A method for predicting local polarization of a flow battery based on a multi-scale model

Abstract

The present invention discloses a method for predicting local polarization of a flow battery based on a multi-scale model, belonging to the field of flow batteries. The present invention constructs a battery-scale model and a pore-scale model. The battery-scale model can simulate the mass transfer process and electrochemical reaction of active substances at the centimeter-to-meter scale, with a large simulation range; while the pore-scale model can simulate the mass transfer process and electrochemical reaction of active substances at the nano-to-micron scale, and can reflect the geometric structure composed of carbon fibers, with high accuracy. The multi-scale model of the present invention learns the relationship between the pore-scale geometric conditions, the battery-scale inlet conditions and the local polarization situation at the pore scale through a machine learning method, realizing the connection between the battery-scale model and the pore-scale model. The present invention can reduce the prediction time of the local polarization situation and improve the accuracy of local polarization prediction.

Description

Technical Field

[0001] The present invention belongs to the field of flow batteries, and particularly relates to a method for predicting local polarization of a flow battery based on a multi-scale model. Background Art

[0002] Under the dual-carbon goal, with the large-scale development of renewable energy including wind energy and solar energy, in order to solve the contradiction between the supply side and the demand side, a safe, efficient, and low-cost energy storage technology capable of integrating renewable energy into the power grid is urgently needed. Among many energy storage technologies, due to characteristics such as modular design, rich raw materials, and flexible design, redox flow batteries are regarded as one of the most promising energy storage technologies. As a large-scale energy storage system, the safety, reliability, and high energy efficiency of redox flow batteries are the first factors to be considered during application. Most improvement measures are directed at different components of the flow battery, including electrodes, membranes, bipolar plates, and electrolytes. In addition, side reactions in the flow battery can generate gases, which may lead to safety problems and are also the key to the safe operation of the flow battery that needs to be solved.

[0003] In the field of flow batteries, all-vanadium flow batteries, zinc-nickel flow batteries, iron-chromium flow batteries, etc. are all aqueous batteries, in which the hydrogen evolution reaction and oxygen evolution reaction that occur will reduce the electrochemical reaction area of the porous electrode, resulting in a decrease in the energy efficiency of the battery. Long-term hydrogen evolution and oxygen evolution reactions may also corrode the electrodes and reduce the operating life of the battery. At high voltages, the side reactions of the battery will intensify, so a low cut-off voltage is commonly used during charge and discharge processes to suppress gas side reactions. However, due to the complexity of the electrode pore structure and the preferential flow path of the electrolyte, local over-polarization in the electrode may occur locally, and local over-polarization will lead to gas side reactions. In order to study the causes of local polarization, a pore-scale model developed based on X-ray tomography technology and the lattice Boltzmann method and a pore network model developed based on the pore structure simplified into a pore network structure have been successively proposed by scholars. Among them, the former is often used for electrode material development, electrode structure optimization, and two-phase flow research, and has high accuracy; while the latter is often used to explore the transfer and reaction process of active substances inside the electrode, with slightly lower accuracy but capable of saving a large amount of computing resources.

[0004] On the other hand, to solve the local polarization problem of flow batteries, flow management techniques are often used to suppress the concentration overpotential and improve the energy efficiency of the battery. The commonly used flow optimization strategies usually consider the contradiction between battery performance and pump power loss: high flow rate can strengthen the ion mass transfer process on the electrode surface, thereby reducing the concentration overpotential; however, the corresponding pump power will also increase, resulting in a decrease in system efficiency. Starting from this point, the flow optimization strategy has evolved from the traditional segmented strategy to the dynamic optimization strategy. Research shows that compared with the constant flow rate, the variable flow rate strategy can improve the overall system efficiency of the battery and reduce the local concentration overpotential. However, the current flow optimization strategy does not consider the gas side reactions that may be caused by local polarization. Formulating the strategy solely based on system efficiency may lead to gas side reactions in specific situations, thereby triggering safety problems.

