Potential water logging behavior early warning method, device and system, and storage medium

Abstract

The invention discloses a potential water logging behavior early warning method, device and system, and a storage medium, and the method comprises the steps: obtaining the operation data of hydrogen fuel cell reaction water in a diversion type flow channel, including the inlet and outlet pressure drop of an anode and a single cell cathode flow channel, the output voltage of a fuel cell stack, and the output current of the fuel cell stack, and carrying out the preprocessing of the measured data, the method comprises the following steps: dividing a training set and a test set according to a set proportion, constructing an LSTM neural network prediction model based on input real-time monitoring data, training the LSTM model, and then predicting a possible flooding condition based on actual operation data of a fuel cell and performing early warning. By adopting the technical scheme provided by the invention, the judgment of the water content and the prevention of water logging possibly occurring in the future are facilitated, and meanwhile, the early warning threshold is defined, so that the stable operation of the proton exchange membrane fuel cell is ensured on the premise of preventing the water logging of the proton exchange membrane fuel cell.

Description

Technical Field

[0001] The present invention belongs to the technical field of hydrogen fuel cell water management, and in particular relates to a potential flooding behavior early warning method and device, system, and storage medium. Background Art

[0002] Traditional backup power sources use internal combustion engines as generators, which have the disadvantages of low efficiency and environmental pollution. Proton exchange membrane fuel cells (PEMFCs) are devices that convert chemical energy released during the electrochemical reaction of hydrogen (fuel) and oxygen into electrical energy. They have the advantages of fast low-temperature startup, high power density, and green environmental protection. They are considered to be one of the most promising alternative energy sources in the future and are widely used in electric vehicles, drones, portable power sources, and other fields. In actual use, PEMFCs may have various faults, and serious ones may even cause safety accidents. Therefore, fault diagnosis of PEMFCs is an important research direction. Hydrogen fuel cells produce water during the combustion process, and gradually penetrate into the flow channel. If it cannot be discharged in time, it will cause water accumulation. When the water exceeds a certain amount, it will cause flooding.

[0003] Existing flood warning methods mostly use anode gas pressure drop as flood warning diagnosis because the anode contains less water than the cathode, the gas pressure drop changes more significantly, and is more sensitive to flood changes.

[0004] Patent 200410010000.6 invented a flooding diagnosis method for a hydrogen / oxygen proton exchange membrane fuel cell stack. The method determines the occurrence of flooding by testing whether the pressure drop at the inlet and outlet of the fuel cell stack exceeds a threshold. However, corresponding measures can only be taken after flooding occurs, which has a lag and cannot prevent flooding in advance.

[0005] Patent 202311268682.X invented a method for predicting flooding failures of proton exchange membrane fuel cells. It determines possible fuel cell flooding in the future by predicting whether the node voltage of each battery exceeds a threshold, and prevents flooding by blowing water. However, the early warning judgment conditions are not perfect, and misjudgment of non-flooding situations may occur. Blowing water reduces the water content, thereby affecting the conductivity of the fuel cell.

[0006] During actual operation, the water produced by the reaction tends to move toward the cathode, causing more water accumulation at the cathode, making it more likely to flood and affect the performance of the fuel cell. In addition, since the cathode flow channel is often a gas-liquid two-phase flow, when flooding occurs, the gas pressure drop changes less than that at the anode, making it more difficult to judge. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a method and device, system and storage medium for early warning of potential flooding of a diversion flow channel caused by reaction water in a hydrogen fuel cell.

[0008] To achieve the above object, the present invention adopts the following technical solution:

[0009] A potential flooding behavior early warning method, comprising:

[0010] Obtaining the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel, and pre-processing the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel;

[0011] The LSTM model is trained based on the pre-processed hydrogen fuel cell reaction water running data in the diversion channel;

[0012] The actual operation data of the fuel cell is input into the trained LSTM model to predict possible flooding and issue early warnings.

[0013] Preferably, the operating data of the hydrogen fuel cell reaction water in the diversion flow channel include: the inlet and outlet pressure drops of the anode and cathode flow channels of the single-chip battery, the output voltage of the fuel cell stack, and the output current of the fuel cell stack.

[0014] Preferably, the preprocessing includes: data reconstruction, smoothing and denoising, and data normalization.

[0015] The present invention also provides a potential flooding behavior warning device, comprising:

[0016] An acquisition module is used to obtain the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel, and pre-process the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel;

[0017] A training module, used to train the LSTM model based on the pre-processed hydrogen fuel cell reaction water running data in the diversion flow channel;

[0018] The prediction module is used to input the actual operation data of the fuel cell into the trained LSTM model to predict possible flooding and issue early warnings.

[0019] Preferably, the operating data of the hydrogen fuel cell reaction water in the diversion flow channel include: the inlet and outlet pressure drops of the anode and cathode flow channels of the single-chip battery, the output voltage of the fuel cell stack, and the output current of the fuel cell stack.

[0020] Preferably, the preprocessing includes: data reconstruction, smoothing and denoising, and data normalization.

