A proton exchange membrane fuel cell control method

Abstract

The present invention discloses a proton exchange membrane fuel cell control method. The method comprises: using an established cell voltage model to monitor the cell state and obtain a predicted voltage; comparing the difference between the predicted voltage and the actual measured voltage with a set threshold to determine whether the cell is in an abnormal state; and, if the cell state is determined to be abnormal, inputting the measured cell voltage, ohmic internal resistance, and anode gas pressure drop into a trained bagging neural network ensemble classifier to determine the cell fault type. This method ensures prediction accuracy and timeliness while reducing the amount of model training data. Furthermore, without increasing the amount of feature data, it can diagnose both membrane dry and flooding faults.

Description

Technical Field

[0001] The present invention relates to the technical field of battery health control, and more specifically, to a proton exchange membrane fuel cell control method. Background Art

[0002] Proton exchange membrane fuel cells (PEMFCs) convert hydrogen into electricity during use. They offer advantages such as quiet operation, zero pollution, high operating current, high specific power, fast startup, environmental friendliness, and a compact structure. They can fundamentally address the environmental and energy challenges plaguing automotive development, and are gaining increasing attention from countries and businesses. However, the slow pace of commercialization of PEMFCs stems largely from concerns about their lifespan and health. Due to variability in external conditions and operating conditions, PEMFCs are prone to failures such as membrane drying and flooding, which can severely degrade the performance and lifespan of the fuel cell stack. Therefore, to avoid significant damage to the fuel cell stack, timely fault detection and repair are crucial.

[0003] In the existing technology, the diagnosis methods for proton exchange membrane fuel cell flooding and membrane dry failure are mainly based on mathematical models, experiments and data-driven diagnostic methods. For the method based on mathematical models, the key is to establish an analytical model that matches the fuel cell, which is crucial for simulating the performance of the battery system. However, this method requires researchers to have an in-depth understanding of the internal structure and reaction kinetics of the fuel cell system. The calculation accuracy of the model often depends on its complexity. For models with higher complexity, although the accuracy of the model will be improved, the calculation cost and calculation time will be significantly increased. In addition, since the fuel cell is a complex system coupled with multiple physical fields, it is difficult to fully and deeply understand the internal principles of the battery. Therefore, the application of model-based methods in real-time diagnosis is very limited, and it is only widely used in fuel cell offline simulation and design.

[0004] Experimental methods primarily include diagnosing flooding and membrane drying faults based on anode gas pressure drop, AC impedance spectroscopy, and membrane impedance, as well as using visualization techniques to study the dynamic processes of flooding and membrane drying. While anode pressure drop can guide operating conditions to restore the PEMFC to normal operation, this method struggles to accurately locate a single faulty cell within the stack. Regarding fault location, methods based on AC impedance spectroscopy and membrane impedance can accurately locate faults based on the impedance characteristics of individual cells and describe the impedance characteristics of the entire stack. However, these processes, such as frequency sweep impedance spectroscopy and averaging multiple measurements to obtain the impedance at a single frequency point, are time-consuming and hinder online fault diagnosis. Visualization techniques can directly observe the distribution of water within the cell and the dynamic changes in the gas-liquid phase, facilitating research into water mechanisms within PEMFCs. However, imaging equipment is expensive, requires transparent cells, and is only suitable for laboratory research, limiting commercial applications and online testing.

[0005] Data-driven fault diagnosis techniques rely primarily on the analysis of extensive historical data and are independent of the specific model of the diagnostic object. Currently, data-driven methods for diagnosing PEMFC flooding and membrane drying have garnered widespread attention. This approach directly utilizes historical PEMFC operational data as training data to generate a corresponding training model. This training data is then processed using various mathematical algorithms to capture the characteristics of flooding, membrane drying, and normal conditions. These characteristics are then used to determine the operating state of the battery under given conditions. Data-driven fault diagnosis methods can be used for online diagnosis of flooding and membrane drying faults in high-power PEMFC stacks and multiple stacks, and offer significant advantages for locating faulty individual cells within a stack.

