Method and system for predicting water content of a proton exchange membrane
By combining the artificial fish swarm algorithm with the sparrow algorithm and improving the training rules of the convolutional neural network, the local optimum problem of predicting the water content of the proton exchange membrane in fuel cells was solved, and efficient and accurate prediction results were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG UNIV
- Filing Date
- 2023-05-30
- Publication Date
- 2026-06-26
AI Technical Summary
Existing convolutional neural networks are prone to getting stuck in local optima when predicting the water content of proton exchange membranes in fuel cells, resulting in low training efficiency and inaccurate prediction results.
The swarming and tail-chasing behaviors from the artificial fish swarm algorithm are incorporated into the follower and scout behaviors from the sparrow algorithm, respectively, to form new position update rules and train a convolutional neural network.
This improved the training efficiency of convolutional neural networks, avoided local optima, and enabled more accurate prediction of proton exchange membrane water content.
Smart Images
Figure CN119067175B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fuel cell technology, specifically to a method and system for predicting the water content of proton exchange membranes. Background Technology
[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.
[0003] Fuel cells use hydrogen and oxygen as fuel to generate electricity, with water as a byproduct, making them a green and clean energy source. During operation, the internal humidity of a fuel cell is a crucial state parameter. Internal humidity refers to the water content of the proton exchange membrane, which directly affects the proton transport process, thus impacting fuel cell efficiency. It can also cause changes in internal resistance, altering parameters such as output voltage, current, and power, thus affecting the fuel cell's output characteristics. When the water content is below a certain range, "membrane drying" may occur; when it exceeds a certain range, "flooding" may occur, potentially leading to cell failure in severe cases. Therefore, the internal humidity of the fuel cell (usually the water content on the cathode side) is used as a control target to maintain fuel cell performance.
[0004] The water content inside a fuel cell cannot be directly measured. Currently, the water content is indirectly obtained by calculating various parameters during fuel cell operation. If a convolutional neural network is used to predict the water content, the convolutional neural network needs to be trained to obtain accurate water content prediction results. During the training period, it is easy to get stuck in local optima, resulting in inaccurate prediction results obtained from the training data. It requires a long training time to meet the requirements, thus indirectly making the parameter training inefficient. Summary of the Invention
[0005] To address the technical problems mentioned above, this invention provides a method and system for predicting the water content of proton exchange membranes. The method integrates the swarming and tail-chasing behaviors from the artificial fish swarm algorithm with the follower and scout behaviors from the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. A convolutional neural network is then trained based on the fused position update rules.
[0006] To achieve the above objectives, the present invention adopts the following technical solution:
[0007] The first aspect of the present invention provides a method for predicting the water content of a proton exchange membrane, comprising the following steps:
[0008] The internal resistance and temperature of the fuel cell membrane electrode, as well as the thickness of the fuel cell membrane, are obtained, and the water content of the corresponding proton exchange membrane is predicted based on the trained convolutional neural network.
[0009] The convolutional neural network uses the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured membrane water content of the fuel cell membrane electrode as training data. It integrates the swarming and tail-chasing behaviors from the artificial fish swarm algorithm with the follower and scout behaviors from the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. The convolutional neural network is then trained based on the fused position update rules.
[0010] By incorporating the swarming behavior from the artificial fish algorithm into the follower position update rule from the sparrow algorithm, the fused follower position update rule is obtained as follows:
[0011]
[0012] In the formula, The optimal position occupied by the discoverer. A represents the worst position globally in the current iteration. + =A T (AA T ) -1 A is a 1×d matrix whose elements are randomly selected as 1 or -1, Q is a random number that follows a normal distribution, L is a matrix whose elements are all 1, Rand() is a random number that follows a uniform distribution between -1 and 1, and ω1 and ω2 represent weight coefficients.
[0013] By incorporating the tail-chasing behavior of artificial fish swarms into the scout position update rule in the Sparrow Algorithm, the fused scout position update rule is obtained as follows:
[0014]
[0015] In the above formula, For the global optimal position, β is the step size control variable that follows a normal distribution with zero mean and unit variance, K is a random number between 1 and -1, and f n It is the current fitness value of the sparrow, f b f is the globally optimal fitness value. w ε is the worst fitness value globally, ε is a sufficiently small normal number, Rand() is a random number between -1 and 1 that follows a uniform distribution, and ω1 and ω2 represent the weight coefficients.