[0021] An embodiment of the present invention further provides a potential flooding behavior warning system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a potential flooding behavior warning method when executed by the processor.

[0022] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes a potential flooding behavior warning method when running.

[0023] The present invention adopts LSTM neural network to warn of potential flooding of diversion flow channels caused by reaction water in hydrogen fuel cells, which helps to judge the water content and prevent possible flooding in the future. At the same time, the warning threshold is defined to ensure the stable operation of proton exchange membrane fuel cells under the premise of preventing flooding of proton exchange membrane fuel cells. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0025] Figure 1 This is a flow chart of a potential flooding behavior warning method according to an embodiment of the present invention;

[0026] Figure 2 It is a principle diagram of a potential flooding behavior early warning method according to an embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of the LSTM structure of an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of the slider window prediction method. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0031] Embodiment 1:

[0032] like Figure 1 , 2 As shown, the present invention provides a method for predicting and warning potential flooding of a flow-guiding flow channel caused by reaction water in a hydrogen fuel cell, comprising the following steps:

[0033] S1. Collect the operating data of the reaction water of the hydrogen fuel cell in the diversion flow channel, and the operating data includes the inlet and outlet pressure drops of the anode and cathode flow channels, the output voltage of the fuel cell, and the output current of the fuel cell. Preprocess the original data, including data reconstruction, smoothing and denoising, data normalization, and division of training sets and test sets; firstly, reconstruct the data of the diversion flow channel operation to reduce the data volume and extract representative typical data; then use the Savitzky-Golay filter (SG filter) and wavelet threshold algorithm to smooth and denoise the data; use Z-score to normalize the feature parameters to convert each input feature to the same order of magnitude, thereby eliminating the influence of the dimension; finally, divide the training set and test set according to the set ratio;

[0034] S2. According to the running data of hydrogen fuel cell reaction water in the flow channel, including the pressure drop of anode and cathode flow channel inlet and outlet, fuel cell output voltage, and fuel cell output current, X t-n , X t-n+1 , ..., X t The above data at the n+1 time point is used as input to build an LSTM neural network model to predict Y t+1 To Y t+m The cathode inlet and outlet pressure drop at the moment, which includes input layer, hidden layer and output layer. The preprocessed data is used as the input layer of the neural network; the hidden layer determines the optimal number of neurons and the number of neuron layers according to the LSTM prediction effect, and determines the optimal activation function, algorithm and subsequent optimization; after learning, the LSTM neural network obtains the predicted data as the output layer, denormalizes it, obtains the real data, and realizes the prediction of the diversion channel operation data;

[0035] S3. The inlet and outlet pressure drop in the cathode flow channel is used as an alarm parameter, and a sliding window method is used for prediction. The input of the prediction model is determined by the sliding window method, and the prediction result is obtained by a direct prediction mechanism. When the predicted value of the inlet and outlet pressure drop exceeds the initial value by 20%, the system will make reasonable self-healing adjustments, reduce the amount of hydrogen introduced, and reduce the intensity of the reaction; when the predicted value of the inlet and outlet pressure drop has not been lower than the initial value for a long time and the predicted value of the output voltage is reduced to 15% of the initial value, the system will alarm.

[0036] As an implementation method of the embodiment of the present invention, in the data reconstruction in the data preprocessing described in step S1, sampling is performed at intervals of one hour, and data points of 10 hours are selected to replace the original data.

[0037] As an implementation method of an embodiment of the present invention, in the smoothing and denoising processing in the data preprocessing described in step S1, the smoothing processing adopts an SG filter, the principle of which is to use the least squares method to smooth the signal through a polynomial function, and can perform filtering operations in the time domain and the frequency domain at the same time; the denoising processing adopts a wavelet threshold algorithm:

[0038] X N =f N +e N

[0039] Among them, f N is the original signal, e N is a Gaussian noise signal.

[0040] As an implementation method of the present invention, the characteristic parameters in the preprocessing described in step S1 are normalized, and the Z-score is standardized, and its conversion function is:

[0041]

[0042] Among them, x n，m is the mth data in the nth feature, u n is the average value of the nth feature in the training set, σ n is the standard deviation of the nth feature in the training set.

[0043] As an implementation manner of an embodiment of the present invention, the division of the data set described in step S1 divides the data into a training set and a test set in a certain ratio to meet the training requirements of the LSTM neural network.

[0044] As an implementation of an embodiment of the present invention, in step S2, the input data of LSTM is a 3D tensor, and the embodiment of the present invention uses the data of the past 20 time steps to predict the data of the next time step. The single sample used to predict the time t can be represented by a 2D tensor of shape (20, 6):

[0045]

[0046] The model is used to predict the cathode inlet and outlet voltage drops and output voltage in the next N time steps. The input model data shape is (N, 20, 6), such as Figure 3 shown.