[0006] Machine learning is a common method among data-driven approaches. For example, a BP neural network-based approach is used for PEMFC fault diagnosis. While this method can accurately determine the type of PEMFC fault, the neural network model requires a large amount of data for training. Furthermore, the more input variables in the model and the more complex the mapping relationships, the more training data is required, increasing the cost and time required to obtain data. Therefore, it is necessary to optimize the neural network model to reduce the amount of data required for model training while ensuring the accuracy and timeliness of the model's predictions. Another example is a support vector machine-based approach for PEMFC fault diagnosis. Current density, membrane water content, and ohmic internal resistance are used as three characteristic data points for the fuel cell and then substituted into a pre-trained support vector machine classifier to determine whether the fuel cell is in a membrane-dry state. This support vector machine-based approach can only determine whether the fuel cell is in a membrane-dry state, not whether it is flooded, which limits its commercial application. Summary of the Invention

[0007] The purpose of the present invention is to overcome the above-mentioned shortcomings of the prior art and provide a proton exchange membrane fuel cell control method. The method comprises:

[0008] Using the established battery voltage model to monitor the battery state and obtain a predicted voltage, the battery is a proton exchange membrane fuel cell;

[0009] comparing the difference between the predicted voltage and the actual measured voltage with a set threshold to determine whether the battery is in an abnormal state;

[0010] When the battery is determined to be in an abnormal state, the measured battery voltage, ohmic internal resistance and anode gas pressure drop are input into the trained Bagging neural network ensemble classifier to obtain the battery fault type.

[0011] Compared with the existing technology, the advantages of the present invention are that, in order to solve the problem of a large amount of training data required for the neural network, bagging ensemble learning is introduced, so that the obtained bagging neural network ensemble model has higher accuracy and requires less model training data; in order to solve the problem of only being able to diagnose the PEMFC membrane dry, ohmic internal resistance, anode gas pressure drop and battery voltage are used as characteristic data of the bagging neural network ensemble. Without increasing the amount of characteristic data, the obtained model can not only determine whether the fuel cell is in a membrane dry state, but also determine whether the fuel cell is in a flooded state.

[0012] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0014] Figure 1 is a flow chart of a proton exchange membrane fuel cell control method according to one embodiment of the present invention;

[0015] Figure 2 is a structural diagram of a single neural network model according to one embodiment of the present invention;

[0016] Figure 3 2 is a schematic diagram of bagging ensemble learning according to one embodiment of the present invention;

[0017] Figure 4 Schematic diagram of the application process of a proton exchange membrane fuel cell control method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention.

[0019] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0020] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0021] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0022] It should be noted that like reference numerals and letters refer to like items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0023] To ensure the long-term and stable operation of proton exchange membrane fuel cells, it is necessary to avoid failures such as flooding and membrane drying. If flooding or membrane drying does occur, timely repairs must be performed to prevent further deterioration. This paper applies bagging ensemble learning to proton exchange membrane fuel cells, addressing the problem of diagnosing flooding and membrane drying failures in these systems and effectively improving their reliability and durability.

[0024] In short, the present invention monitors the status of a proton exchange membrane fuel cell based on a proton exchange membrane fuel cell voltage model, wherein the proton exchange membrane fuel cell voltage model is a voltage model of a proton exchange membrane fuel cell in a healthy state; if the proton exchange membrane fuel cell is in an abnormal state, based on the fuel cell internal resistance model, the EIS (electrochemical impedance spectroscopy) method is used to select the real part and imaginary part corresponding to a certain frequency (for example, 20KHz) of the electrochemical impedance spectroscopy to solve the ohmic internal resistance of the battery; the anode gas pressure drop is obtained through the anode gas inlet and outlet pressure sensors; the battery voltage at this time is measured by a voltmeter connected in parallel in the circuit; the obtained ohmic internal resistance, anode gas pressure drop and battery voltage are input into a trained Bagging neural network integrated classifier model to perform fault diagnosis on the proton exchange membrane fuel cell; and then fault recovery is implemented to achieve performance recovery of the proton exchange membrane fuel cell.

[0025] Specifically, see Figure 1As shown, the provided proton exchange membrane fuel cell control method includes the following steps.

[0026] Step S110 : establishing a proton exchange membrane fuel cell voltage model to obtain a predicted voltage.

