battery heat sink.

[0029] Figure 2 Surface temperature results of the microchannel composite plate from experiments and CFD simulations.

[0030] Figure 3 Comparison diagram of simulated temperature on the top surface of composite plates with parallel microchannel structures of rectangular and circular cross sections.

[0031] Figure 4 Diagram of the parametric modeling of parallel microchannel structures in carbon / epoxy composite plates.

[0032] Figure 5 Comparison of the top surface temperature of the composite cooling plate with microchannel structures at different parallel distances.

[0033] Figure 6 T for RSM max Predicted values and actual values.

[0034] Figure 7 It is a local sensitivity analysis of design variables.

[0035] Figure 8 3D plot of the response surface.

[0036] Figure 9 Schematic diagram for optimizing the surface temperature of the microchannel network under cooling conditions. DETAILED DESCRIPTION

[0037] The present invention will be further described in detail below with reference to the embodiments.

[0038] In this embodiment, the battery heat dissipation plate is made of a carbon / epoxy composite plate with a panel size of 200mm×150mm×3mm, including two longitudinally arranged collecting and distributing channels, multiple cooling channels connected transversely between the two collecting and distributing channels, and a coolant inlet and a coolant outlet respectively arranged at the ends of the two collecting and distributing channels. The coolant inlet and the coolant outlet are arranged along the diagonal line, such as Figure 1 The parallel microchannels in the composite plate have a rectangular cross-section of 1.02 mm × 0.84 mm, the fiber volume fraction of the carbon / epoxy composite plate is 45%, and the coolant is 50 / 50 water-ethylene glycol.

[0039] 1. Simulation Modeling

[0040] The finite volume-based CFD software FLUENT is used to solve the control equations and related boundary conditions. The control equations for incompressible fluid flow are:

[0041] Continuity equation:

[0042]

[0043] Momentum equation:

[0044]

[0045]

[0046]

[0047] Energy equation:

[0048]

[0049] Among them, u, v, w are the velocity components in the x, y and z directions respectively, ρ, P c ,μ,c p 、k f and T are coolant density, coolant pressure, coolant dynamic viscosity coefficient, coolant specific heat capacity, coolant thermal conductivity, and coolant temperature, respectively.

[0050] The cooling performance of the composite plate was simulated using the commercial CFD software ANSYS Fluent. For the channel structure, 0.17 mm tetrahedral elements were used after mesh convergence testing. To better simulate the actual flow of the coolant, this example employed a finer mesh and implemented five boundary layers for the fluid closure walls. The mesh quality of the tetrahedral elements was controlled, such as no negative volume, a skewness less than 0.89, and an aspect ratio less than 3.0. The final mesh contained 3.703 million fluid elements and 8.912 million solid elements.

[0051] One surface of the panel was heated using a polyimide flexible heater with thermal grease, while the other surface was painted for thermal imaging. During the test, the panel surface temperature was recorded with an infrared thermal imager, the coolant inlet and outlet temperatures were measured with thermocouples embedded in the pipes, and the coolant pressure drop was measured with a pressure sensor.

[0052] A 500Wm -2 The steady-state heat flux simulates the actual experimental temperature field. According to the experimental setup, the top surface is open to convection and radiation, the emissivity is 0.97, and the convection coefficient is defined as h = 8Wm -2 K. The side wall of the panel is insulated, and the coolant is cooled at 28.2 ml min -1 A mass flow rate of circulates through the channel. The initial temperature at the inlet is set to 22°C and measured by an inlet thermocouple. A pressure boundary is applied at the outlet, with zero pressure. The material properties are listed in Table 1.

[0053] Table 1. Material properties

[0054]

[0055] According to the dimensions of the microchannel, the Reynolds number can be calculated as:

[0056] Re=ρv f d h / μ (6)

[0057] Where Re is the Reynolds number, ρ is the density of the coolant, and v f is the flow rate (m / s), μ is the dynamic viscosity, d h is the diameter of the circular microchannel. For rectangular channels, the equivalent diameter of the rectangular cross-section microchannel (d h ) can be converted by the following formula:

[0058] d h =2ab / (a+b) (7)

[0059] where a and b are the length and width of the rectangular cross section, respectively. In this case, the Reynolds number is Re = 149 (< 2300), so the flow is considered laminar.

