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

Abstract

The application discloses a potential water flooding behavior early warning method and device, system and storage medium, and comprises the following steps: obtaining hydrogen fuel cell reaction water in a flow guide type flow channel operation data, including anode and monolithic cell cathode flow channel inlet and outlet pressure drop, fuel cell stack output voltage, fuel cell stack output current, pre-processing the measured data, dividing the training set and the test set in a set proportion, constructing an LSTM neural network prediction model based on the input real-time monitoring data, training the LSTM model, and then predicting and warning the possible water flooding based on the actual fuel cell operation data. The technical scheme of the application is helpful to judge the water content and prevent the future possible water flooding, and the 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 flooding of the proton exchange membrane fuel cell.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of hydrogen fuel cell water management, and particularly relates to a potential water flooding behavior early warning method and device, system and storage medium. BACKGROUND

[0002] Traditional backup power uses internal combustion engines as generators, which has the disadvantages of low efficiency and environmental pollution. Proton exchange membrane fuel cell (PEMFC) is a device that converts chemical energy released in the electrochemical reaction process of hydrogen (fuel) and oxygen into electrical energy, which has the advantages of low temperature start, fast speed, high power density, green environmental protection and the like, and is considered as one of the most promising alternative energy in the future, and is widely used in electric vehicles, unmanned aerial vehicles, portable power sources and the like. In actual use, PEMFC will have various faults, and serious faults will even cause safety accidents, so fault diagnosis of PEMFC is an important research direction. Hydrogen fuel cell generates water in the combustion process, and the water gradually penetrates into the flow channel, if the water cannot be discharged in time, water accumulation will occur, and when the water exceeds a certain amount, water flooding will occur.

[0003] The existing water flooding early warning method adopts anode gas pressure drop as water flooding early warning diagnosis because the anode contains less water than the cathode, the gas pressure drop changes more obviously, and the water flooding changes more sensitively.

[0004] Patent 200410010000.6 invents a water flooding diagnosis method of hydrogen / oxygen proton exchange membrane fuel cell stack, which judges the occurrence of water flooding phenomenon by testing whether the inlet and outlet pressure drops of the stack exceed the threshold value, but can only take corresponding measures after water flooding occurs, which has hysteresis and cannot prevent water flooding in advance.

[0005] Patent 202311268682.X invents a water flooding fault prediction method of proton exchange membrane fuel cell, which judges the future possible fuel cell water flooding by predicting whether the node voltage of each cell exceeds the threshold value, and prevents the occurrence of water flooding by blowing water, but the early warning judgment condition is not perfect, and misjudgment may occur in non-water flooding situation, blowing water reduces water content, and further affects the electrical conductivity of the fuel cell.

[0006] In actual operation, the reaction generated water tends to move to the cathode, causing more water accumulation in the cathode, which is more prone to water flooding and affects the performance of the fuel cell, and since the cathode flow channel is usually gas-liquid two-phase flow, when water flooding occurs, the anode gas pressure drop changes less, so it is more difficult to judge. SUMMARY

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

[0008] To achieve the above object, the application adopts the following technical scheme:

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

[0010] Obtain the hydrogen fuel cell reaction water in the flow guide type flow channel operation data, and pretreat the hydrogen fuel cell reaction water in the flow guide type flow channel operation data;

[0011] According to the pretreated hydrogen fuel cell reaction water in the flow guide type flow channel operation data, train an LSTM model;

[0012] Input the actual operation data of the fuel cell into the trained LSTM model, predict and make early warning on the possible water flooding situation.

[0013] Preferably, the hydrogen fuel cell reaction water in the flow guide type flow channel operation data includes the anode and monolithic cell cathode flow channel inlet and outlet pressure drop, fuel cell stack output voltage and fuel cell stack output current.

[0014] Preferably, the pretreatment includes data reconstruction, smoothing and denoising processing, and data normalization.

[0015] The application also provides a potential water flooding behavior early warning device, comprising:

[0016] An acquisition module is configured to obtain the hydrogen fuel cell reaction water in the flow guide type flow channel operation data, and pretreat the hydrogen fuel cell reaction water in the flow guide type flow channel operation data;

[0017] A training module is configured to train an LSTM model according to the pretreated hydrogen fuel cell reaction water in the flow guide type flow channel operation data;

[0018] A prediction module is configured to input the actual operation data of the fuel cell into the trained LSTM model, predict and make early warning on the possible water flooding situation.

[0019] Preferably, the hydrogen fuel cell reaction water in the flow guide type flow channel operation data includes the anode and monolithic cell cathode flow channel inlet and outlet pressure drop, fuel cell stack output voltage and fuel cell stack output current.

[0020] Preferably, the pretreatment includes data reconstruction, smoothing and denoising processing, and data normalization.

[0021] The application also provides a potential water flooding behavior early warning system, comprising a memory and a processor, wherein the memory stores a computer program run by the processor, and the computer program performs the potential water flooding behavior early warning method when run by the processor.