[0027] For example, the voltage model of a proton exchange membrane fuel cell is expressed as:

[0028] V st =N*(E nernst -V act -V ohm -V conc ) (1)

[0029] Among them, V st is the voltage of the proton exchange membrane fuel cell, N is the number of proton exchange membrane fuel cell cells, E nernst is the Nernst voltage, V act is the activation loss, V ohm is the ohmic loss, V conc is the concentration loss.

[0030] In one embodiment, the Nernst voltage E nernst Expressed as:

[0031]

[0032] Among them, T st is the actual temperature of the proton exchange membrane fuel cell, and Actual measured hydrogen and oxygen partial pressures for proton exchange membrane fuel cells.

[0033] Activation loss V act Expressed as:

[0034] V act =V0+V a (1-e -10i ) (3)

[0035] Among them, V0 and V a are the first and second empirical parameters, and i is the actual measured current density.

[0036] Ohmic loss V ohm Expressed as:

[0037]

[0038] Among them, λ m is the membrane water content.

[0039] Concentration loss V conc Expressed as:

[0040]

[0041] Among them, a is the empirical parameter, i max is the maximum current density.

[0042] Furthermore, the predicted voltage V obtained by the proton exchange membrane fuel cell voltage model is st The difference is obtained by subtracting the theoretical voltage (or theoretical voltage) from the actual measured value of the proton exchange membrane fuel cell voltage. If the difference is greater than the preset fault voltage threshold, it means that the proton exchange membrane fuel cell has failed. The fault voltage threshold can be set according to actual working conditions and experience;

[0043] Step S120 : establishing a fuel cell internal resistance model to obtain ohmic internal resistance.

[0044] Due to the existence of polarization, the actual output voltage of the fuel cell is slightly lower than the theoretical voltage. According to the different causes and characteristics of the polarization phenomenon, the resistance can be divided into activation internal resistance R f 、Ohmic internal resistance R m and concentration internal resistance R d . According to the internal mechanism of fuel cells, the equivalent internal resistance model of fuel cells is established.

[0045] For example, the total internal resistance of the battery stack R stack Expressed as:

[0046] R stack =R f +R m +R d (6)

[0047] The expression of AC impedance is as follows:

[0048]

[0049] Among them, C d 、C d1 is the double layer capacitance, j is the imaginary unit, and w represents the frequency.

[0050] Step S130: determining characteristic data characterizing membrane dry fault and water flooding fault.

[0051] Membrane drying is a failure caused by excessively high internal fuel cell temperatures or insufficient membrane humidity. When a fuel cell stack experiences membrane drying, the proton exchange membrane becomes dry, hindering electrode hydration. This significantly increases the membrane's ohmic internal resistance and reduces conductivity, hindering proton access to the catalyst layer surface. Over time, output performance continues to decline. Waterlogging is a failure caused by excessively low internal fuel cell temperatures or excessive membrane humidity. When a fuel cell stack experiences waterlogging, excess water floods the pores of the catalyst layer (CL) and gas diffusion layer (GDL), hindering reactant transport and significantly increasing gas pressure drop. The anode gas pressure drop varies more significantly with water content than the cathode gas pressure drop, leading to a severe decline in fuel cell performance or even failure. Whether membrane drying or waterlogging occurs, the cell voltage will fluctuate.

[0052] Based on the analysis of the failure principle of membrane dryness and flooding, the index parameters defining membrane dryness and flooding failure are determined. For example, the battery voltage V st , ohmic internal resistance R m and anode gas pressure drop The parameters are used as characteristic data to characterize membrane drying and flooding failures.

[0053] Step S140: construct a Bagging neural network ensemble classifier and perform training using the determined feature data.

[0054] Take the 3-layer neural network model as an example to illustrate. Figure 2 The single neural network model shown in the figure has three layers: input layer, hidden layer (or called hidden layer) and output layer. The characteristic data is used as the input variable of the neural network, that is, the battery voltage V st 、Ohmic internal resistance R m and anode gas pressure drop As the input variable of the neural network, the fault type of the proton exchange membrane fuel cell is used as the output variable of the output layer.