[0005] In summary, in the models used to explore gas side reactions, scholars often use pore-scale models to analyze the structure of porous media and the influence of inlet conditions on local polarization. When controlling and regulating local polarization, flow optimization strategies at the battery scale are often mentioned. The mismatch in scale between the two makes it difficult to unify the research and adjustment measures for gas side reactions, thus causing a technical bottleneck in controlling local polarization and gas side reactions. To address this bottleneck, the key to the research is how to expand the pore-scale model to the battery scale including the entire electrode. Summary of the Invention

[0006] Aiming at the problem of the mismatch in the scale of battery models existing in the above-mentioned prior art, the present invention proposes a method for predicting local polarization of a flow battery based on a multi-scale model. This multi-scale model is applied to the field of renewable energy battery energy storage. For the problem of scale mismatch between the pore-scale model and the flow optimization at the battery scale existing during the operation of the flow battery, a deep neural network is trained to predict the local polarization inside the electrode, avoiding the occurrence of gas side reactions.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A method for predicting local polarization of a flow battery based on a multi-scale model, comprising the following steps:

[0009] S1. Establish a pore-scale model of the electrode in the flow battery based on the pore network model; generate different microscopic pore structures of the electrode as the sample input of the first training sample, and substitute the microscopic pore structures of the electrode in each first training sample into the pore-scale model for solution, and obtain the porosity, permeability and specific surface area corresponding to the microscopic pore structure of the electrode as the sample label of the first training sample, thereby constructing the first training sample set;

[0010] S2. Use the first training sample set to train the first deep neural network until the network converges, so that it can predict the corresponding porosity, permeability, and specific surface area based on the microscopic pore structure of the electrode;

[0011] S3. Combine the Naiver-Stokes equation, Brinkmann equation, Nernst-Planck equation, Butler-Volmer equation, mass conservation equation, and charge conservation equation, and establish a battery-scale model describing the velocity field, concentration field, and potential field inside the flow battery based on the finite volume method to simulate the electrolyte flow and active material mass transfer process; generate different electrolyte active material concentrations, electrolyte input flow rates, applied voltages on the battery, and microscopic pore structures of the electrode as sample inputs of the second training samples, and use the trained first deep neural network to predict the porosity, permeability, and specific surface area corresponding to the microscopic pore structure of the electrode in each second training sample. Then, substitute the predicted porosity, permeability, and specific surface area, the electrolyte active material concentration, electrolyte input flow rate, and applied voltage on the battery in the same second training sample into the battery-scale model. After solving for the velocity field, concentration field, and potential field inside the flow battery, use them as the inlet conditions of the pore-scale model. The pore-scale model calculates the local reaction rate of the active material inside each pore in the electrode based on the inlet conditions and the microscopic pore structure of the electrode and uses it as the sample label of the corresponding second training sample, thereby constructing the second training sample set;

[0012] S4. Use the second training sample set to train the second deep neural network until the network converges, so that it can predict the local reaction rate of the active material inside each pore in the electrode based on the electrolyte active material concentration, electrolyte input flow rate, applied voltage on the battery, and microscopic pore structure of the electrode, thereby reflecting the local polarization situation of the flow battery.

[0013] Preferably, the pore-scale model satisfies the following assumptions: physical properties such as pressure and concentration inside the pores do not change, the electrolyte is a dilute solution, and the electrochemical reaction only occurs inside the pores and not in the throats.

[0014] Preferably, the battery-scale model satisfies the following assumptions: all components in the battery are adiabatic, the fluid in the battery is incompressible, gas side reactions in the battery are ignored, and the transmembrane migration of active materials and water in the battery is ignored.

[0015] Preferably, when solving the pore-scale model and the battery-scale model, the space of the battery and the electrode is discretized, and numerical solutions are performed for each discrete unit separately.

[0016] Preferably, both the first deep neural network and the second deep neural network use BP neural networks.

[0017] Preferably, in the first deep neural network, the BP neural network includes an input layer, four hidden layers, and an output layer, and each hidden layer includes 100 neurons.

[0018] Preferably, in the second deep neural network, the BP neural network includes an input layer, four hidden layers, and an output layer, and each hidden layer includes 400 neurons.

[0019] Preferably, in the first and second deep neural networks, the microscopic pore structure of the electrode is obtained by discretizing the space where the electrode is located into units, and then multiplying the coordinates [x i , y i , z i of each unit by the pore size d pi in the grid, and accumulating the products. The accumulated vector is used as the network input.

[0020] Compared with the prior art, the present invention has the following characteristics:

[0021] 1. Compared with traditional pore-scale models or battery-scale models, the present invention combines pore-scale models and battery-scale models through machine learning to form a multi-scale model with both high accuracy and large scale, which can accurately predict local polarization affected by the electrode fiber structure.