[0016] The convolutional neural network is trained using the internal resistance, temperature, and thickness of the proton exchange membrane of the fuel cell membrane electrode as inputs, and the pre-measured membrane water content as output.
[0017] The internal resistance of the membrane electrode of a fuel cell is measured by impedance method. Specifically, a sine wave is used as the excitation signal to disturb the electrode, the response signal is measured, and the response signal is processed and analyzed to obtain the corresponding impedance.
[0018] The water content of the fuel cell membrane is measured by the drying and weighing method. Specifically, the fuel cell is placed in a drying oven and dried. The water content of the fuel cell membrane is obtained by measuring the weight difference before and after drying.
[0019] In the Sparrow Algorithm, each individual in the sparrow population corresponds to a feasible solution to the optimization problem. Let the population size be N. In each iteration, the follower will move closer to the discoverer with a high fitness value and forage for food. When the fitness value of the follower at its position is better than that of the discoverer, i.e., i≤N / 2, the follower's position will replace the discoverer's position. When the follower is in a state of extreme hunger, i.e., i>N / 2, the follower will change its position to forage in other areas.
[0020] A second aspect of the present invention provides a system for implementing the above-described method, comprising:
[0021] The water content prediction module is configured to: obtain the internal resistance and temperature of the fuel cell membrane electrode, as well as the thickness of the fuel cell membrane, and predict the corresponding water content of the proton exchange membrane based on the trained convolutional neural network.
[0022] The convolutional neural network uses the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured membrane water content of the fuel cell membrane electrode as training data. It integrates the swarming and tail-chasing behaviors from the artificial fish swarm algorithm with the follower and scout behaviors from the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. The convolutional neural network is then trained based on the fused position update rules.
[0023] A third aspect of the present invention provides a computer-readable storage medium.
[0024] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the proton exchange membrane water content prediction method described above.
[0025] A fourth aspect of the present invention provides a computer device.
[0026] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the proton exchange membrane water content prediction method described above.
[0027] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects:
[0028] 1. Using the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured water content of the fuel cell membrane electrode as training data, the swarming and tail-chasing behaviors in the artificial fish swarm algorithm are integrated into the follower and scout behaviors in the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. Based on the fused position update rules, a convolutional neural network is trained, so that the convolutional neural network does not get stuck in local optima during training, thus obtaining the required parameters efficiently.
[0029] 2. The swarming behavior in the artificial fish swarm algorithm is integrated into the follower behavior in the sparrow algorithm, and the tail-chasing behavior in the artificial fish swarm algorithm is integrated into the scout behavior in the sparrow algorithm. This results in a new position update rule, which improves the performance of the sparrow algorithm and forms a combined algorithm. Based on this combined algorithm, a convolutional neural network is trained so that the convolutional neural network no longer gets trapped in local optima. Attached Figure Description
[0030] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0031] Figure 1 This is a schematic diagram of the proton exchange membrane water content prediction process provided by one or more embodiments of the present invention. Detailed Implementation
[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0033] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0034] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0035] The following embodiments present a method and system for predicting the water content of proton exchange membranes (PEMs). The method combines the Sparrow Algorithm and the Artificial Fish Swarm Algorithm, and utilizes this combined algorithm to optimize a convolutional neural network. The newly obtained convolutional neural network is trained on a database of the internal resistance, temperature, thickness, and water content of the fuel cell membrane electrode assembly (MEA). After training, the internal resistance and temperature of the MEA are collected, and the thickness of the fuel cell membrane is obtained. Based on the trained convolutional neural network, the corresponding water content of the EME is predicted.
[0036] Example 1:
[0037] like Figure 1 As shown, the method for predicting the water content of a proton exchange membrane includes the following steps:
[0038] The internal resistance and temperature of the fuel cell membrane electrode, as well as the thickness of the fuel cell membrane, are obtained, and the water content of the corresponding proton exchange membrane is predicted based on the trained convolutional neural network.
[0039] The convolutional neural network uses the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured membrane water content of the fuel cell membrane electrode as training data. It integrates the swarming and tail-chasing behaviors from the artificial fish swarm algorithm with the follower and scout behaviors from the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. The convolutional neural network is then trained based on the fused position update rules.
