Method and apparatus for obtaining prediction model, prediction method, electronic device and medium

Abstract

The application relates to the field of fuel cells and discloses a method and device for obtaining a prediction model, a prediction method, electronic equipment and a medium. The method for obtaining a fuel cell life prediction model comprises the following steps: obtaining P pieces of real operation data corresponding to P fuel cells under test working conditions, wherein P is a positive integer; constructing a fuel cell stack simulation model, obtaining Q pieces of simulation operation data corresponding to Q fuel cells under simulated test working conditions through the stack simulation model, wherein Q is a positive integer; obtaining original data according to the P pieces of real operation data and the Q pieces of simulation operation data, obtaining training data according to the original data; training a preset model through the training data to obtain a fuel cell life prediction model. According to the method for obtaining the fuel cell life prediction model, the test cost and test time cost for fuel cell life prediction can be reduced.

Description

Technical Field

[0001] This application relates to the field of fuel cells, and more particularly to a method, apparatus, prediction method, electronic device, and medium for obtaining a prediction model. Background Technology

[0002] Fuel cells, especially proton exchange membrane fuel cells (PEMFCs), are widely used in the transportation energy sector due to their advantages such as high efficiency and environmental friendliness. Fuel cell lifespan is affected by multiple factors, including material degradation and fluctuations in operating conditions; therefore, accurately predicting the remaining lifespan of fuel cells is the core of fuel cell health management.

[0003] In related technologies, fuel cell life prediction schemes require long-term operation of the fuel cell stack to obtain the stack's operating parameters throughout its entire life cycle. Such fuel cell life prediction schemes have high testing costs and testing time costs. Summary of the Invention

[0004] This application provides a method, apparatus, prediction method, electronic device, and medium for obtaining a prediction model, which facilitates the reduction of testing costs and testing time costs for fuel cell life prediction.

[0005] In a first aspect, embodiments of this application provide a method for obtaining a fuel cell life prediction model, comprising: obtaining P real operating data corresponding to P fuel cells under test conditions, where P is a positive integer; constructing a fuel cell stack simulation model, and obtaining Q simulated operating data corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; obtaining raw data based on the P real operating data and the Q simulated operating data, and obtaining training data based on the raw data; and training a preset model using the training data to obtain a fuel cell life prediction model.

[0006] According to some of the foregoing embodiments of the first aspect of this application, obtaining P real operating data corresponding to P fuel cells under test conditions includes: collecting the real values ​​of preset operating parameters of each fuel cell at multiple time points at a preset sampling frequency to obtain the real operating data of each fuel cell in a preset time series; obtaining Q simulated operating data corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model includes: obtaining the predicted values ​​of the preset operating parameters of each fuel cell in the preset time series through the fuel cell stack simulation model to obtain the simulated operating data.

[0007] According to some of the aforementioned embodiments of the first aspect of this application, the preset operating parameters of the real operating data and the simulated operating data include: output voltage, output current, output current density, relative humidity of the anode and cathode air intakes, and operating temperature.

[0008] According to some of the foregoing embodiments of the first aspect of this application, the method for obtaining a fuel cell life prediction model further includes: after obtaining Q sets of simulation operation data, determining whether the fuel cell stack simulation model is accurate based on at least one set of real operation data and at least one set of simulation operation data; if the fuel cell stack simulation model is determined to be accurate, retaining the Q sets of simulation operation data; if the fuel cell stack simulation model is determined to be inaccurate, repeatedly performing the steps of adjusting the fuel cell stack simulation model, obtaining the Q sets of simulation operation data, and determining whether the fuel cell stack simulation model is accurate, until the fuel cell stack simulation model is determined to be accurate.

[0009] According to some of the foregoing embodiments of the first aspect of this application, determining whether the fuel cell stack simulation model is accurate based on at least one of the actual operating data and at least one of the simulated operating data includes: obtaining error information of one of the simulated operating data relative to one of the actual operating data; if the error information meets a preset error requirement, determining that the fuel cell stack simulation model is accurate; if the error information does not meet the preset error requirement, determining that the fuel cell stack simulation model is inaccurate.

