Method for accelerating charging dynamics through self-driven optimization porous structure

Through the pore network model and automatic optimization framework, the problems of slow charge transfer rate and model simplification deviation during the charging process of traditional porous electrodes are solved, and accurate optimization of porous electrodes and charging efficiency are achieved.

CN119089727APending Publication Date: 2024-12-06EAST CHINA UNIV OF SCI & TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411010869.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-02-22
Filing Date
2024-07-26
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

During the charging process, traditional porous electrodes have slow charge transfer rates due to the limitation of pore structure. The existing methods have deviations in simplifying models, which are costly and time-consuming, making it difficult to accurately describe and optimize disordered porous structures.

Method used

The pore network model combined with the automatic optimization framework is used to reconstruct the three-dimensional structure of the pore electrode through simulated annealing algorithm, extract the pore network topology, build the pore network model of the supercapacitor, and optimize the structural parameters using the genetic algorithm to convert them into an equivalent circuit model for calculation. Relaxation time to achieve acceleration of charging dynamics.

Benefits of technology

Accurate description and optimization of the charging dynamics of porous electrodes is achieved, the optimal structural parameters are quickly found, the experimental data dependence is reduced, the charging efficiency is improved, and a general image-structure-performance-optimization platform is provided.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089727A_ABST
    Figure CN119089727A_ABST
Patent Text Reader

Abstract

The invention discloses a method for accelerating charging dynamics through self-driven optimization of a porous structure, the method uses a pore network model to describe the porous structure and accelerates the charging dynamics process through an automatic optimization framework, and the method comprises the following steps: step S1, using a simulated annealing algorithm to reconstruct a three-dimensional structure of a porous electrode from an SEM graph; s2, extracting a topological structure of a pore network from the porous electrode by adopting a maximum sphere algorithm; s3, constructing a pore network model of the supercapacitor; s4, converting the pore network model into an equivalent circuit model, and deriving expressions of potential relaxation time tau phi and concentration relaxation time tau c; and step S5, obtaining the minimum relaxation time. By adopting the method, the problems of disordered porous structure description accuracy, model convertibility, rapid relaxation time calculation and direct input of experimental measurement structure parameters can be solved at the same time, and the charging dynamics in the porous carbon-based electrode can be described more accurately and optimized more comprehensively.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of batteries, and in particular to a method for accelerating charging dynamics by self-driven optimization of porous structures. Background Art

[0002] Porous electrodes are one of the most important components of some chemical power sources and also play a key role in the development of high-performance supercapacitors. A porous electrode refers to a solid electrode with many pores in it, and the pores must have a certain degree of connectivity. Under certain external conditions, fluids can flow in the pores. The porous electrode structure can usually be divided into two major parts: the solid electrode skeleton and the pores. The flow of electrolyte inside the porous electrode can sometimes be regarded as a multiphase system in which the fluid and the pore structure are coupled together.

[0003] Supercapacitors, also known as double-layer capacitors, are a better energy storage element between traditional physical capacitors and batteries. They have the advantages of both the high specific energy of chemical batteries and the high specific power of traditional physical capacitors. The basic principle of double-layer capacitors is to use charged porous electrode materials to adsorb ions in the electrolyte. The microporous surfaces of the positive and negative porous electrode materials adsorb ions of opposite charges respectively, forming a double layer at the electrode / electrolyte interface to store energy. Compared with traditional batteries, supercapacitors have the characteristics of high power density, long cycle life, fast charging and discharging, and a wide temperature range.

[0004] In traditional porous electrodes, the pore structure limits the charge transfer rate because the charge diffusion path in the pores is long, resulting in a slow charging process. A common method to accelerate charging is to increase the charge transfer rate by regulating the pore structure, such as controlling the pore size, distribution, and shape. By reducing the pore size, the diffusion path of the charge in the pores can be shortened, thereby speeding up the charging speed. In addition, the connectivity of the pores can be increased to improve the efficiency of charge transfer in the electrode. Another method is to change the material of the porous electrode to improve the conductivity and electrochemical activity. For example, choosing a material with good conductivity as the electrode substrate can reduce the impedance of charge transfer and increase the charging speed. At the same time, choosing a material with high electrochemical activity as the electrode can increase the reaction rate between the charge and the electrolyte, further accelerating the charging process.

[0005] Traditional methods often use a certain degree of simplification in the treatment of porous electrode kinetic processes, generally simplifying them into two-dimensional models or one-dimensional models, and the construction of their porous structures often cannot truly reflect the real structural information. Therefore, it will bring large deviations. Before the emergence of the pore network model, traditional physical experimental methods or empirical empirical models were usually used to study the pore structure and transport characteristics of the medium. For physical means, physical experimental means (such as permeation experiments, mercury injection experiments, etc.) are used to determine the pore structure parameters of the medium and observe the transport behavior of the fluid in the medium. The experimental process of this method is relatively difficult, and the cost is high and time-consuming, which is not conducive to efficient research and promotion. Previous theoretical research mainly used empirical models and numerical simulations, using empirical models of known medium pore structure and transport characteristics to estimate or predict new media. These models may be based on empirical formulas, empirical parameters or analogy methods. Or use numerical calculation methods (such as finite element method, finite difference method, etc.) to simulate and calculate the pore structure and fluid transport behavior of the medium to obtain the pore network information of the medium. Both methods usually face the problem of excessive deviation due to their oversimplification. In addition, empirical models are difficult to generalize to new systems, and traditional finite element calculation methods take a long time or are difficult to converge under complex geometric and boundary conditions. Summary of the invention

[0006] In order to overcome the above technical defects, the object of the present invention is to provide a method for accelerating charging dynamics by self-driven optimization of porous structures.