[0040] Specifically:
[0041] The swarming and tail-chasing behaviors in the artificial fish swarm algorithm are combined with the follower and scout behaviors in the sparrow algorithm to obtain a combined algorithm. The combined algorithm is then used to optimize the convolutional neural network.
[0042] The internal resistance of the fuel cell membrane electrode was measured by AC impedance method, and the temperature was measured at the same time. The water content of the fuel cell proton exchange membrane was obtained by drying and weighing method. A database of the internal resistance, temperature, thickness and water content of the fuel cell membrane electrode was obtained.
[0043] An optimized convolutional neural network was trained on a database of fuel cell membrane electrode assembly (MEA) internal resistance, temperature, proton exchange membrane (PEM) thickness, and membrane water content to obtain a method for predicting PEM water content. The internal resistance and temperature of the MEA were collected, and the thickness of the fuel cell membrane was obtained. Based on the trained convolutional neural network, the corresponding PEM water content was predicted.
[0044] A proton exchange membrane fuel cell has components such as current collectors, bipolar plates, membrane electrode assemblies (MEAs), and sealing gaskets. The MEAs of a proton exchange membrane fuel cell include a gas diffusion layer, a microporous layer, a catalyst layer, and a proton exchange membrane.
[0045] The method for measuring the internal resistance of the membrane electrode assembly (MEA) of a fuel cell is the AC impedance method. This method uses a small sine wave as an excitation signal to disturb the system, then measures its response signal, processes and analyzes the response signal to obtain the corresponding impedance, and simultaneously measures its temperature.
[0046] The water content of the fuel cell membrane was measured using a drying and weighing method. The fuel cell was placed in a drying chamber and dried, and the water content of the membrane was obtained by measuring its weight before and after drying. Data on the fuel cell's internal resistance, temperature, proton exchange membrane thickness, and water content were used as training data.
[0047] The combinatorial algorithm integrates the swarming behavior of the artificial fish swarm algorithm with the follower behavior of the sparrow algorithm, and the tail-chasing behavior of the artificial fish swarm algorithm with the scout behavior of the sparrow algorithm, to obtain new position update rules to improve the performance of the sparrow algorithm, thus forming a combinatorial algorithm.
[0048] In the Sparrow Algorithm, each individual sparrow in the population represents a feasible solution to the corresponding optimization problem. Let the population size be N, and the position of the nth sparrow be... Where d represents the dimension of the optimization variable, and the element value of each dimension is the value of the corresponding optimization variable.
[0049] In each iteration, followers will approach the discoverer with a high fitness value and forage for food. When a follower's fitness value at its current location is better than the discoverer's (i≤N / 2), the follower's location will replace the discoverer's location. When a follower is in a state of extreme hunger (i>N / 2), it will change its location to forage in other areas. The follower's location update rule is as follows:
[0050]
[0051] In the above formula, The optimal position occupied by the discoverer. A represents the worst position globally in the current iteration. + =A T (AA T ) -1 A is a 1×d matrix whose elements are randomly selected as 1 or -1, Q is a random number that follows a normal distribution, and L is a matrix whose elements are all 1.
[0052] Sparrow populations randomly select a portion of their population to act as scouts. When a scout senses danger, individuals at the edge of the area move towards a safe zone, while sparrows in the center move randomly. The position update pattern is as follows:
[0053]
[0054] In the above formula, For the global optimal position, β is the step size control variable that follows a normal distribution with zero mean and unit variance, K is a random number between 1 and -1, and f n It is the current fitness value of the sparrow, f b f is the globally optimal fitness value. w ε is the worst fitness value globally, and ε is a sufficiently small positive constant.
[0055] The artificial fish swarm algorithm uses N artificial fish in a d-dimensional search space. The state and position of the artificial fish are represented by a vector X = (x1, x2, ..., xn). d ) indicates that ||X i -X j || represents the spatial distance between any two artificial fish, the field of vision of an artificial fish is Visual, the crowding degree of the artificial fish group is represented by the crowding factor δ, the step length of the artificial fish is step, and the food concentration at the location of the artificial fish is Y=f(x).