[0010] According to some embodiments of the first aspect of this application, the real operating data includes the real value of the output voltage of the fuel cell over a preset time series, and the simulated operating data includes the predicted value of the output voltage of the fuel cell over the preset time series; obtaining error information of one of the simulated operating data relative to one of the real operating data includes: obtaining the real service life of the fuel cell based on the real value of the output voltage of the real operating data, and obtaining the predicted service life of the fuel cell based on the predicted value of the output voltage of the simulated operating data; obtaining the prediction error based on the predicted service life and the real service life; the error information satisfies a preset error requirement, including: the prediction error is less than a first threshold.

[0011] According to some of the foregoing embodiments of the first aspect of this application, obtaining error information of one of the simulated running data relative to one of the real running data further includes: calculating the root mean square error and coefficient of determination of the true value of the output voltage of the real running data and the predicted value of the output voltage of the simulated running data; the error information satisfies preset error requirements, including: the prediction error is less than a first threshold, the root mean square error is less than a second threshold, and the coefficient of determination is greater than a third threshold.

[0012] According to some of the aforementioned embodiments of the first aspect of this application, the ratio of P to Q is 2 / 8 to 4 / 6.

[0013] According to some of the foregoing embodiments of the first aspect of this application, obtaining training data from the original data includes: performing noise reduction processing on the original data to obtain noise-reduced data; dividing the noise-reduced data into a training set, a validation set, and a test set; and performing data standardization processing on the training set, the validation set, and the test set to obtain training data.

[0014] According to some of the foregoing embodiments of the first aspect of this application, the step of performing noise reduction processing on the original data to obtain noise-reduced data includes: performing wavelet threshold noise reduction processing on the original data to obtain noise-reduced data.

[0015] According to some of the foregoing embodiments of the first aspect of this application, the step of training a preset model with the training data to obtain a fuel cell life prediction model includes: training a preset long short-term memory neural network model with the training data to obtain a fuel cell life prediction model.

[0016] Secondly, embodiments of this application provide a battery life prediction method, which includes: obtaining a fuel cell life prediction model by means of a fuel cell life prediction model acquisition method according to any of the foregoing embodiments of the first aspect of this application; and obtaining the remaining life information of the fuel cell through the fuel cell life prediction model.

[0017] Thirdly, embodiments of this application provide a device for acquiring a fuel cell life prediction model, comprising: a sampling module for acquiring P real operating data corresponding to P fuel cells under test conditions, where P is a positive integer; a simulation module for constructing a fuel cell stack simulation model and acquiring Q simulated operating data corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; a data processing module for obtaining raw data based on the P real operating data and the Q simulated operating data, and obtaining training data based on the raw data; and a training module for training a preset long short-term memory neural network model using the training data to obtain a fuel cell life prediction model.

[0018] Fourthly, embodiments of this application provide an electronic device, which includes a memory and a processor. The memory is communicatively connected to the processor. The memory stores instructions, and the processor invokes the instructions in the memory to cause the electronic device to execute a method for obtaining a fuel cell lifetime prediction model according to any of the foregoing embodiments of the first aspect of this application.

[0019] Fifthly, embodiments of this application provide a computer-readable storage medium storing instructions that, when executed by a processor, implement a method for obtaining a fuel cell lifetime prediction model according to any of the foregoing embodiments of the first aspect of this application.

[0020] According to the fuel cell life prediction model acquisition method of this application embodiment, P real operating data points corresponding to P fuel cells under test conditions are obtained, where P is a positive integer; a fuel cell stack simulation model is constructed, and Q simulated operating data points corresponding to Q fuel cells under simulated test conditions are obtained through the fuel cell stack simulation model, where Q is a positive integer; raw data is obtained based on the P real operating data points and the Q simulated operating data points, and training data is obtained based on the raw data points; a preset model is trained using the training data to obtain the fuel cell life prediction model. The obtained fuel cell life prediction model can be used to predict the life of fuel cells. In the above scheme, the raw data used to obtain the training data is obtained from real operating data and simulated operating data. Simulated operating data can replace a large amount of real operating data, thereby reducing the testing cost and testing time cost for fuel cell life prediction. Attached Figure Description

[0021] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0022] Figure 1 This is a flowchart of one embodiment of the method for obtaining a fuel cell lifetime prediction model according to this application; Figure 2 This is a flowchart illustrating the process of obtaining training data from raw data in one embodiment of the fuel cell life prediction model acquisition method according to this application; Figure 3 This is a schematic diagram of the original data and the noise-reduced data in one embodiment of the fuel cell life prediction model acquisition method according to this application; Figure 4 This is a schematic diagram of the structure of neurons in a long short-term memory neural network layer in one embodiment of the fuel cell lifetime prediction model acquisition method according to this application; Figure 5 This is a schematic diagram of an embodiment of the apparatus for obtaining a fuel cell lifetime prediction model according to this application; Figure 6 This is a schematic diagram of the hardware structure of an embodiment of the electronic device according to this application. Detailed Implementation

[0023] The technical solutions in the embodiments (or "implementations") of this application will be clearly and completely described herein with reference to the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements.