[0022] 2. The present invention uses machine learning methods to learn the relationship between pore-scale geometric conditions, battery-scale inlet conditions, and local polarization at the pore scale, realizes the connection between the battery-scale model and the pore-scale model, greatly reduces the calculation time of the model, and reduces the calculation resources required by the model. The present invention can realize local polarization prediction at the battery scale and extend the pore-scale model to the entire electrode or even the entire stack.

[0023] 3. The present invention can predict local polarization in different situations through a multi-scale model, which can be used to deeply study the mechanism and control strategy of local polarization in flow batteries, and provide a theoretical basis for future local polarization control technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic flow chart of a method for predicting local polarization of a flow battery based on a multi-scale model in the present invention.

[0025] Figure 2 It is a schematic diagram of the deep neural network adopted by the present invention.

[0026] Figure 3 It is the model verification of the battery-scale model (left figure) and the pore-scale model (right figure) in the embodiment of the present invention.

[0027] Figure 4 This is the local polarization distribution diagram under different electrolyte flow rates and different charge states in the embodiments of the present invention. Detailed implementation manners

[0028] The present invention will be further described and explained below in conjunction with the accompanying drawings and specific implementation manners.

[0029] As Figure 1 shown, in a preferred embodiment of the present invention, a method for predicting the local polarization of a flow battery based on a multi-scale model is provided. The multi-scale model includes a battery-scale model, a pore-scale model, and a machine learning model. Among them, the battery-scale model adopted in the present invention has the function of calculating and predicting the required inlet conditions through input parameters, and is used to simulate the electrolyte flow and active material mass transfer processes; the pore-scale model has the functions of constructing the microstructure and calculating the local polarization, and is used to predict the local polarization affected by the microstructure, electrode permeability, specific surface area, and electrochemical reaction; the machine learning model uses a deep neural network and has the function of learning the relationship between the inlet conditions, microstructure and parameters, and local polarization. The present invention combines and connects the functions of these three parts to invent a multi-scale model that can quickly predict the local polarization.

[0030] The construction methods of the battery-scale model, pore-scale model, and machine learning model in the present invention will be described in detail below.

[0031] The battery-scale model in the present invention is used to describe the flow velocity, concentration, and potential distribution inside the electrode. This model is a model that describes the velocity field, concentration field, and potential field inside the flow battery based on the finite volume method by combining the Naiver-Stokes equation, Brinkmann equation, Nernst-Planck equation, Butler-Volmer equation, mass conservation equation, and charge conservation equation. Among them, the flow velocity distribution in the flow channel is described by the Navier-Stokes equation, and the flow velocity distribution in the electrode is described by the Brinkmann equation. These equations are as follows:

[0032]

[0033]

[0034] Among them, ρ is the fluid density, u is the fluid velocity, p is the pressure, μ is the kinematic viscosity, ε p is the porosity, κ is the permeability, F b is the body force, β F is the Forchheimer coefficient, which takes into account the density, porosity, permeability, and zero-dimensional friction coefficient.

[0035] The ion concentration in the electrolyte is described by the mass conservation equation as follows:

[0036]

[0037] where c i is the concentration of species i in the electrolyte, S i is the source term, ε p is the electrode porosity, is the flux of species i in the electrolyte, and the flux of species in the electrolyte can be calculated using the Nernst-Planck equation:

[0038]

[0039] where F is the Faraday constant, φ e is the ion potential, and are the effective diffusivity and effective ion mobility, D i is the diffusivity of species i in the electrolyte, u i is the ion mobility of species i in the electrolyte, z i is the number of electrons transferred in the battery reaction.

[0040] The charge conservation equation in the model is as follows:

[0041]

[0042] where j is the local current density, and are the ion current density and electron current density, which can be calculated by the following equations:

[0043]

[0044]

[0045] where φ s is the electron potential in the current collector plate, σ s is the conductivity of the current collector plate. The local current density can be described by the Butler-Volmer equation:

[0046]

[0047] where a is the specific surface area, k is the reaction rate constant, α is the charge transfer rate, c O is the concentration of the oxide in the electrolyte, is the concentration of the oxide on the electrode surface, c R is the concentration of the reductant in the electrolyte, is the concentration of the reductant on the electrode surface, R is the gas constant, T is the ambient temperature; Eeq is the equilibrium potential and is defined as:

[0048]

[0049] where E0 represents the standard equilibrium potential of the redox couple.

[0050] In addition, in Equation (8), the concentration of substances on the electrode surface can be calculated from the balance between the electrochemical reaction rate and the reactant mass transfer rate:

[0051]

[0052] where k m is the mass transfer coefficient.