[0022] The embodiment of the present application also provides a storage medium, which stores a computer program, and the computer program performs the potential water flooding behavior early warning method when running.

[0023] The present application adopts an LSTM neural network to early warn potential water flooding behavior of a hydrogen fuel cell reaction water caused by a flow guide type flow channel, which is helpful for judging water content and preventing future possible water flooding, and the early warning threshold is defined to ensure stable operation of the proton exchange membrane fuel cell under the premise of preventing water flooding of the proton exchange membrane fuel cell. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only constitute the embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0025] Figure 1 The flow chart of the potential water flooding behavior early warning method of the embodiment of the present application;

[0026] Figure 2 The principle diagram of the potential water flooding behavior early warning method of the embodiment of the present application;

[0027] Figure 3 The schematic diagram of the LSTM structure of the embodiment of the present application;

[0028] Figure 4 The schematic diagram of the sliding window prediction mode. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

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

[0031] Embodiment 1:

[0032] As shown in Figure 1 , 2 The present application provides a potential water flooding behavior early warning method of a hydrogen fuel cell reaction water caused by a flow guide type flow channel, which comprises the following steps:

[0033] S1, collect the operation data of the hydrogen fuel cell reaction water in the flow guide flow channel, the operation data including the pressure drop of the anode and cathode flow channel inlet and outlet, the fuel cell output voltage, and the fuel cell output current. The raw data is preprocessed, specifically including data reconstruction, smoothing and denoising processing, data normalization, division of the training set and test set; the operation data of the flow guide flow channel is first reconstructed to reduce the data amount, and representative typical data is extracted; then the Savitzky-Golay filter (S-G filter) and wavelet threshold algorithm are used to smooth and denoise the data; the Z-score is used for parameter feature normalization to convert each input feature to the same order of magnitude, so as to eliminate the influence of the dimension; finally, the training set and test set are divided according to the set proportion;

[0034] S2, according to the operation data of the hydrogen fuel cell reaction water in the flow guide flow channel, including the pressure drop of the anode and cathode flow channel inlet and outlet, the fuel cell output voltage, and the fuel cell output current, X t-n , X t-n+1 , …, X t The above data at the n+1 time points are taken as inputs to construct an LSTM neural network model to predict the cathode inlet and outlet pressure drop at Y t+1 , Y t+m , which includes an input layer, a hidden layer, and an output layer. The preprocessed data is used as the input layer of the neural network; the number of neurons and the number of layers of the hidden layer are determined according to the prediction effect of the LSTM, and the best activation function, algorithm, and subsequent optimization are determined; after learning, the LSTM neural network obtains the prediction data as the output layer, which is subjected to inverse normalization to obtain the real data, thereby realizing the prediction of the operation data of the flow guide flow channel.

[0035] S3, the inlet and outlet pressure drop of the cathode flow channel is taken as an alarm parameter, and a sliding window is used for prediction; the input of the prediction model is determined by using a sliding window, and the prediction result is obtained by using a direct prediction mechanism. When the predicted value of the inlet and outlet pressure drop exceeds the initial value by 20%, the system performs reasonable self-healing adjustment, which reduces the amount of hydrogen gas input and slows down the reaction intensity; when the predicted value of the inlet and outlet pressure drop has not been below 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 embodiment of the present application, the data reconstruction in the data preprocessing of step S1 takes one hour as an interval for sampling, and 10 hours of data points are selected instead of the original data.

[0037] As an embodiment of the present application, the smoothing and denoising processing in the data preprocessing of step S1, the smoothing processing adopts S-G filter, the principle is to use least square method, the signal is smoothed by polynomial function, and the filtering operation can be performed in time domain and frequency domain at the same time; the denoising processing adopts wavelet threshold algorithm:

[0038] X N =f N +e N

[0039] Wherein, f N is the original signal, e N is the Gaussian noise signal.

[0040] As an embodiment of the present application, the feature parameter normalization in the preprocessing of step S1, the Z-score standardization, the conversion function is:

[0041]

[0042] Wherein, wherein, 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, and sigma n is the standard deviation of the nth feature in the training set.

[0043] As an embodiment of the present application, the data set division of step S1, the data is divided into training set and test set according to certain proportion, so as to meet the training requirement of LSTM neural network.

[0044] As an embodiment of the present application, in step S2, the input data of LSTM is a 3D tensor, and the data of the past 20 time steps is used to predict the data of the next time step in the embodiment of the present application, and a 2D tensor with shape (20, 6) can be used to represent a single sample for predicting t time:

[0045]

[0046] The model is used to predict the future N time steps of the cathode import and export pressure drop and output voltage respectively, and the input model data shape is (N, 20, 6), as shown in Figure 3 .

[0047] As an embodiment of the present application, the LSTM neural network of step S2, the activation function adopts sigmoid function, the loss function selects RMSE function, the optimizer selects Adam, the number of first layer neurons in hidden layer is 64, the number of second layer neurons is 32, the learning rate is 0.001, and the maximum iteration number is 500.