[0024] If the embodiments of this application contain terms relating to directional indications or positional relationships (such as up, down, left, right, front, back, inside, outside, top, bottom, center, vertical, horizontal, longitudinal, transverse, length, width, counterclockwise, clockwise, axial, radial, circumferential, etc.), such terms are only used to explain the relative positional relationships and movement of the components in a specific posture (as shown in the attached figures); if the specific posture changes, the directional indications or positional relationships will also change accordingly. Furthermore, the terms "first" and "second" used in the embodiments of this application are only for descriptive convenience and should not be construed as indicating or implying relative importance.

[0025] This application provides a method for obtaining a fuel cell lifetime prediction model. Figure 1 This is a flowchart of one embodiment of the method for obtaining a fuel cell lifetime prediction model according to this application. The method for obtaining the fuel cell lifetime prediction model includes steps S110 to S140.

[0026] In step S110, P real operating data corresponding to P fuel cells under test conditions are obtained, where P is a positive integer.

[0027] In some embodiments, the test condition is a durability test condition. This durability test condition is defined, for example, by limiting the following operating environment parameters: ambient temperature of the fuel cell, air-side humidity, anode and cathode inlet pressure, hydrogen excess coefficient, air excess coefficient, gas pipeline diameter, flow channel volume, etc.

[0028] In some embodiments, acquiring P real operating data points corresponding to P fuel cells under test conditions includes: collecting the real values ​​of preset operating parameters of each fuel cell at multiple time points using a preset sampling frequency, and obtaining the real operating data of each fuel cell over a preset time series. In some embodiments, the sampled data is used to form a time-performance degradation trajectory baseline, which facilitates the resolution of operating condition distortion problems in subsequent fuel cell stack simulation models and can also provide data support for the verification of fuel cell stack simulation models.

[0029] In one example, the preset operating parameters for actual operating data include: output voltage, output current, output current density, relative humidity of the anode and cathode intake air, and operating temperature.

[0030] In step S120, a fuel cell stack simulation model is constructed, and Q simulation operation data corresponding to Q fuel cells under simulated test conditions are obtained through the stack simulation model, where Q is a positive integer. In some embodiments, the fuel cell stack simulation model is constructed using a visual simulation modeling tool. The stack simulation model integrates electrochemical, thermodynamic, and hydrodynamic models to quantify the key factors of stack lifespan degradation. Electrochemically, time-varying parameters are used to simulate catalyst activity decline and membrane aging; thermodynamically, the effect of high temperature on material aging is calculated.

[0031] In some embodiments, obtaining Q simulation operation data corresponding to Q fuel cells under simulated test conditions through a fuel cell stack simulation model includes: obtaining the predicted values ​​of preset operation parameters of each fuel cell in a preset time series through the fuel cell stack simulation model to obtain simulation operation data.

[0032] The simulation data and the real operation data use the values ​​of preset operation parameters under the same working condition (test condition) and the same time series (preset time series).

[0033] In some embodiments, the preset operating parameters of real operating data and simulated operating data include: output voltage, output current, output current density, relative humidity of anode and cathode intake air, and operating temperature.

[0034] Optionally, in some embodiments, the method for obtaining the fuel cell life prediction model further includes: after obtaining Q simulation operation data, determining whether the fuel cell stack simulation model is accurate based on at least one real operation data and at least one simulation operation data; if the fuel cell stack simulation model is determined to be accurate, retaining the Q simulation operation data; if the fuel cell stack simulation model is determined to be inaccurate, repeatedly performing the steps of adjusting the fuel cell stack simulation model, obtaining the Q simulation operation data, and determining whether the fuel cell stack simulation model is accurate, until the fuel cell stack simulation model is determined to be accurate.