[0007] To this end, the present invention provides a method for accelerating charging kinetics by self-driven optimization of porous structures, using a pore network model to describe the porous structure and accelerating the charging kinetics process through an automatic optimization framework, comprising the following steps:

[0008] Step S1, using a simulated annealing algorithm to reconstruct the three-dimensional structure of the porous electrode from a scanning electron microscope (SEM) image; that is, using SEM to scan the porous electrode to obtain a reference image; using image segmentation technology, identifying the solid phase and the pore phase in the reference image; generating a random initial 3D voxel point cloud with the same porosity as the reference image; applying a nested multi-resolution hierarchical simulated annealing algorithm, by changing the phase of the voxel (solid phase or pore phase), randomly reconstructing the point cloud so that its features gradually approach the reference image; converting the reconstructed 3D voxel point cloud into volume data with a pore phase and a solid phase.

[0009] Step S2, using the maximum sphere algorithm to extract the topological structure of the pore network from the porous electrode; extracting the representative elementary volume (REV) of the pore phase; selecting any voxel point in REV and searching for adjacent voxel points in a specific direction around it; using the shrinkage algorithm to find the upper and lower limits of the maximum sphere and its radius according to the formation range of the voxel, and determine the size and position of the maximum sphere; continuing to search for the next maximum sphere according to the above method; classifying all maximum spheres as pores or throats.

[0010] Step S3, based on the extracted pore network, construct a pore network model of the supercapacitor; according to the extracted pore network and the design requirements of the supercapacitor, determine the pore geometry and size distribution, introduce electrode and electrolyte materials, and determine the interface characteristic parameters to complete the construction of the pore network model of the supercapacitor. The porous electrode is regarded as an ordered or disordered pore network composed of many different spherical pores and cylindrical throats. In the present invention, the ball-and-stick model and the cube-cuboid model are used to characterize the pore network of the porous electrode and the bulk phase, respectively.

[0011] Step S4, convert the pore network model into an equivalent circuit model and derive the potential relaxation time τ φ and concentration relaxation time τ c The local electrostatic potential φ(x,t) and the ion concentration c(x,t) are modeled by the simplified Poisson-Nernst-Planck (PNP) equation. The simplified equation assumes neutral charge in the pore, monovalent salt solution, and the same diffusion coefficients for anions and cations.

[0012] Step S5, combining the genetic algorithm with the derived relaxation time expression, and automatically searching for the minimum relaxation time by optimizing the structural parameters of the porous structure. The optimization process is as follows:

[0013] Initialize individuals (i.e., one way of pore arrangement) and populations (i.e., multiple ways of pore arrangement) and encode them; before the termination condition is met, perform fitness calculation, parent selection, crossover and mutation operations on the decoded individuals; after termination, select the individual with the best fitness (i.e., the optimal way of pore arrangement) as the optimal solution.

[0014] Furthermore, by simplifying the topological structure and geometry of the pore network model, it is converted into a transmission line model and a stacked electrode model to improve the computational efficiency and applicability of the model while maintaining an accurate description of the disordered porous structure.

[0015] Furthermore, in step S4, a differential equation is established to describe the change law of charge, potential or concentration, and the local electrostatic potential φ(x, t) and ion concentration c(x, t) are modeled. Assuming that the pores are electrically neutral, monovalent salt solutions, and anions and cations have the same diffusion coefficients, the corresponding potential φ in pore i is i (t) and ion concentration c i (t) The formula is as follows:

[0016] Porous Electrode

[0017]

[0018] Body phase

[0019]

[0020] Among them, C i is the capacitance of aperture i, V i is the volume of pore i, Z i is the connectivity of pore i, z count is the ion valence, F is the Faraday constant, is the diffusion conductance between pore i and pore j, is the ionic conductance between pore i and pore j; capacitance C i and conductivity The calculation formula is:

[0021]

[0022] Where tr is d or Ω, S i is the surface area of ​​pore i minus the cross-sectional area of ​​throat ij, ε0 is the vacuum dielectric constant, ε r is the relative dielectric constant of the solution, is the Debye length of pore i, D is the diffusion coefficient of the ions, A i is the cross-sectional area of ​​pore i or throat ij, l i is the length of pore i or throat ij, R is the ideal gas constant, and T is the temperature.