[0055] In order to more accurately identify the fault type, in one embodiment, a "heuristic method" is introduced to determine the optimal number of nodes in the hidden layer and construct a back propagation (BP) neural network. The "heuristic method" is specifically:

[0056] First, set the starting number of nodes X0 and the starting step length h0, expressed as:

[0057] X0=log2N in

[0058] h0=(N im -N out) / 3 (8)

[0059] Among them, X0 is the starting number of nodes in the hidden layer, N in is the number of input units, N out is the number of output units.

[0060] Then, determine the number of comparison nodes X1, which is calculated as follows:

[0061] X1=X0+h0 (9)

[0062] Calculate the neural network prediction error values ​​f0(x) and f1(x) corresponding to the two points (i.e., the starting node number and the comparison node number) in sequence, and determine the position of the next point (i.e., the next comparison node number) based on the size relationship between the two points, which can be expressed as:

[0063]

[0064] X2 is the position of the next point, and the above process is repeated until the optimal point X is found. i This is the optimal number of hidden layer nodes.

[0065] In one embodiment, the constructed Bagging neural network ensemble classifier model is as follows: Figure 3 As shown, it contains multiple weak learners, marked as T, each weak learner is trained with a sampling set, each sampling set contains M samples, and each sampling set is obtained by randomly sampling the original sample training set.

[0066] The training process of the Bagging neural network ensemble model is: given a training set containing M samples, first randomly take out a sample and put it into the sampling set, and then put the sample back into the initial training set, so that the sample may still be selected in the next sampling. After M random sampling operations, a sampling set containing M samples is obtained, such as sampling set 1. According to this method, T sampling sets containing M samples can be sampled, and then a weak learner can be trained based on each sampling set. These weak learners are then combined by voting, that is, the prediction result is the category with the most votes, thereby constructing a Bagging neural network ensemble model. In the constructed Bagging neural network ensemble model, the weak learner can be used Figure 2 3-layer BP neural network model.

[0067] Furthermore, a working bagging neural network ensemble classifier was constructed using relevant software using a training set of data for membrane dry and flooding fault classification. First, a portion of preprocessed data was selected as the training set. The bagging neural network ensemble classifier was then constructed using the relevant software based on this training set. The remaining preprocessed data served as the test set, which was then used to validate the bagging neural network ensemble classifier constructed using the training set data. If the remaining data could be correctly validated, the model was considered effective and reliable.

[0068] Step S150 : Using the trained Bagging neural network ensemble classifier to identify the battery fault type and execute a corresponding fault recovery strategy.

[0069] After the model training is completed, it can be applied to the actual online detection of battery faults. Figure 4 As shown in Figure 2, the model application process generally includes the following steps:

[0070] Step S501 : monitoring the state of the proton exchange membrane fuel cell to determine whether the cell is in an abnormal state.

[0071] For example, the voltage V obtained by the proton exchange membrane fuel cell voltage model st The difference is obtained by subtracting the actual measured value of the proton exchange membrane fuel cell voltage. If the difference is greater than the preset fault voltage threshold, it indicates that the proton exchange membrane fuel cell has failed.

[0072] Step S502 : When the proton exchange membrane fuel cell is in an abnormal state, the ohmic internal resistance, voltage and anode gas pressure drop parameters of the fuel cell are measured as characteristic data.

[0073] For example, using the electrochemical impedance spectroscopy (EIS) method, a set of small-amplitude AC potential wave signals with different frequencies are applied to the fuel cell system, and the ohmic internal resistance R can be read on the impedance spectrum tester. m And the total internal resistance of the stack R stack , in which the total internal resistance of the stack R is measured at low frequency stack , high frequency measurement of ohmic internal resistance R m The fuel cell internal resistance model is shown in equations (6) and (7).

[0074] Use the voltmeter connected in parallel in the circuit to get the corresponding battery voltage V st .

[0075] The corresponding anode gas pressure drop is obtained by using the anode gas inlet and outlet pressure sensors. Anode gas pressure drop Expressed as:

[0076]

[0077] Among them, P in is the anode inlet gas pressure, P out is the anode outlet gas pressure.

[0078] Step S503: input the feature data into the trained Bagging neural network ensemble classifier to obtain the battery fault type.