[0035] Specifically, determining whether the fuel cell stack simulation model is accurate based on at least one real operating data and at least one simulated operating data includes: obtaining error information of one of the simulated operating data relative to one of the real operating data; if the error information meets the preset error requirements, the fuel cell stack simulation model is determined to be accurate; if the error information does not meet the preset error requirements, the fuel cell stack simulation model is determined to be inaccurate.

[0036] In this embodiment, the real operating data includes the actual value of the output voltage of the fuel cell over a preset time series, and the simulated operating data includes the predicted value of the output voltage of the fuel cell over a preset time series.

[0037] For fuel cells, a failure voltage threshold is set. When the output voltage of the fuel cell drops below the failure voltage threshold, the fuel cell is considered to have reached the end of its life.

[0038] In some embodiments, any simulation run data is taken, and the last one in terms of generation time among Q simulation run data is taken to calculate the error information between the two.

[0039] The preset time series includes the failure time point of the fuel cell in the time span. That is, for real operating data, the preset time series includes the time point when the actual value of the output voltage changes from greater than the failure voltage threshold to less than or equal to the failure voltage threshold in the time span. For simulated operating data, the preset time series includes the time point when the predicted value of the output voltage changes from greater than the failure voltage threshold to less than or equal to the failure voltage threshold in the time span.

[0040] In some embodiments, obtaining error information of one of the simulated running data relative to one of the real running data includes: obtaining the real service life of the fuel cell based on the true value of the output voltage of the real running data, and obtaining the predicted service life of the fuel cell based on the predicted value of the output voltage of the simulated running data; obtaining the prediction error based on the predicted service life and the real service life; the error information meets preset error requirements, including: the prediction error is less than a first threshold.

[0041] In one example, the actual lifespan of the fuel cell is the duration from the start of the preset time series to the failure time of the fuel cell in real operating data; the predicted lifespan of the fuel cell is the duration from the start of the preset time series to the failure time of the fuel cell in simulated operating data.

[0042] In one example, the prediction error is obtained based on the predicted service life and the actual service life, including the prediction error obtained according to the following formula: ; In the above formula, For prediction error, To predict service life, This represents the actual service life.

[0043] This prediction error is used to measure the deviation between the predicted service life and the actual service life.

[0044] In one example, the error information meets the preset error requirements, including: the prediction error is less than 5%.

[0045] In some embodiments, obtaining error information of one of the simulated running data relative to one of the real running data further includes: calculating the root mean square error and coefficient of determination of the true value of the output voltage of the real running data and the predicted value of the output voltage of the simulated running data; the error information meets preset error requirements, including: the prediction error is less than a first threshold, the root mean square error is less than a second threshold, and the coefficient of determination is greater than a third threshold.

[0046] In one example, the root mean square error of the true value of the output voltage from real-world operating data and the predicted value of the output voltage from simulated operating data is calculated using the following formula: ; In the above formula, The root mean square error, The total number of sampling points for the preset time series. For the simulation run data of the first The predicted value of the output voltage at each sampling point. The first for real running data The true value of the output voltage at each sampling point.

[0047] The root mean square error (RMSE) is used to measure the deviation between the predicted and actual output voltage values. The smaller the RMSE value, the more accurate the fuel cell stack simulation model.

[0048] In one example, the coefficient of determination for the true value of the output voltage from real-world operating data and the predicted value of the output voltage from simulated operating data is calculated using the following formula: ; In the above formula, As the coefficient of determination, The total number of sampling points for the preset time series. For the simulation run data of the first The predicted value of the output voltage at each sampling point. The first for real running data The true value of the output voltage at each sampling point For real running data The average of the true values ​​of the output voltage at each sampling point.

[0049] The coefficient of determination is used to measure how well the fuel cell stack simulation model fits the predicted output voltage to the actual value. The coefficient of determination is less than or equal to 1. The closer the coefficient of determination is to 1, the better the fuel cell stack simulation model fits the predicted output voltage to the actual value.

[0050] In one example, the error information meets the preset error requirements, including: prediction error less than 5%, root mean square error less than 0.2, and coefficient of determination greater than 0.9.

[0051] In step S130, raw data is obtained based on P real running data and Q simulated running data, and training data is obtained based on the raw data.