[0023] Furthermore, in step S4, the initial ion concentration is uniform, and a positive potential +Ψ and a negative potential -Ψ are applied to the cathode and the anode, respectively, and the initial conditions are:

[0024] Ic i (t=0)=c0

[0025] φ cathode (t=0)=+Ψ=vV m

[0026] φ anode (t=0)=-Ψ=-vV m ,

[0027] Where c0 is the initial ion concentration, v is the scaling factor for the thermovoltage, V m =(k B T) / e is the thermal voltage, where k B is the Boltzmann constant and e is the elementary charge.

[0028] Furthermore, τ c The calculation is to expand the equivalent circuit model by changing the physical meaning of capacitors and resistors, where the capacitance is equal to the volume of the pores and the resistance is equal to the inverse of the diffusion conductance of the conduits.

[0029] Further, in step S4, based on the equivalent circuit model, the relaxation time τ is used to describe the time for the potential and concentration to reach equilibrium. The charging process of the supercapacitor uses the potential relaxation time τ φ and concentration relaxation time τ c To describe, τ φ and τ c They are the matrices M (n×n) and N (n×n) The second smallest eigenvalue λ min2 , the formula is as follows:

[0030]

[0031] Among them, min2() is the function of finding the second minimum value, and eig() is the function of finding the eigenvalue of the matrix. -1 is the operator for finding the inverse of a matrix, C (n×n) and V (n×n) are the diagonal matrices corresponding to the capacitance C and volume V weights of each pore, and is the ionic conductivity G per throat Ω and the diffusion conductivity G d The Laplacian matrix of weights, M (n×n) and N (n×n) All heterogeneous structural parameters of the pore network are included, including the location, connectivity, and geometric parameters of pores and throats, as well as the properties of the electrolyte, including the initial concentration and diffusion coefficients of ions.

[0032] Furthermore, in step S4, the relaxation time is numerically calculated in the pore network model to verify the results of the analytical solution. According to the pure exponential charge accumulation, the time-dependent function τ in pore i is defined as φ,i (t) and τ c,i (t), the calculation formula is as follows:

[0033]

[0034] τ φ,sPNP (t) = max(τ φ,i |τ φ,i (t) = t)

[0035] τ c,sPNP (t) = max(τ c,i |τ c,i (t) = t)

[0036] The subscripts eq and 0 represent the equilibrium state and the initial state, respectively, and τ i Equal to τ i The relaxation time of the system is defined as the maximum τ among all pores. φ,i and τ c,i The value of .

[0037] Furthermore, in step S5, the genetic algorithm evolves based on the population of candidate solutions under the influence of natural diversity and selective operations, without considering whether the objective function is differentiable, and directly performs iterative calculations.

[0038] Furthermore, this method uses a combination of multiscale modeling, numerical simulation and experimental verification to find the optimal structural parameters of the electrode, including pore size distribution, connectivity distribution, porosity and the properties of the electrolyte.

[0039] Compared with the prior art, the above technical solution has the following beneficial effects:

[0040] 1. It can simultaneously solve the problems of disordered porous structure description accuracy, model convertibility (i.e., whether the "basic structure" of the model is the same), fast calculation of relaxation time, and direct input of experimentally measured structural parameters. Under the condition of taking into account both accuracy and solution cost, the charging dynamics in porous carbon-based electrodes can be more accurately described and more comprehensively optimized.

[0041] 2. The self-driven optimization framework proposed in the present invention couples the relaxation time expression with the genetic algorithm, which can quickly find the optimal structural parameters, such as pore size distribution, connectivity distribution and porosity, and even the properties of the electrolyte. It does not require a large amount of experimental data to achieve the optimization of the porous electrode, thereby accelerating the charging dynamics of the porous carbon-based electrodes.

[0042] 3. This invention provides a general image-structure-performance-optimization platform for understanding and accelerating charging dynamics in porous electrodes. In the future, more accurate governing equations can be implemented in PNM, such as the modified Poisson-Nernst-Planck (MPNP) equation that takes into account ion size, to achieve more realistic simulations in systems with smaller pore sizes. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flow chart of the method for accelerating charging kinetics by self-driven optimization of porous structures according to the present invention;

[0044] Figure 2 Framework for self-driven optimization of charging dynamics of porous electrodes.

[0045] Figure 3 Graph of relaxation time in real porous electrodes. (a) SEM image of activated carbon in the experiment and the corresponding binary segmentation result. The white area is the pore phase and the black area is the solid phase. (b) Reconstructed electrode image obtained by simulated annealing algorithm. (c) Pore network diagram of the pore phase extracted from (b). (d) Pore network model diagram of supercapacitor, whose electrode and diaphragm are activated carbon and glass microfiber (MFV5) respectively. (e) Connectivity distribution diagram of (c). (f) Experimental and analyzed τ φ Comparison chart of .

[0046] Figure 4 The theoretical model diagram of supercapacitor. (a) PNM diagram of supercapacitor. (b) Equivalent circuit model diagram (ECM) of PNM. (c) Transmission line (TL) model and its ECM. (d) Stacked electrode (SE) model and its ECM. (e) Effective medium approximation (EMA) and its ECM.