[0048] Further, the Sigmoid function output range is limited, the optimization is stable, and it is a continuous function, which is easy to derive, and the calculation formula is:

[0049]

[0050] Further, the RMSE is of the same level as the data, and it is easier to perceive the data, and the calculation formula is:

[0051]

[0052] As an embodiment of the present application, the first order momentum and the second order momentum of the historical gradient are used to update the weight and the bias using the Adam algorithm, and the update rule of the weight is:

[0053]

[0054] As an embodiment of the present application, in step S3, the prediction mode of the sliding window has a slide rail length of 40, and the last ten data are predicted, as shown in Figure 4 In step S3, the data set x constructed by the sliding window needs to use the first 40 data, and is specifically represented as:

[0055]

[0056] Further, in step S3, the test data set x is substituted into the constructed ten-step prediction model, and the ten-step predicted voltage data set y can be obtained, and is specifically represented as:

[0057]

[0058] Further, in step S3, the hydrogen fuel cell reaction water causes the potential water flooding behavior prediction model of the flow guide type flow channel obtained by training in step S2, and when the future cathode flow channel inlet and outlet pressure drop sharply increases by more than 20% of the original pressure drop, the self-healing means will be taken to prevent the occurrence of water flooding phenomenon in advance; when the cathode flow channel inlet and outlet pressure drop is high and the fuel cell output voltage is predicted to sharply decrease by 15% of the original, the system will issue an alarm.

[0059] Embodiment 2:

[0060] The present application also provides a potential water flooding behavior warning device, comprising:

[0061] The acquisition module is used to obtain the hydrogen fuel cell reaction water in the flow guide type flow channel running data, and simultaneously pre-process the hydrogen fuel cell reaction water in the flow guide type flow channel running data;

[0062] The training module is configured to train an LSTM model according to the preprocessed hydrogen fuel cell reaction water in the flow guide type flow channel operation data.

[0063] The prediction module is configured to input the actual fuel cell operation data into the trained LSTM model, predict the possible water flooding and give a warning.

[0064] As an embodiment of the present application, the hydrogen fuel cell reaction water in the flow guide type flow channel operation data includes the anode and monolithic cell cathode flow channel inlet and outlet pressure drop, fuel cell stack output voltage and fuel cell stack output current.

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

[0066] Embodiment 3

[0067] The present application also provides a potential water flooding behavior warning system, which comprises a memory and a processor, the memory stores a computer program which is run by the processor, and the computer program executes the potential water flooding behavior warning method when run by the processor.

[0068] Embodiment 4

[0069] The present application also provides a storage medium, which stores a computer program, and the computer program executes the potential water flooding behavior warning method when run.

[0070] The above embodiments are only descriptions of the preferred modes of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements of the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A method for early warning of potential flooding behavior, characterized in that, include: Data on the operation of hydrogen fuel cell reaction water in a guided flow channel is obtained, and this data is preprocessed. The data includes: inlet and outlet pressure drops of the anode and single-cell cathode flow channels, fuel cell stack output voltage, and fuel cell stack output current. The preprocessing includes: data reconstruction, smoothing and noise reduction, and data normalization. An LSTM model was trained based on the data of pretreated hydrogen fuel cell reaction water operating in a guided flow channel; where X... t-n X t-n+1 , ..., X t Using the pretreated hydrogen fuel cell reactor water operating data in the guided flow channel at these n+1 time points as input, an LSTM neural network model is constructed to predict Y. t+1 To Y t+m The inlet and outlet pressure drop of the cathode flow channel and the output voltage of the hydrogen fuel cell stack at any given time; The system inputs actual fuel cell operating data into a trained LSTM model to predict and warn of potential flooding. Specifically, when it predicts that the pressure drop at the cathode inlet and outlet will increase by more than 20% of the original pressure drop, it will initiate self-healing measures to prevent flooding in advance. When it predicts that the pressure drop at the cathode inlet and outlet will not fall below the original pressure drop for an extended period and that the fuel cell stack output voltage will drop to 15% of the original voltage, the system will issue an alarm.

2. A potential flooding behavior early warning device for implementing the potential flooding behavior early warning method of claim 1, characterized in that, include: The acquisition module is used to obtain the operating data of hydrogen fuel cell reaction water in the guide channel, and at the same time preprocess the operating data of hydrogen fuel cell reaction water in the guide channel. The training module is used to train an LSTM model based on the data of the pretreated hydrogen fuel cell reaction water running in the guided flow channel. The prediction module is used to input the actual operating data of the fuel cell into the trained LSTM model to predict and issue early warnings about possible flooding.

3. A potential flooding behavior early warning system, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executed by the processor, the computer program performing the potential flooding behavior warning method as described in claim 1 when executed by the processor.

4. A storage medium, characterized in that, The storage medium stores a computer program that, when running, executes the potential flooding behavior early warning method as described in claim 1.

Images