[0079] After obtaining the three characteristic data of the unknown fault stack, namely the battery voltage V st , ohmic internal resistance R m and anode gas pressure drop The new data under different conditions are normalized and then fed into the trained Bagging neural network ensemble classifier model to identify the fault state, determining whether the stack is in a membrane dry state or a flooded state.

[0080] Step S504: executing different recovery strategies according to the identified battery fault type.

[0081] Different fault recovery measures are used for different faults, thereby improving the reliability and durability of proton exchange membrane fuel cells. Specifically:

[0082] If it is a membrane dry fault, reduce the operating temperature of the proton exchange membrane fuel cell, reduce the stoichiometric ratio of the intake air, and increase the relative humidity RH of the intake air;

[0083] If the fault is water flooding, the operating temperature of the proton exchange membrane fuel cell is increased, the stoichiometric ratio of the intake air is increased, and the relative humidity RH of the intake air is reduced.

[0084] After executing the fault recovery strategy, you can continue to monitor whether the battery status returns to normal, which will not be repeated here.

[0085] It should be noted that, without departing from the spirit and scope of the present invention, those skilled in the art may make appropriate changes or modifications to the above embodiments. For example, other existing methods may be used to obtain the predicted voltage and measure the ohmic internal resistance and anode gas pressure drop. For another example, the number and structure of weak learners included in the bagging neural network ensemble classifier can be set according to actual needs and are not limited by the present invention.

[0086] In summary, compared with the prior art, the present invention has the following advantages:

[0087] 1) Existing neural network-based PEMFC fault diagnosis solutions require a large amount of historical data. To address this issue, the present invention introduces bagging ensemble learning to optimize the neural network model. This results in a bagging neural network ensemble model that requires less training data while ensuring accurate and timely model predictions.

[0088] 2) To address the problem that existing solutions can only diagnose membrane drying, the present invention combines the characteristics of the fault type without increasing the amount of feature data. It uses ohmic internal resistance, anode gas pressure drop, and battery voltage as feature data integrated by the bagging neural network, so that the resulting model can diagnose both membrane drying and flooding.

[0089] 3) The present invention applies active fault-tolerant control to the model, which can promptly repair the fuel cell in the membrane dry and flooded states, thereby improving the durability and reliability of the PEMFC.

[0090] The present invention is suitable for various applications requiring PEM fuel cells, including miniature PEM fuel cell portable power supplies, small PEM fuel cell mobile power supplies, medium- and high-power PEM fuel cell power generation systems, and PEM fuel cell systems used as automotive propulsion. Computer simulations have fully verified the feasibility of the present invention, demonstrating favorable results.

[0091] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0092] Computer-readable storage medium can be a tangible device that can keep and store the instructions used by the instruction execution device.Computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device or any suitable combination thereof.More specific examples (non-exhaustive list) of computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove having instructions stored thereon, and any suitable combination thereof.Computer-readable storage medium used herein is not interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagated by waveguides or other transmission media (for example, light pulses by fiber optic cables), or electrical signals transmitted by wires.

[0093] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0094] The computer program instructions for performing the operation of the present invention can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, Python, and conventional procedural programming languages ​​such as "C" language or similar programming languages. The computer readable program instructions can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), is personalized by utilizing the state information of the computer readable program instructions, and the electronic circuit can execute the computer readable program instructions, thereby realizing various aspects of the present invention.

[0095] Various aspects of the present invention are described herein with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0096] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0097] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0098] The flowcharts and block diagrams in the accompanying drawings show the possible implementation architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of an instruction, and the module, program segment or part of the instruction contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that implementation by hardware, implementation by software, and implementation by a combination of software and hardware are all equivalent.

[0099] While various embodiments of the present invention have been described above, the foregoing description is intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technological improvements in the marketplace, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of the present invention is defined by the appended claims.