[0047] Figure 5 This is the equivalent circuit model diagram of PNM in a simple case.

[0048] Figure 6 is the relaxation process and relaxation time diagram. (a) is the relaxation process at {0, 0.1, 1, 10}τ φ The potential distribution snapshot at time (b) is at {0,0.1,1,10}τ c The concentration distribution snapshot at time τ φ and τ c are the relaxation time of potential and concentration, respectively. (c) and (e) are the analytical solutions τ obtained from the ECM derived from the PNM. ECM The numerical solution τ calculated with the simplified PNP equation sPNP (d) and (f) are the absolute value relative error distribution δ diagrams of ECM, δ=|(τ ECM ―τ sPNP ) / τ sPNP |. (g) and (h) are time cost comparison diagrams of the ECM derived from PNM and the simplified PNP equation.

[0049] Figure 7This is the self-driving optimization logic diagram, where Equation 1 is τ φ and τ c Two expressions of equations;

[0050] Figure 8 To minimize τ φ As the goal, the structure diagram of PNM is optimized through the self-driven optimization framework. Where (a) is the initial structure diagram. (b) is the optimized structure diagram, and the porosity is the same as (a). (c) is the minimum τ of each generation φ Fig. (d) is the pore size distribution diagram before and after optimization. DETAILED DESCRIPTION

[0051] The advantages of the present invention are further described below in conjunction with the accompanying drawings and specific embodiments. Those skilled in the art should understand that the following specific description is illustrative rather than restrictive, and should not be used to limit the scope of protection of the present invention.

[0052] The pore network model (PNM) can be used to optimize porous electrodes to accelerate the charging process. PNM regards the porous electrode as an ordered or disordered pore network consisting of many different spherical pores and cylindrical throats. By establishing a pore network model, the transport and diffusion of charges in porous electrodes can be simulated. By adjusting the pore structure parameters (such as pore size, distribution, and shape), the charge transfer path can be optimized to accelerate the charging process.

[0053] The PNM method divides the bulk phase into many cubes in the process of describing the porous electrode to reflect the complex connection between the bulk phase and the electrode pores. However, due to the complex structure, it is difficult to directly calculate the partial differential equations in PNM using numerical methods. This embodiment converts the pore network model into ECM to achieve the purpose of calculating the relaxation time in PNM faster without solving complex partial differential equations. In ECM, nodes can represent the bulk phase and pores of the porous electrode, and lines represent the connections between nodes. The charge relaxation response of the supercapacitor can be modeled by different combinations of capacitors and resistors. The capacitor represents the energy storage capacity of the pore, while the resistor represents the difficulty of ion transmission in the pore and bulk phase.

[0054] The present invention applies a genetic algorithm to optimize the structural parameters of porous electrodes, such as pore size, distribution and shape. Through encoding, crossover and mutation operations, the genetic algorithm generates a set of candidate solutions, i.e., different electrode structures. Then, by evaluating the performance indicators of each candidate solution, such as charge transfer rate and energy storage capacity, the candidate solution with higher fitness is selected. Self-driven optimization can be applied to the evolution process of the genetic algorithm. Through self-learning and adaptive mechanisms, the self-driven optimization dynamically adjusts the parameters and strategies of the genetic algorithm according to the performance of each candidate solution and the changes in the optimization objectives. By combining the genetic algorithm and the self-driven optimization, the global optimization of the porous electrode can be achieved. The genetic algorithm searches for a variety of electrode structures, while the self-driven optimization automatically adjusts the parameters and strategies of the algorithm according to the changes in performance indicators to adapt to different optimization objectives and constraints. This optimization method can improve the performance of porous electrodes and the efficiency of the charging process, providing better performance for applications such as supercapacitors.

[0055] like Figure 1 The flowchart of the method for accelerating charging kinetics by self-driving optimized porous structure provided in this embodiment is shown, comprising the following steps:

[0056] like Figure 2 The PNM has been applied to accurately model electrochemical devices such as lithium-ion batteries, flow batteries, and supercapacitors.

[0057] First, the porous electrode is scanned by SEM to obtain a reference image; the solid phase and pore phase in the reference image are identified by image segmentation technology; a random initial 3D voxel point cloud with the same porosity as the reference image is generated; a nested multi-resolution hierarchical simulated annealing algorithm is applied to randomly reconstruct the point cloud by changing the phase (solid phase or pore phase) of the voxel so that its features gradually approach the reference image; the reconstructed 3D voxel point cloud is converted into volume data with pore phase and solid phase. The parameters of the algorithm are set as follows: gridlevel, which determines how many grid resolutions are used in sequence for each entry, and its value is [3,2]; length, which specifies the length of the structural descriptor used to calculate a specific phase for each entry, and its value is [30,30]; nps, which is the number of exchanges used to determine the initial temperature T0, and its value is 30; nt, which specifies the maximum number of iterations used to calculate a given phase at each grid level, and its value is [1×10 9 ,1×10 9 ]; nstop, which is the critical value of the number of consecutive rejections of pixel exchange for a given phase, will stop the iteration when it is reached, and its value is 500; acc, which is the expected accuracy of the structural descriptor error at each lag distance, and its value is 1×10 -5; sgd, the size of the reconstructed grid, its value is [1024,1024,1024]; lam, the cooling factor, its value is 0.9; nrdir, the number of directions for calculating the structure descriptor, its value is 6. The code of the algorithm is taken from the relevant literature.