[0060] Steady-state simulations were performed using the SIMPLE pressure-velocity coupling scheme, Green-Gauss node-based gradient discretization, second-order pressure discretization, third-order MUSCL momentum discretization, and third-order MUSCL energy discretization. This solution is particularly suitable for solving the conservation of mass, momentum, and energy for incompressible Newtonian fluids with laminar flow. Convergence was achieved after iteration until the velocity and continuity residuals reached 10 -4 , the energy residual reaches 10 -8 These thresholds are sufficient to make the temperature and pressure fields converge.

[0061] The average surface temperature T of the composite cooling plate av , maximum surface temperature T max , surface temperature standard deviation σ T and coolant temperature rise ΔT c The simulation results were verified by other methods. Figure 2 As shown in the figure, the simulated temperature distribution (right) is consistent with the experimental results (left), that is, hot spots are formed between the cooling channels, while low temperature areas appear along the channels. In addition, the surface temperature data such as T av , T max , σ T and ΔT c etc. were well predicted. av , T max , σ T and ΔT c All absolute errors are within 0.5°C.

[0062] As can be seen from the figure, there are obvious hot spots between the cooling channels, and the microchannel structure shows poor cooling performance, which means that the temperature distribution is obviously uneven. Therefore, it is necessary to optimize the design of the microchannel structure.

[0063] 2. Optimization design of microchannel structure

[0064] Before the optimization design, the microchannel with a rectangular cross section was simplified to a circular channel to save calculation time. The equivalent diameter was determined according to equation (7). For example, a 1.02 mm × 0.84 mm rectangular cross section is equivalent to a circular cross section with a diameter of 0.92 mm. The equivalence was verified by comparing the simulation temperature results, as shown in Figure 3 Excellent consistency was obtained in the temperature distribution, including temperature contours, T max , T av and σ T Therefore, the following optimized design of parallel networks is based on the equivalent channel structure.

[0065] In this embodiment, the geometry of the parallel microchannel structure is discretized by a parametric modeling scheme, e.g. Figure 4 The positions of the entrance, exit and four diagonal points remain fixed, and the distances between adjacent channels are different, i.e. Figure 4 P1~P14、P R and P L , a total of 16 design parameters are obtained, among which P1~P7 are the distances of the first 7 node segments set in sequence from the end away from the coolant outlet to the coolant outlet end on the collecting channel where the coolant outlet is located, P R is the distance of the 8th node segment in this direction; P8~P14 are the distances of the first 7 node segments arranged in sequence from one end of the coolant inlet to the other end on the collecting channel where the coolant inlet is located, P L is the distance of the 8th node segment in this direction. It should be noted that the sum of P1~P7 and P8~P14 is constrained to be less than 184.0mm, and P1~P7 and P R As well as P8~P14 and P L The sum of is constrained to be equal to 184 mm to keep the panel area constant during the optimization process. max Used as target output, representing the cooling performance of the cooling system.

[0066] In order to determine the design variable P i To find the upper bound of , it is necessary to simulate the model with different spacings of cooling channels to find the critical distance between adjacent channels where hot spots form. In this embodiment, five different values of P1, P2, P8 and P9 (such as 20mm, 30mm, 40mm, 50mm and 60mm) were tested, and the cooling performance of a set of parallel structures was obtained, such as Figure 5 As shown. It can be seen that the greater the distance between adjacent channels, the greater the possibility of generating hot spots, and T max The higher the distance, the more hot spots will appear. 14 The upper boundary of the design parameters is determined to be 40 mm. Specifically, according to the actual processing requirements of the microchannel structure, P i The lower boundary is 3mm.

[0067] Therefore, the mathematical model of the optimal design is obtained:

[0068]

[0069] Agent model and optimization method

[0070] A surrogate model was developed to model the relationship between design variables and output responses. Further optimization was performed based on the accurate surrogate model. A design of experiments (DOE) was performed to randomly generate sampling points for design optimization. The performance of the composite cooling plate was subsequently evaluated through CFD simulation.

[0071] In this work, a surrogate model is constructed from a response surface model (RSM) to improve its efficiency. Optimal space filling design (OSFD) is used as a DOE method, which allows users to specify multiple design points and is suitable for complex response surface methods.