[0053] The above battery-scale model needs to meet the following assumptions:

[0054] 1. All components in the battery are adiabatic;

[0055] 2. The fluid in the battery is an incompressible fluid;

[0056] 3. The gas side reactions in the battery are negligible;

[0057] 4. The transmembrane of active substances and water in the battery is negligible.

[0058] When performing model calculations for the above battery-scale model, it is necessary to discretize the calculation range space of the entire battery. Each unit obtained after discretization can be solved and calculated according to its respective parameters. In this embodiment, the calculation range of the battery-scale model is 3.24 cm 2 , which is divided into 81 square regions to correspond to the 4 mm 2 calculation range of the pore-scale model.

[0059] The pore-scale model in the present invention is a pore network model. The geometric structure of the model is a pore network composed of spheres and cylinders. The spheres represent the pores of the porous medium where the electrochemical reaction occurs, and the cylinders represent the throats connecting two pores. The control equations of the pore-scale model are similar to those of the battery scale, except for the mass conservation equation and the material conservation equation. The mass conservation equation is:

[0060]

[0061] where n i is the number of adjacent pores of pore i, u i is the fluid velocity from pore i to pore j, and A ij is the cross-sectional area of the connecting throat. To accurately describe the velocity, the Hagen-Poiseuille equation is introduced:

[0062] uij = α ij (p i - p j )(12)

[0063] where p i and p j are the pressures of pore i and pore j, and α ij = S ij / 8πμl ij is the hydraulic conductivity at the throat length l ij Another mass conservation equation is:

[0064]

[0065] where, R i is the reaction rate of pore i, i.e., the local reaction rate, and m ij is the mass flux from pore i to pore j, which can be calculated by the following equation:

[0066]

[0067] where, c i and c j are the substance concentrations of pore i and pore j, and D represents the hydraulic diameter of the throat.

[0068] The above pore-scale model needs to satisfy the following assumptions:

[0069] 1. Physical properties such as pressure and concentration in the pores hardly change;

[0070] 2. The electrolyte is a dilute solution;

[0071] 3. The electrochemical reaction only occurs in the pores and not in the throats.

[0072] Similarly, when performing model calculations for the above pore-scale model, it is necessary to discretize the calculation range space of the entire electrode. Each unit obtained after discretization can be solved and calculated according to its respective parameters. In this embodiment, the modeling range of the pore-scale model is a cuboid with an area of 4 mm 2 and a thickness of 0.5 mm, which is divided into 32, 32, and 8 grids in each direction.

[0073] The machine learning model in this embodiment adopts a BP deep neural network model, and the network structure is as Figure 2As shown in the figure, it is composed of an input layer, a hidden layer and an output layer. The specific structure and principle of the BP network belong to the prior art and will not be elaborated here. The battery-scale model can simulate the mass transfer process and electrochemical reaction of the active material at the centimeter to meter scale, and has a large simulation range; the pore-scale model can simulate the mass transfer process and electrochemical reaction of the active material at the nanometer to micrometer scale, and can reflect the geometric structure composed of carbon fibers, with high accuracy. The multi-scale model of the present invention learns the relationship between the pore-scale geometric conditions, the battery-scale inlet conditions and the local polarization conditions at the pore scale through a machine learning method, can realize the connection between the battery-scale model and the pore-scale model, has a multi-scale model with high accuracy and large scale, can reduce the prediction time of the local polarization conditions, and improves the accuracy of the local polarization prediction.

[0074] The following will describe in detail the specific method of learning the relationship between the inlet conditions, the microstructure and parameters, and the local polarization through the machine learning model.

[0075] In this embodiment, based on the above battery-scale model and pore-scale model, the method for constructing a multi-scale model through machine learning and predicting the local polarization of the flow battery specifically includes the following steps:

[0076] S1. After establishing the pore-scale model of the electrode in the flow battery based on the above pore network model, different electrode micro-pore structures are generated as the sample input of the first training sample, and the electrode micro-pore structures in each first training sample are substituted into the pore-scale model for solution, and the porosity, permeability and specific surface area corresponding to the electrode micro-pore structures are obtained and used as the sample labels of the first training sample, thereby constructing the first training sample set.

[0077] S2. Use the first training sample set to train the first deep neural network until the network converges, so that it can predict the corresponding porosity, permeability and specific surface area based on the electrode micro-pore structure.