[0058] Figure 3 (a) and Figure 3 (b) SEM images of activated carbon in the experiment and the reconstructed activated carbon electrode, respectively. The reconstructed electrode retains the details of the activated carbon to the greatest extent.

[0059] Then, Figure 3 (c) shows the conversion of the reconstructed electrode into a disordered pore network using the maximum sphere algorithm. The extraction process is as follows: extract the REV of the pore phase; select any voxel point in the REV and search for adjacent voxel points in a specific direction around it; according to the formation range of the voxel, use the contraction algorithm to find the upper and lower limits of the maximum sphere and its radius, and determine the size and position of the maximum sphere; continue to search for the next maximum sphere according to the above method; classify all maximum spheres as pores or throats. The algorithm parameters are set as follows: axial connectivity: Neighborhood, that is, the number of neighbors, its value is 26; separation object: Method, that is, the method of separating objects, Skeleton–Aggressive is selected; Neighborhood, that is, the number of neighbors, its value is 26; Markerextent, its value is 2. This embodiment is implemented by the software Avizo.

[0060] Figure 3 (d) shows the construction of the PNM for the supercapacitor based on the extracted pore networks. The thickness of the electrode and separator are 0.38 mm and 0.42 mm, respectively. The pore networks of the cathode and anode are composed of 20 extracted pore networks, respectively, to meet the actual thickness of the electrode. In addition, periodic boundary conditions are applied along the yz direction to reflect the actual situation.

[0061] In this example, the local electrostatic potential φ(x, t) and ion concentration c(x, t) are modeled by the simplified Poisson-Nernst-Planck (PNP) equation in the setting of the control equation. The simplified equation assumes that the pores are electrically neutral, monovalent salt solutions, and anions and cations have the same diffusion coefficients. The PNM describes the corresponding potential φi(t) and ion concentration c in pore i. i (t) are as follows:

[0062] Porous Electrode

[0063]

[0064] Body phase

[0065]

[0066] Here C i is the capacitance of aperture i, V i is the volume of pore i, Z i is the connectivity of pore i, z count is the ion valence, F is the Faraday constant, is the diffusion conductance between pore i and pore j, is the ionic conductance between pore i and pore j.

[0067] The capacitance C in the equation i and conductivity (tr is d or Ω) is:

[0068]

[0069] Here S i is the surface area of ​​pore i minus the cross-sectional area of ​​throat ij, ε0 is the vacuum dielectric constant, ε r is the relative dielectric constant of the solution, is the Debye length of pore i, D is the diffusion coefficient of the ions, A i is the cross-sectional area of ​​pore i or throat ij, l i is the length of pore i or throat ij, R is the ideal gas constant, and T is the temperature.

[0070] In this example, the initial conditions are set such that the initial ion concentration is uniform, and a positive potential +Ψ and a negative potential -Ψ are applied to the cathode and anode, respectively:

[0071] Ic i (t=0)=c0

[0072] φ cathode (t=0)=+Ψ=vV m

[0073] φ anode (t=0)=-Ψ=-vV m

[0074] Here c0 is the initial ion concentration, v is the scaling factor for the thermovoltage, V m =(k B T) / e is the thermal voltage, where k B is the Boltzmann constant and e is the elementary charge.

[0075] This example uses the following default parameter set: + =-z - =1, T=298.15K, D=2×10-9 m 2 / s,ε r =71, c0=100 mol / m 3 Therefore, the initial Debye length λ D and thermal voltage V m Equal to 9.15×10 -10 m and 0.02526 V. The governing equations above are solved in the COMSOL Multiphysics finite element solver using the specified initial conditions and a time range of 10 -5 τ φ ~10 2 τ c .

[0076] In order to accurately describe the effect of PNM parameters on charging kinetics, especially relaxation time, the ECM of PNM is introduced in this example. Figure 4 (a) and Figure 4 (b) shows the PNM of the supercapacitor and its equivalent circuit model (ECM), respectively. Figure 4 As shown in (b), in the ECM, nodes can represent the bulk and pores of the porous electrode, and lines represent the connections between nodes. The charge relaxation response of the supercapacitor can be modeled by different combinations of capacitors and resistors. The capacitance represents the energy storage capacity of the pores, and the resistance represents the difficulty of ion transport in the pores and bulk.