Claims

1. A proton exchange membrane fuel cell control method comprising the following steps: Using the established battery voltage model to monitor the battery state and obtain a predicted voltage, the battery is a proton exchange membrane fuel cell; comparing the difference between the predicted voltage and the actual measured voltage with a set threshold to determine whether the battery is in an abnormal state; When the battery is determined to be in an abnormal state, the measured battery voltage, ohmic internal resistance and anode gas pressure drop are input into the trained Bagging neural network ensemble classifier to obtain the battery fault type; The Bagging neural network ensemble classifier includes multiple weak learners, each weak learner corresponds to a neural network model, and the neural network model includes an input layer, a hidden layer, and an output layer. The measured battery voltage, ohmic internal resistance, and anode gas pressure drop are used as input variables of the input layer, and the battery fault type is used as the output variable of the output layer. The Bagging neural network ensemble classifier is trained according to the following steps: Take the training set containing M samples as the initial data set, first randomly take out a sample and put it into the sampling set, then put the sample back into the initial data set, after M random sampling operations, a sampling set containing M samples is obtained, and then T sampling sets containing M samples are sampled, where M and T are set integers; Using the obtained T sampling sets to train T weak learners included in the Bagging neural network ensemble classifier, each sampling set corresponds to a weak learner; The T weak learners are combined through voting, and the category with the most votes is used as the predicted battery fault type.

2. The method according to claim 1, characterized in that The number of hidden layer nodes of the neural network model is determined according to the following steps: Set the starting number of nodes X0 and the starting step length h0 of the hidden layer, which are expressed as: X0=log2N in <h2 style=";text-align:left;direction:ltr">h0=(N<h2 style=";text-align:left;direction:ltr"> in <h2 style=";text-align:left;direction:ltr"> -N<h2 style=";text-align:left;direction:ltr"> out <h2 style=";text-align:left;direction:ltr"> ) / 3 Among them, X0 is the number of starting nodes in the hidden layer, N in is the number of input units in the input layer, N out is the number of units output by the output layer; Design the number of comparison nodes X1 in the hidden layer, expressed as: X1=X0+h0 Calculate the neural network model prediction error values ​​f0(x) and f1(x) corresponding to the starting node number and the comparison node number to determine the next comparison node number X2, which is expressed as: After repeated iterations, until the optimal number of nodes X is found i , as the optimal number of hidden layer nodes.

3. The method according to claim 1, characterized in that The battery voltage model is expressed as: V st =N*(E nernst -V act -V ohm -V conc ) Among them, V st is the battery voltage, N is the number of battery cells, E nernst is the Nernst voltage, V act is the activation loss, V ohm is the ohmic loss, V conc is the concentration loss; Among them, the Nernst voltage E nernst Expressed as: Among them, T st is the actual temperature of the battery, and The actual measured partial pressures of hydrogen and oxygen for the battery; Among them, the activation loss V act Expressed as: V act =V0+V a (1-e -10i ) Among them, V0 and V a are the first and second empirical parameters respectively, i is the actual measured current density; Among them, the ohmic loss V ohm Expressed as: Among them, λ m is the membrane water content; Among them, the concentration loss V conc Expressed as: Among them, a is the empirical parameter, i max is the maximum current density.

4. The method according to claim 1, wherein The anode gas pressure drop is expressed as: Among them, P in is the anode inlet gas pressure, P out is the anode outlet gas pressure.

5. The method according to claim 1, wherein The ohmic internal resistance is obtained according to the following steps: Establish the battery equivalent internal resistance model, which is expressed as: R stack =R f +R m +R d Among them, R f is the activation internal resistance, R m is the ohmic internal resistance, R d is the concentration resistance, R stack is the total internal resistance of the battery stack; Electrochemical impedance spectroscopy is used to apply a set of AC potential wave signals with different frequencies and set amplitudes to the battery system, and the ohmic internal resistance R is read on the impedance spectrum tester. m And the total internal resistance of the stack R stack , in which the total internal resistance R of the stack is measured at low frequency stack , high frequency measurement of ohmic internal resistance R m .

6. The method according to claim 1, further comprising: When the battery fault type is a membrane dry fault or a flood fault, perform fault recovery according to the following steps: For membrane dry failure, reduce the battery operating temperature, lower the stoichiometric ratio of the intake air, and increase the relative humidity of the intake air; For flooding faults, increase the operating temperature of the battery, increase the stoichiometric ratio of the intake air and reduce the relative humidity of the intake air.

7. A computer-readable storage medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

8. A computer device comprising a memory and a processor, wherein a computer program capable of being run on the processor is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images