[0078] It should be noted that the specific training method of the first deep neural network can refer to the conventional BP network training method. The first training sample set can be pre-divided into a training set and a test set, and the network parameters are optimized through the gradient descent algorithm until the network converges.

[0079] S3. After establishing the battery-scale model that describes the velocity field, concentration field, and electric potential field inside the flow battery based on the finite volume method by combining the Navier-Stokes equation, Brinkmann equation, Nernst-Planck equation, Butler-Volmer equation, mass conservation equation, and charge conservation equation, the electrolyte flow and active material mass transfer processes can be simulated. Thus, different electrolyte active material concentrations, electrolyte input flow rates, applied voltages on the battery, and electrode micro-porous structures are generated as the sample inputs of the second training samples, and the porosity, permeability, and specific surface area corresponding to the electrode micro-porous structures in each second training sample are predicted using the trained first deep neural network. Then, the predicted porosity, permeability, and specific surface area, together with the electrolyte active material concentration, electrolyte input flow rate, and applied voltage on the battery in the same second training sample, are substituted into the above battery-scale model. After solving for the velocity field, concentration field, and electric potential field inside the flow battery, they are used as the inlet conditions for the above pore-scale model. The pore-scale model calculates the local reaction rate of the active material inside each pore in the electrode based on the inlet conditions and the electrode micro-porous structure and uses it as the sample label for the corresponding second training sample, thereby constructing the second training sample set.

[0080] S4. Use the second training sample set to train the second deep neural network until the network converges, enabling it to predict the local reaction rate of the active material inside each pore in the electrode based on the electrolyte active material concentration, electrolyte input flow rate, applied voltage on the battery, and electrode micro-porous structure, thereby reflecting the local polarization situation of the flow battery. In actual applications, any electrolyte active material concentration, electrolyte input flow rate, applied voltage on the battery, and electrode micro-porous structure can be input into the trained second deep neural network, and the local reaction rate of the active material inside each pore in the electrode can be quickly predicted by the model, thereby reflecting the local polarization situation without the need for complex numerical solutions.

[0081] It should be noted that the specific training method of the second deep neural network can refer to the conventional BP network training method. The second training sample set can be pre-divided into a training set and a test set, and the network parameters can be optimized through the gradient descent algorithm until the network converges.

[0082] It should be noted that for the above first deep neural network and second deep neural network, the specific network parameters can be adjusted according to the actual situation. In this embodiment, in the first deep neural network, the BP neural network includes an input layer, 4 hidden layers, and an output layer, and each hidden layer includes 100 neurons. In the second deep neural network, the BP neural network includes an input layer, 4 hidden layers, and an output layer, and each hidden layer includes 400 neurons. The number of neurons in the input layer and output layer of the first deep neural network and the second deep neural network can be adjusted according to the dimensions of the input vector and the output vector.

[0083] In addition, it should be noted that in order to describe the microscopic structural characteristics of the porous medium, in the first deep neural network and the second deep neural network of this embodiment, the microscopic pore structure of the electrode is discretized into units in the space where the electrode is located, and then the product of each unit coordinate [x i , y i , z i and the pore diameter d pi of the pores in the grid is accumulated, and the accumulated vector is used as the network input.

[0084] As the training data of the deep neural network, the number of samples needs to meet the sample number requirements for training. In this embodiment, the first training sample set contains 500 samples constructed based on the microscopic pore structures generated from real electrodes, while the second training sample set contains 10,000 training samples.

[0085] To verify the accuracy of the battery-scale model and pore-scale model constructed by the present invention, they are compared with the actual experimental results. Finally, the comparison between the model of the present invention and the experimental results is as Figure 3 shown. The results show that when the flow rate increases from 10 mL / min to 20 mL / min, the battery-scale model has a high degree of agreement with the experimental results, and the maximum error is 0.5%, which may be due to the fact that side reactions and transmembrane effects are not considered in the model. The pore-scale model has a good degree of agreement at high current densities, and the error reaches 3% at low current densities, which may be due to the inaccurate method of simulating pores in the pore network model, resulting in excessive activation polarization. Generally speaking, the models used in this embodiment have high accuracy and high credibility.

[0086] Figure 4The local polarization conditions obtained by using the second deep neural network finally trained in the embodiments of the present invention to predict different discrete units in the electrode are shown in the figure for different flow rates and different charge states. The results show that the second deep neural network finally obtained in this embodiment can relatively clearly obtain the local polarization conditions under different states, which has certain guiding significance for the formulation of flow rate strategies and the design direction of electrode structures.