[0077] Figure 5 A simple but representative PNM ECM is shown. The capacitance of the bulk phase tends to zero (i.e., C5→0), and the circuit of the bulk phase is simplified to a single resistor (i.e., G 2,3 =(1 / G 2,5 +1 / G 5,3 ) -1 ). The charge obtained by charging the capacitor is Q1(t)=C1Δφ1(t)), ... and Q4(t)=C4 Δ φ4(t)), where Δφ1(t), ... and Δφ4(t) are the time-varying voltage differences of the capacitors. The circuit satisfies the condition that the anode and cathode absorb equal amounts of charge, Q(t) = Q1(t) + Q2(t) = -Q3(t) - Q4(t). At the same time, Where i is the hole number. The entire circuit obeys Ohm's law and Kirchhoff's law.

[0078] Based on the PNM ECM, τ is obtained by analytical derivation. φ and τ c Two general expressions of τ φ and τ c They are the matrices M (n×n) and N (n×n)The second smallest eigenvalue λ min2 :

[0079]

[0080] Among them, min2() is the function of finding the second minimum value, and eig() is the function of finding the eigenvalue of the matrix. -1 is the operator for finding the inverse of a matrix. (n×n) and V (n×n) are the diagonal matrices corresponding to the capacitance C and volume V weights of each pore, respectively. and is the ionic conductivity G per throat Ω and the diffusion conductivity G d The Laplacian matrix of weights. M (n×n) and N (n×n) It contains all the heterogeneous structural parameters of the pore network, including the location, connectivity, and geometric parameters of the pores and throats, as well as some properties of the electrolyte, including the initial concentration and diffusion coefficient of the ions. This example uses the Python package SciPy to solve the second smallest eigenvalue of the matrix.

[0081] Figure 4 (c) and Figure 4 (d) shows that PNM can be down-converted into TL model and SE model by simplifying the topology and geometry.

[0082] The TL model simplifies the disordered porous structure into a cylindrical pore with a cross-sectional area A and a length L. Ions are transported along the axial direction of the cylindrical pore and adsorbed on the pore wall to form a double electrical layer. The τ derived from the TL model is φ The expression is:

[0083]

[0084] Where T RC is the charge relaxation time of the flat electrode, and d is the ion diameter.

[0085] The SE model simplifies the porous electrode into a stack of n infinitely large flat electrodes with a gap of average pore size h. The thickness of the porous electrode is H = (n-1)h and the length of the bulk phase is 2L. Ions can be transported through the flat plates or adsorbed on both sides of the plates. The τ derived from the SE model is φ,SE The expression is:

[0086]

[0087] like Figure 4As shown in (e), the topological structure of EMA is approximately replaced by the mathematical expression of a single pore to represent the disordered porous structure at a statistical level, considering only the effects of connectivity and pore size distribution.

[0088] Although these simplified models demonstrate the scalability and applicability of this method, they ignore the influence of disordered porous structure parameters on τφ and τc. Therefore, this case uses numerical solutions to ensure that τ φ and τ c The structural parameters studied include the pore size standard deviation σ (i.e., pore size disorder), the average electrode connectivity Z, and the pore arrangement (i.e., microscopic surface topology).

[0089] In this example, the relaxation time is numerically calculated in PNM to verify the results of the analytical solution. According to the purely exponential charge accumulation, the time-dependent function τ in pore i is defined as φ,i (t) and τ c,i (t):

[0090]

[0091] τ φ,sPNP (t) = max(τ φ,i |τ φ, i(t)=t)

[0092] τ c,sPNP (t) = max(τ c,i |τ c,i (t) = t)

[0093] Here the subscripts eq and 0 denote the equilibrium state and the initial state, respectively. i Equal to τ i The relaxation time of the system (i.e., τ φ and τ c ) is defined as the largest τ among all pores φ,i and τ c,i The value of .

[0094] The default structural parameters of the pore network in this example are set as follows. The topology is cubic. The spacing between nodes in the electrode and bulk phase is 100nm and 50nm respectively. The dimensions of the cathode and anode are both 5×5×5. The dimensions of the bulk phase are 10×10×10. The pore size follows a normal distribution (mean pore size is 50nm and standard deviation is 10nm). The average connectivity of the cathode and anode is 2.5. The random seed controls the order of pore arrangement and the default value is zero. All pore network models were generated using the Python package OpenPNM.

[0095] Figure 6 (a) and Figure 6 (b) shows four potential distribution snapshots {0, 0.1, 1, 10}τ φ and concentration distribution snapshots {0, 0.1, 1, 10}τ c , visualize the numerical results, 0.1τ φ The potential at 0.1τc and the concentration at 10τc are far from equilibrium, while φ The potential at and 10τ c The concentration at is in equilibrium. Therefore, τ φ and τ c It can represent the time scale for potential and concentration to reach equilibrium.

[0096] Figure 6 (c)~ Figure 6 (h) shows the analytical solution T derived from ECM ECM and the numerical solution τ calculated from the simplified PNP equation sPNP Difference in accuracy and time cost. Figure 6 (c) and Figure 6 (e) shows the horizontal axis is T ECM , the vertical axis is τ sPNP The distribution of points. These points are evenly distributed on both sides of y = x. Figure 6 (d) and Figure 6 As shown in (f), for τ φ and τ c , the absolute value of the relative error δ of most ECMs is less than 1, where δ=|(τ ECM ―τ sPNP ) / τ sPNP |. Therefore, τ can be calculated with high precision using the expression derived from ECM φ and τ c .like Figure 6 As shown in (g), the time cost of ECM is about 1000 times lower than that of PNP. Figure 6 (h) shows that ECM has an effect on τ φ and τ c Have the same solving efficiency.