[0072] Select the Genetic Aggregation (GA) algorithm to generate the most suitable output response T max The GA generates populations of different response surfaces and solves them in parallel, selecting the best population based on the fitness function of each response surface. The GA response surface is written as a set using the weighted average of different metamodels:

[0073]

[0074]

[0075] in, For the integrated prediction, is the prediction of the lth response surface, N m is the number of metamodels, w l is the weight factor of the lth response surface; The root mean square error (RMSE) of the DOE samples and the The root mean square error (PRESS) of the same design point of cross validation RMSE ) is used to estimate the optimal weight factor value and satisfies the following formula:

[0076]

[0077]

[0078]

[0079] Among them, x j is the jth design point, y(x j ) is x j The output parameter value at is the prediction of the lth response surface constructed without the jth design point, where N is the number of design points. The GA algorithm calculates the fitness of each individual in the initial population for the target expectation and passes the optimized individuals to the next generation, ultimately delivering the optimal solution through continuous comparison with the target value.

[0080] 3. Conclusion and Verification

[0081] Verification of RSM: Predicted T on cooling panel surface max Used to validate RSM. The variability of RSM responses was assessed using the MSE and coefficient of determination (R2) indices:

[0082] R 2 =SSR / SST=1-SSE / SST

[0083] Where SSE is the sum of squared errors, SSR is the sum of squared responses, and SST is the sum of squared sums. 2 The larger , the more variability is explained by the linear regression model.

[0084] DOE obtained a total of 784 design points and predicted T max The curve with the actual value is as follows Figure 6 As shown. Obtain the horizontal coordinate target T from CFD max , predicting the ordinate value from RSM. The predicted value shows a good linear relationship with the target value. The MSE and R² values are 0.1059 and 0.9890, respectively, demonstrating the high accuracy of RSM.

[0085] Based on RSM, local sensitivity analysis of design variables is plotted, such as Figure 7 As shown. Negative sensitivity value means T max As the independent design variables increase, the effect of T decreases, and vice versa. max P8, P9, P10 and P11 have different effects on T max is positively sensitive to T, while other design variables are max This indicates that the branch channel near the inlet is more conducive to cooling the plate to reduce T max And achieve a more uniform surface temperature. P1, P2, P3, P4 on Tmax It is negatively sensitive, that is, increasing P1, P2, P3, and P4 will significantly reduce T max .

[0086] Optimized channel structure: The relationship between design variables and T was further investigated by 3D plots of the RSM response. max The relationship between Figure 8 In this 3D diagram, T max As the z-axis, the two design variables controlling two adjacent channels are used as the x- and y-axes (such as P1 and P8, P2 and P9, P3 and P10, P5 and P12, P6 and P13, P7 and P14).

[0087] Figure 8 T in (a), (b) and (c) max The relationship between its corresponding x and y coordinates shows a consistent trend, that is, T max It decreases with the increase of P1, P2, and P3 and the decrease of P8, P9, and P10. Figure 8 (e) presents Figure 8 (f) The same pattern, i.e. T max It first decreases and then increases with the increase of P13 and P14, and continues to decrease with the increase of P6 and P7. In particular, T max As P5 and P12 increase, they first decrease and then increase. Figure 8 (d). In this case, the lowest T max Located in the middle of the x and y axes, this can be seen from T max It is clearly observed in the projection on the xy plane ( Figure 8 (d)).

[0088] Since P8, P9, and P10 are close to the entrance, and P6 and P7 are close to the exit ( Figure 4 ), response surface ( Figure 8 ) also shows that branch channels close to the inlet and outlet will help reduce T max The results are in good agreement with the local sensitivity analysis of the design variables mentioned above. For one inlet and outlet in this project, it is recommended to arrange them in an inclined linear distribution rather than in parallel in the channel structure design.

[0089] Solved by RSM and GA Figure 4 Three groups of optimized microchannel structures of the composite cooling plate are shown in Table 2.

[0090] Table 2. Optimization of microchannel structures by RSM and GA solutions

[0091]

[0092] In the table above, "Ori." represents the original parallel channel structure, while "OP1," "OP2," and "OP3" represent optimized channel structures. All three optimized channel structures achieved a Tmax close to 33°C, 5°C lower than the original Tmax (38°C).

[0093] The optimized microchannel structure and the surface temperature on the cooling composite plate are as follows Figure 9 As shown in the figure, after optimization, the original hot spots on the cooling panel have been significantly reduced. The temperature distribution on the optimized panel surface is more uniform than that of the original cooling panel. In the optimized channel structure, the inflow branch channel is close to the inlet, and the outflow branch is close to the outlet. This channel arrangement results in a more uniform flow rate in the cooling channel than in a parallel structure, and subsequently a more uniform temperature distribution on the cooling panel surface.