[0087] The above-described embodiments are only a preferred solution of the present invention, but they are not intended to limit the present invention. Those of ordinary skill in the relevant technical fields can make various changes and modifications without departing from the spirit and scope of the present invention. For example, the battery scale model and the pore scale model can also be replaced by other models as long as the same technical effects can be achieved. Therefore, all technical solutions obtained by means of equivalent replacement or equivalent transformation fall within the protection scope of the invention.

Claims

1. A method for predicting local polarization of a flow battery based on a multi-scale model, characterized in that, It includes the following steps: S1. Establish a pore-scale model of the electrode in the flow battery based on the pore network model; generate different microscopic pore structures of the electrode as the sample inputs of the first training samples, and substitute the microscopic pore structures of the electrodes in each first training sample into the pore-scale model for solution, and obtain the porosity, permeability, and specific surface area corresponding to the microscopic pore structure of the electrode as the sample labels of the first training samples, so as to construct the first training sample set; S2. Use the first training sample set to train the first deep neural network until the network converges, so that it can predict the corresponding porosity, permeability, and specific surface area based on the microscopic pore structure of the electrode; S3. Combine the Naiver-Stokes equation, Brinkmann equation, Nernst-Planck equation, Butler-Volmer equation, mass conservation equation, and charge conservation equation, and establish a battery-scale model describing the velocity field, concentration field, and potential field inside the flow battery based on the finite volume method to simulate the electrolyte flow and active substance mass transfer process; Generate different electrolyte active substance concentrations, electrolyte input flow rates, applied voltages applied to the battery, and microscopic pore structures of the electrode as the sample inputs of the second training samples, and use the trained first deep neural network to predict the porosity, permeability, and specific surface area corresponding to the microscopic pore structure of the electrode in each second training sample, and then substitute the predicted porosity, permeability, and specific surface area, the electrolyte active substance concentration, the electrolyte input flow rate, and the applied voltage applied to the battery in the same second training sample into the battery-scale model, and after solving to obtain the velocity field, concentration field, and potential field inside the flow battery, use them as the inlet conditions of the pore-scale model. According to the inlet conditions and the microscopic pore structure of the electrode, the local reaction rate of the active substance in each pore of the electrode is calculated and used as the sample label of the corresponding second training sample, so as to construct the second training sample set; S4. Use the second training sample set to train the second deep neural network until the network converges, so that it can predict the local reaction rate of the active substance in each pore of the electrode based on the electrolyte active substance concentration, the electrolyte input flow rate, the applied voltage applied to the battery, and the microscopic pore structure of the electrode, so as to reflect the local polarization situation of the flow battery.

2. The local polarization prediction method for a flow battery based on a multi-scale model according to claim 1, wherein The pore-scale model satisfies the following assumptions: the physical properties such as pressure and concentration in the pores do not change, the electrolyte is a dilute solution, and the electrochemical reaction only occurs in the pores and does not occur in the throats.

3. The local polarization prediction method for a flow battery based on a multi-scale model according to claim 1, characterized in that, The battery-scale model satisfies the following assumptions: all components in the battery are adiabatic, the fluid in the battery is an incompressible fluid, the gas side reactions in the battery are ignored, and the transmembrane migration of active substances and water in the battery is ignored.

4. The local polarization prediction method for flow batteries based on a multi-scale model according to claim 1, wherein When solving the pore-scale model and the battery-scale model, the space of the battery and the electrode is discretized, and numerical solutions are carried out for each discrete unit respectively.

5. The local polarization prediction method for a flow battery based on a multi-scale model according to claim 1, wherein Both the first deep neural network and the second deep neural network adopt BP neural networks.

6. The method for predicting local polarization of a flow battery based on a multi-scale model according to claim 5, wherein In the first deep neural network, the BP neural network includes an input layer, four hidden layers, and an output layer, and each hidden layer contains 100 neurons.

7. The local polarization prediction method of the flow battery based on the multi-scale model according to claim 5, characterized in that In the second deep neural network, the BP neural network includes an input layer, four hidden layers, and an output layer, and each hidden layer contains 400 neurons.

8. The method for predicting local polarization of a flow battery based on a multi-scale model according to claim 1, wherein In the first deep neural network and the second deep neural network, the microscopic pore structure of the electrode is obtained by discretizing the space where the electrode is located into units, and then accumulating the product of the coordinates [x i , y i , z i of each unit and the pore diameter d pi in the grid. The accumulated vector is used as the network input.

Images