[0047] As an implementation method of an embodiment of the present invention, the LSTM neural network in step S2, the activation function micro-sigmoid function loss function adopted is selected as the RMSE function, the optimizer is selected as Adam, the number of neurons in the first layer of hidden layer neurons is 64, the number of neurons in the second layer is 32, the learning rate is 0.001, and the maximum number of iterations is 500.

[0048] Furthermore, the Sigmoid function has a limited output range, stable optimization, and is a continuous function, which is easy to derive. The calculation formula is:

[0049]

[0050] Furthermore, RMSE is at the same level as the data, making it easier to perceive the data. The calculation formula is:

[0051]

[0052] As an implementation method of an embodiment of the present invention, the Adam algorithm is used to update the weights and biases of the first-order momentum and the second-order momentum of the historical gradient, and the weight update rule is:

[0053]

[0054] As an implementation method of the present invention, in step S3, the prediction mode of the sliding window has a sliding track length of 40, and the next ten data are predicted, such as Figure 4 As shown, in step S3, the data set x constructed using the sliding window needs to use the first 40 data, which is specifically expressed as:

[0055]

[0056] Further, in step S3, the test data set x is substituted into the constructed ten-step prediction model to obtain a voltage data set y predicted ten steps in advance, which is specifically expressed as:

[0057]

[0058] Further, in step S3, the prediction model for potential flooding of the diversion flow channel caused by the reaction water of the hydrogen fuel cell obtained through the training in step S2 predicts that when the pressure drop at the inlet and outlet of the cathode flow channel increases sharply and exceeds 20% of the original pressure drop in the future, self-healing measures will be taken to prevent the occurrence of flooding in advance; when the pressure drop at the inlet and outlet of the cathode flow channel remains high for a long time and it is predicted that the output voltage of the fuel cell will drop sharply by 15% of the original, the system will sound an alarm.

[0059] Embodiment 2:

[0060] The embodiment of the present invention further provides a potential flooding behavior warning device, comprising:

[0061] An acquisition module is used to obtain the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel, and pre-process the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel;

[0062] A training module, used to train the LSTM model based on the pre-processed hydrogen fuel cell reaction water running data in the diversion flow channel;

[0063] The prediction module is used to input the actual operation data of the fuel cell into the trained LSTM model to predict possible flooding and issue early warnings.

[0064] As an implementation method of an embodiment of the present invention, the operating data of the hydrogen fuel cell reaction water in the guide flow channel includes: the inlet and outlet pressure drops of the anode and cathode flow channels of the single-chip battery, the output voltage of the fuel cell stack, and the output current of the fuel cell stack.

[0065] As an implementation method of the embodiment of the present invention, the preprocessing includes: data reconstruction, smoothing and denoising processing, and data normalization.

[0066] Embodiment 3:

[0067] An embodiment of the present invention further provides a potential flooding behavior warning system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a potential flooding behavior warning method when executed by the processor.

[0068] Embodiment 4:

[0069] An embodiment of the present invention further provides a storage medium, on which a computer program is stored, and the computer program executes a potential flooding behavior warning method when running.

[0070] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the protection scope determined by the claims of the present invention.

Claims

1. A potential flooding behavior early warning method, characterized in that: include: Obtaining the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel, and pre-processing the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel; The LSTM model is trained based on the pre-processed hydrogen fuel cell reaction water running data in the diversion channel; The actual operation data of the fuel cell is input into the trained LSTM model to predict possible flooding and issue early warnings.

2. The potential flooding behavior early warning method according to claim 1, characterized in that: The operating data of hydrogen fuel cell reaction water in the diversion flow channel: including the inlet and outlet pressure drops of the anode and cathode flow channels of the single cell, the output voltage of the fuel cell stack, and the output current of the fuel cell stack.

3. The potential flooding behavior early warning method according to claim 2, characterized in that: The preprocessing includes: data reconstruction, smoothing and denoising, and data normalization.

4. A potential flooding behavior warning device, characterized in that: include: An acquisition module is used to obtain the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel, and pre-process the running data of the hydrogen fuel cell reaction water in the flow-guiding flow channel; A training module, used to train the LSTM model based on the pre-processed hydrogen fuel cell reaction water running data in the diversion flow channel; The prediction module is used to input the actual operation data of the fuel cell into the trained LSTM model to predict possible flooding and issue early warnings.

5. The potential flooding behavior warning device as claimed in claim 4, characterized in that: The operating data of hydrogen fuel cell reaction water in the diversion flow channel: including the inlet and outlet pressure drops of the anode and cathode flow channels of the single cell, the output voltage of the fuel cell stack, and the output current of the fuel cell stack.

6. The potential flooding behavior warning device as claimed in claim 5, characterized in that: The preprocessing includes: data reconstruction, smoothing and denoising, and data normalization.

7. A potential flooding behavior warning system, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the potential flooding behavior warning method as claimed in any one of claims 1 to 3 is executed.

8. A storage medium, characterized in that: The storage medium stores a computer program, which executes the potential flooding behavior warning method according to any one of claims 1 to 3 when running.

Images