[0097] Figure 3 (e) shows that the average connectivity Z of the reconstructed electrodes is 6.4, which satisfies (Key parameters affecting δ) Here δ is the relative error of ECM, δ = |(τ ECM ―τ sPNP ) / τ sPNP |, Therefore, the expression derived in this example is accurate enough to calculate τ φ .

[0098] Figure 3(f) shows the analytical equation for τ φ = 16.4 s, which is not much different from the relaxation time observed in the experimental data (12.5 s). Considering the simplicity of the governing equations, the difference is within a reasonable range. Therefore, this example can well describe the effect of the pore structure of porous carbon-based electrodes on the charging kinetics through PNM.

[0099] Figure 7 The part of Equation 1 combined with the genetic algorithm in the self-driving optimization framework is shown, where Equation 1 is τ φ and τ c The scaling step in the framework is to make the porosity the same as the initial porosity by scaling the pore size of each pore in the porous electrode.

[0100] The self-driven optimization framework built in this example combines the genetic algorithm with the derived relaxation time expression, and automatically searches for the minimum relaxation time by optimizing the structural parameters of the porous structure. The optimization process is as follows: initialize individuals (i.e., one way of pore arrangement) and populations (i.e., multiple ways of pore arrangement) and encode them; before the termination condition is met, perform fitness calculation, parent selection, crossover and mutation operations on the decoded individuals; after termination, select the individual with the best fitness (i.e., the optimal pore arrangement) as the optimal solution. The algorithm parameter settings are as follows: initial pore size distribution: d av , i.e., the average pore size, whose value is 50nm; σ, i.e., the standard deviation of the pore size, whose value is 15nm; Problem: var, i.e., the pore size of each pore in the electrode; lb, i.e., the lower limit of the pore size of each pore in the electrode, whose value is 20nm; ub, i.e., the upper limit of the pore size of each pore in the electrode, whose value is 70nm; Population: NIND, i.e., the initial number of individuals in the population, whose value is 50; Encoding, i.e., the encoding scheme, which is RI; Algorithm: type, i.e., the type of genetic algorithm, which is soea_EGA_templet; MAXGEN, i.e., the maximum number of iterations, whose value is 100. In this embodiment, the genetic algorithm is implemented by Python package Geatpy.

[0101] like Figure 8 The method shown is to minimize τ φ The structure diagram of PNM is optimized by the self-driven optimization framework, showing a simple example of the proposed framework, that is, by optimizing Figure 3 (c) The pore size distribution of the electrode is used to accelerate the charging kinetics (i.e., to minimize τ φ ). The initial structure and the optimized structure are shown in Figure 8 (a) and Figure 8(b). In this example, the porosity of the electrode does not change before and after optimization. The pore size distribution is controlled by changing the size of each pore in the electrode, rather than by changing the form of the distribution function. Figure 8 (c) shows τ φ As the number of iterations increases, τ φ From the initial value of 5.37×10 -2 s is reduced to the optimal value of 5.04×10 -2 s. Therefore, by optimizing the pore size distribution of the electrode, τ φ Finally, this example compares the pore size distribution before and after optimization, such as Figure 8 (d) The results show that appropriately increasing the proportion of micropores can accelerate the charging kinetics.

[0102] In summary, the proposed method introduces PNM to study the charging dynamics in porous carbon-based electrodes and accelerates the charging dynamics by optimizing the porous structure through self-driven optimization. The ECM of PNM is constructed to quantitatively describe the effect of disordered structure in PNM on the charging process and is further compared with the numerical calculation of simplified PNP equation. By using ECM, a simple mathematical expression is proposed to quantify τ φ and τ c , taking into account the effects of electrolyte properties and disordered electrode structure, the expressions are verified by numerical solutions and experimental data. In addition, this example proposes a self-driven optimization framework to accelerate the charging dynamics of porous carbon-based electrodes by coupling our derived expressions and genetic algorithms to optimize structural parameters. However, this theoretical approach can implement more accurate governing equations in PNMs, such as the modified Poisson-Nernst-Planck (MPNP) equation that considers ion size to achieve more realistic simulations in systems with smaller pore sizes.