[0056] Fish naturally exhibit gregarious behavior. When encountering predators or foraging in their environment, they cooperate in groups to increase their chances of survival and foraging efficiency. (State X) i Artificial fish, in d i,j <Count the number of other artificial fish individuals n within the Visual range> f And calculate its average state X. C :
[0057]
[0058] If Y C / n f >δ*Y i If true, then state X exists within the optimization space. C This location offers high adaptability and can accommodate a large number of artificial fish individuals. These artificial fish individuals can move towards state X. C They will gather in one place, otherwise they will continue to forage. Location updated to:
[0059]
[0060] In the above formula, Indicates state X i The artificial fish in the (t+1)th iteration, Indicates state X i In the t-th iteration of the artificial fish, Rand() is a random number between -1 and 1 that follows a uniform distribution.
[0061] Fish groups, driven by their natural instincts to move towards food and avoid predators, will follow the movement of some individuals in a particular direction. The state is X. i Artificial fish, in d i,j <Count the number of other artificial fish individuals n within the Visual range> f And within this range, identify the individual X with the highest fitness. j Its fitness value is Y j If Y j / n f >δ*Y i Then X in the search space j The area exhibits high adaptability and still offers space for continued grouping; artificially bred fish can be introduced to X. j They will move elsewhere, or continue foraging.
[0062]
[0063] Combining the Sparrow Algorithm and the Artificial Fish Swarm Algorithm: The swarming behavior from the Artificial Fish Algorithm is incorporated into the follower position update rule of the Sparrow Algorithm, resulting in a new follower position update rule:
[0064]
[0065] In the formula, The optimal position occupied by the discoverer. A represents the worst position globally in the current iteration. + =A T (AA T ) -1 A is a 1×d matrix whose elements are randomly selected as 1 or -1, Q is a random number that follows a normal distribution, L is a matrix whose elements are all 1, Rand() is a random number that follows a uniform distribution between -1 and 1, and ω1 and ω2 represent weight coefficients.
[0066] By incorporating the tail-chasing behavior of artificial fish swarms into the scout position update rule in the sparrow algorithm, a new scout position update rule is obtained:
[0067]
[0068] In the above formula, For the global optimal position, β is the step size control variable that follows a normal distribution with zero mean and unit variance, K is a random number between 1 and -1, and f n It is the current fitness value of the sparrow, f b f is the globally optimal fitness value. wε is the worst fitness value globally, ε is a sufficiently small normal number, Rand() is a random number between -1 and 1 that follows a uniform distribution, and ω1 and ω2 represent the weight coefficients.
[0069] This led to the combination algorithm.
[0070] Convolutional Neural Networks (CNNs) construct their network structure based on local connectivity, weight sharing, and downsampling. They consist of an input layer, alternating convolutional and downsampling layers, fully connected layers, and an output layer. A combined algorithm is then used to train the parameters within the CNN.
[0071] The parameters include the kernel weights w in the convolutional layer. c,k (k = 1, 2, ..., Nc), bias b c,k (k = 1, 2, ..., Nc), fully connected layer weights w L4 Bias b L4 Nc represents the number of convolutional kernels. This leads to an improved convolutional neural network based on a combinatorial algorithm. The improved convolutional neural network is then used to train and test a database of fuel cell internal resistance, temperature, proton exchange membrane thickness, and membrane water content.
[0072] After training, the internal resistance and temperature of the fuel cell membrane electrode are collected, and the thickness of the fuel cell membrane is obtained. Based on the trained convolutional neural network, the water content of the corresponding proton exchange membrane is predicted.
[0073] Example 2:
[0074] A system for implementing the above method includes:
[0075] The water content prediction module is configured to: obtain the internal resistance and temperature of the fuel cell membrane electrode, as well as the thickness of the fuel cell membrane, and predict the corresponding water content of the proton exchange membrane based on the trained convolutional neural network.
[0076] The convolutional neural network uses the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured membrane water content of the fuel cell membrane electrode as training data. It integrates the swarming and tail-chasing behaviors from the artificial fish swarm algorithm with the follower and scout behaviors from the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. The convolutional neural network is then trained based on the fused position update rules.
[0077] Example 3:
[0078] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the proton exchange membrane water content prediction method as described in Embodiment 1 above.