[0094] Table 3. Surface temperature data of optimized microchannel structure

[0095]

[0096] Table 3 lists the surface temperature data of the optimized microchannel structure. After optimization, T max , T av and σ T It can be seen that the cooling liquid temperature rise ΔT of the three optimized structures is c This is higher than before optimization, indicating that the coolant transfers more heat to the cooling plate surface. This shows that the optimized channel structure has higher cooling efficiency than the original channel.

[0097] Furthermore, the inlet pressure P of optimized structures (such as OP1) is lower than that of the original cooling plate. Since higher pressures require more energy to pump the coolant, requiring tighter sealing of the liquid cooling circulation system, stronger pipes, and higher pump performance, the inlet pressure should be limited to a lower value. In this way, the optimized microchannel structure also contributes to better practicality in current projects due to the lower inlet pressure.

[0098] The above description is only a preferred embodiment of the present invention and does not limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing the microchannel structure of a battery heat sink, wherein the battery heat sink comprises two longitudinally arranged collecting and distributing channels and a plurality of cooling channels transversely connected between the two collecting and distributing channels, wherein the collecting and distributing channels and the cooling channels form a rectangular area, and the connection between the two forms a node, and the two collecting and distributing channels are respectively provided with a coolant inlet and a coolant outlet, and the coolant inlet and the coolant outlet are arranged along the diagonal lines of the rectangular area; characterized in that: The optimization steps are as follows: S1. Build a solid model based on the dimensions of the battery heat sink and the initial channel structure, and use ANSYS to extract the channel structure model and perform meshing. Import the mesh model into Fluent to set simulation conditions and simulate and correct the battery heat sink's cooling performance. S2. Establish an optimized mathematical model for the channel structure: ， Where: is the distance between two nodes on the coolant inlet manifold facing the coolant outlet end, is the distance between two nodes on the manifold where the coolant outlet is located, From the end away from the coolant outlet on the same manifold or The number of adjacent node segments between The order is from the end away from the coolant outlet to the end of the distribution channel where the coolant outlet is located. The distance between two adjacent nodes, The order is from one end of the coolant inlet to the distribution channel where the coolant inlet is located. The distance between two adjacent nodes, is the maximum temperature, which is used as the target output; is the minimum distance between two adjacent nodes, is the maximum distance between two adjacent nodes; The maximum distance between the two outermost cooling channels of the battery heat sink in the longitudinal direction; The maximum distance between two adjacent nodes Determine using the following steps: Under the condition that multiple cooling channels are parallel to each other and arranged horizontally, the distance between two adjacent cooling channels is gradually increased or decreased, and the simulation model in step S1 is used to simulate the cooling performance of the cooling channels at different distances. The highest temperature is selected. The distance in the simulation model close to the set temperature and greater than the set temperature is taken as ; S3. Use response surface model to build the relationship between design variables and output responses, and use optimal space filling design as DOE model to optimize the design by randomly generating sampling points; S4. Generate the most suitable output response using genetic aggregation algorithm RSM type: Genetic algorithm is used to generate populations of different response surfaces and solve them in parallel, and the best population is selected according to the fitness function of each response surface.

2. The microchannel structure optimization method of a battery heat dissipation plate according to claim 1, characterized in that: In step S4, the genetic algorithm solution step includes: The response surface is represented as an ensemble using the weighted average of the different metamodels: ， ， in, For the integrated prediction, For the The prediction of a response surface, is the number of metamodels, For the The weighting factors of the response surfaces; The root mean square error (RMSE) of the DOE samples and the The root mean square error of the same design point for cross validation ( ) is used to estimate the optimal weight factor value and satisfies the following formula: ， ， ， in, For the Design points, yes The output parameter value at Is there no The first design point is constructed The predictions of a response surface are given, and N is the number of design points.

3. The method for optimizing the microchannel structure of a battery heat dissipation plate according to claim 1, wherein: In step S1, when modeling, the diameter of the microchannel with a rectangular cross section is converted into the equivalent diameter of the microchannel with a circular cross section using the following formula: : ， Where, and are the length and width of the rectangular cross section respectively.

Images