[0103] It should be noted that the embodiments of the present invention have better practicability and do not impose any form of limitation on the present invention. Any technician familiar with the field may use the technical content disclosed above to change or modify it into an equivalent effective embodiment. However, any modification or equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for accelerating charging kinetics by self-driven optimization of porous structures, characterized in that: The pore network model is used to describe the porous structure, and the charging kinetics process is accelerated through an automatic optimization framework, including the following steps: Step S1, using a simulated annealing algorithm to reconstruct the three-dimensional structure of the porous electrode from the SEM image; Step S2, extracting the topological structure of the pore network from the porous electrode using the maximum sphere algorithm; Step S3, constructing a pore network model of the supercapacitor based on the extracted pore network; Step S4, convert the pore network model into an equivalent circuit model and derive the potential relaxation time τ φ and concentration relaxation time τ c Expression of Step S5, combining the genetic algorithm with the derived relaxation time expression, and automatically searching for the minimum relaxation time by optimizing the structural parameters of the porous structure.

2. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: By simplifying the topological structure and geometry of the pore network model and converting it into a transmission line model and a stacked electrode model, the computational efficiency and applicability of the model are improved while maintaining an accurate description of the disordered porous structure.

3. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: In step S4, a differential equation is established to describe the change law of charge, potential or concentration, and the local electrostatic potential φ(x, t) and ion concentration c(x, t) are modeled. Assuming that the pores are electrically neutral, monovalent salt solutions, and anions and cations have the same diffusion coefficients, the corresponding potential φ in pore i is i (t) and ion concentration c i (t) The formula is as follows: Porous Electrode Physical appearance Among them, C i is the capacitance of aperture i, V i is the volume of pore i, Z i is the connectivity of pore i, z count is the ion valence, F is the Faraday constant, is the diffusion conductance between pore i and pore j, is the ionic conductance between pore i and pore j; capacitance C i and conductivity The calculation formula is: Where tr is d or Ω, S i is the surface area of ​​pore i minus the cross-sectional area of ​​throat ij, ε0 is the vacuum dielectric constant, ε r is the relative dielectric constant of the solution, is the Debye length of pore i, D is the diffusion coefficient of the ions, A i is the cross-sectional area of ​​pore i or throat ij, l i is the length of pore i or throat ij, R is the ideal gas constant, and T is the temperature.

4. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: In step S4, the initial ion concentration is uniform, and a positive potential +Ψ and a negative potential -Ψ are applied to the cathode and the anode respectively, and the initial conditions are: I.C.c i (t=0)=c0 φ cathode (t=0)=+Ψ=vV m φ anode (t=0)=-Ψ=-vV m , Where c0 is the initial ion concentration, v is the scaling factor for the thermovoltage, V m =(k B T) / e is the thermal voltage, where k B is the Boltzmann constant and e is the elementary charge.

5. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: τ c The calculation is to expand the equivalent circuit model by changing the physical meaning of capacitors and resistors, where the capacitance is equal to the volume of the pores and the resistance is equal to the inverse of the diffusion conductance of the conduits.

6. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: In step S4, based on the equivalent circuit model, the relaxation time τ is used to describe the time for the potential and concentration to reach equilibrium. The charging process of the supercapacitor uses the potential relaxation time τ φ and concentration relaxation time τ c To describe, τ φ and τ c They are the matrices M (n×n) and N (n×n) The second smallest eigenvalue λ min2 , the formula is as follows: Among them, min2() is the function of finding the second minimum value, and eig() is the function of finding the eigenvalue of the matrix. —1 is the operator for finding the inverse of a matrix, C (n×n) and V (n×n) are the diagonal matrices corresponding to the capacitance C and volume V weights of each pore, and is the ionic conductivity G per throat Ω and the diffusion conductivity G d The Laplacian matrix of weights, M (n×n) and N (n×n) All heterogeneous structural parameters of the pore network are included, including the location, connectivity, and geometric parameters of pores and throats, as well as the properties of the electrolyte, including the initial concentration and diffusion coefficients of ions.

7. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: In step S4, the relaxation time is numerically calculated in the pore network model to verify the results of the analytical solution. According to the pure exponential charge accumulation, the time-dependent function τ in pore i is defined as φ,i (t) and τ c,i (t), the calculation formula is as follows: t φ,sPNP (t)=max(τ φ,i |t φ,i (t)=t) t c,sPNP (t)=max(τ c,i |t c,i (t)=t) The subscripts eq and 0 represent the equilibrium state and the initial state, respectively, and τ i Equal to τ i The relaxation time of the system is defined as the maximum τ among all pores. φ,i and τ c,i The value of .

8. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: The optimization process of optimizing the porous structure in step S5 is as follows: first, individuals and populations are initialized and encoded; then, fitness calculation, parent selection, crossover and mutation operations of the decoded individuals are performed before the termination condition is met; after termination, the individual with the best fitness is selected as the optimal solution.

9. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: In step S5, the genetic algorithm evolves based on the population of candidate solutions under the influence of natural diversity and selective operations, without considering whether the objective function is differentiable, and directly performs iterative calculations.

10. A method for accelerating charging kinetics by self-driven optimization of porous structures according to claim 1, characterized in that: This method uses a combination of multi-scale modeling, numerical simulation and experimental verification to find the optimal structural parameters of the electrode, which include pore size distribution, connectivity distribution, porosity and the properties of the electrolyte.

Citation Information

Cited By

  • Method for constructing nested network model of porous material

    CN120809024A