[0079] Example 4:
[0080] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the proton exchange membrane water content prediction method as described in Embodiment 1 above.
[0081] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.
[0082] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for predicting the water content of a proton exchange membrane, characterized in that, Includes the following steps: The internal resistance and temperature of the fuel cell membrane electrode, as well as the thickness of the fuel cell membrane, are obtained, and the water content of the corresponding proton exchange membrane is predicted based on the trained convolutional neural network. The convolutional neural network uses the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured membrane water content of the fuel cell membrane electrode as training data. It integrates the swarming behavior and tail-chasing behavior in the artificial fish swarm algorithm into the follower behavior and scout behavior in the sparrow algorithm, respectively, to obtain the fused follower behavior and scout behavior position update rules, and trains the convolutional neural network based on the fused position update rules. By incorporating the swarming behavior from the artificial fish algorithm into the follower position update rule from the sparrow algorithm, the fused follower position update rule is obtained as follows: ； In the formula, The optimal position for the discoverer. This represents the worst position globally in the current iteration. Let A be a 1×d matrix whose elements are randomly selected as 1 or -1, Q be random numbers that follow a normal distribution, L be a matrix whose elements are all 1, and Rand() be a random number that follows a uniform distribution between -1 and 1. Indicates the weighting coefficient; By incorporating the tail-chasing behavior of artificial fish swarms into the scout position update rule in the Sparrow Algorithm, the fused scout position update rule is obtained as follows: ； In the above formula, For the best position globally, The step size control variable follows a normal distribution with zero mean and unit variance, and K is a random number between 1 and -1. This is the current fitness value of the sparrow. This represents the globally optimal fitness value. This represents the worst-case fitness value globally. It is a sufficiently small positive constant, and Rand() is a random number between -1 and 1 that follows a uniform distribution. This represents the weighting coefficient.
2. The method for predicting the water content of a proton exchange membrane as described in claim 1, characterized in that, The convolutional neural network is trained using the internal resistance, temperature, and thickness of the proton exchange membrane of the fuel cell membrane electrode as inputs, and the pre-measured membrane water content as output.
3. The method for predicting the water content of a proton exchange membrane as described in claim 2, characterized in that, The internal resistance of the membrane electrode of a fuel cell is measured by impedance method. Specifically, a sine wave is used as the excitation signal to disturb the electrode, the response signal is measured, and the response signal is processed and analyzed to obtain the corresponding impedance.
4. The method for predicting the water content of a proton exchange membrane as described in claim 2, characterized in that, The water content of the fuel cell membrane is measured by the drying and weighing method. Specifically, the fuel cell is placed in a drying oven and dried. The water content of the fuel cell membrane is obtained by measuring the weight difference before and after drying.
5. The method for predicting the water content of a proton exchange membrane as described in claim 1, characterized in that, In the Sparrow Algorithm, each individual in the sparrow population corresponds to a feasible solution to the optimization problem. Let the population size be N. In each iteration, the follower will move closer to the discoverer with a high fitness value and forage for food. When the fitness value of the follower at its position is better than that of the discoverer, i.e., i≤N / 2, the follower's position will replace the discoverer's position. When the follower is in a state of extreme hunger, i.e., i>N / 2, the follower will change its position to forage in other areas.
6. A proton exchange membrane water content prediction system, employing the proton exchange membrane water content prediction method as described in any one of claims 1-5, characterized in that, include: The water content prediction module is configured to: obtain the internal resistance and temperature of the fuel cell membrane electrode, as well as the thickness of the fuel cell membrane, and predict the corresponding water content of the proton exchange membrane based on the trained convolutional neural network. The convolutional neural network uses the internal resistance, temperature, thickness of the proton exchange membrane, and pre-measured membrane water content of the fuel cell membrane electrode as training data. It integrates the swarming and tail-chasing behaviors from the artificial fish swarm algorithm with the follower and scout behaviors from the sparrow algorithm, respectively, to obtain the fused follower and scout behavior position update rules. The convolutional neural network is then trained based on the fused position update rules.
7. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the proton exchange membrane water content prediction method as described in any one of claims 1-5.
8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the proton exchange membrane water content prediction method as described in any one of claims 1-5.