[0052] In some embodiments, the ratio of P to Q is 2 / 8 to 4 / 6, meaning that the ratio of actual running data to simulated running data in the original data is 2 / 8 to 4 / 6. In one example, the ratio of P to Q is 3 / 7, meaning that the ratio of actual running data to simulated running data in the original data is 3 / 7.

[0053] The original data contains a large amount of noisy data, which can easily cause the subsequent fuel cell life prediction model to fail to accurately identify the data trend.

[0054] Figure 2 This is a flowchart illustrating the process of obtaining training data from raw data in one embodiment of the fuel cell life prediction model acquisition method according to this application. In some embodiments, obtaining training data from raw data includes steps S131 to S133.

[0055] In step S131, the original data is denoised to obtain denoised data.

[0056] In some embodiments, step S131, which involves denoising the original data to obtain denoised data, includes: performing wavelet threshold denoising on the original data to obtain denoised data.

[0057] In one example, wavelet thresholding denoising specifically includes: wavelet decomposition, selecting an appropriate wavelet basis function based on signal characteristics to perform multi-level wavelet decomposition of the original signal; thresholding, applying threshold rules to the high-frequency components of each level; and inverse wavelet transform, performing wavelet reconstruction layer by layer to obtain the denoised data. In one example, the parameter settings for wavelet thresholding denoising are as follows: the wavelet basis function is Daubechies 5 (db5) wavelet; the decomposition level is 8 levels; the denoising method is the minmax thresholding method; the threshold rule is the soft thresholding rule; and the thresholds for each level are the default values.

[0058] In the example above, the degradation signal of the fuel cell output voltage is a slow-changing low-frequency trend superimposed with mid-to-high-frequency sensor noise. A decomposition level of 8 layers is sufficient to effectively separate the degradation trend from the noise. Too many decomposition levels would increase computational complexity and potentially lead to over-smoothing and loss of local degradation features, hindering subsequent models from identifying degradation details. An 8-layer decomposition level strikes a balance between noise reduction and feature preservation, adapting to the learning requirements of subsequent preset models for temporal degradation features.

[0059] In the example above, a soft thresholding rule is used, which shrinks high-frequency coefficients instead of setting them to zero as in the hard thresholding rule. The signal processed by the soft thresholding rule is smoother, avoiding the pseudo-Gibbs oscillation phenomenon caused by the hard thresholding rule. This makes the voltage degradation trend more continuous, better conforming to the physical law of monotonic performance degradation in fuel cells, and facilitating the subsequent learning of a stable degradation trajectory by the pre-set model.

[0060] In the example above, the wavelet basis function chosen is the Daubechies 5 (db5) wavelet, which has tight support, orthogonality and appropriate vanishing moments. It is suitable for processing non-stationary, slowly varying signals such as fuel cell output voltage and can better characterize the local abrupt changes in degradation trends.

[0061] Figure 3 This is a schematic diagram of the original data and the noise-reduced data in one embodiment of the fuel cell lifetime prediction model acquisition method according to this application. Figure 3 The original data is time-series data; the output voltage from the original data will be used as an example for explanation. Figure 3 This demonstrates that the denoised data has removed significantly more noise compared to the original data.

[0062] In step S132, the denoised data is divided into a training set, a validation set, and a test set.

[0063] In one example, the denoised data is divided into training, validation, and test sets in a 6:2:2 ratio.

[0064] In step S133, the training set, validation set, and test set are standardized to obtain training data.

[0065] In some embodiments, data standardization is performed on the training set, validation set, and test set to obtain training data, including: zero-mean (Z-score) standardization is performed on the training set, validation set, and test set to obtain training data.

[0066] Taking the training set as an example, the transformation function for standardization when performing zero-mean standardization is: ; In the above formula, where This refers to the m-th unstandardized data in the n-th feature; This refers to the m-th standardized data point in the n-th feature; The average value of the nth feature in the training set; Let be the standard deviation of the nth feature in the training set.

[0067] In step S140, the preset model is trained using training data to obtain a fuel cell life prediction model.

[0068] In some embodiments, step S140 includes: training a preset Long Short-Term Memory (LSTM) neural network model using training data to obtain a fuel cell life prediction model.

[0069] The fuel cell life prediction model is obtained by training a pre-set long short-term memory neural network model. The long short-term memory neural network model is more sensitive to time series information, which helps to improve the accuracy of fuel cell life prediction.

[0070] In some embodiments, a fuel cell life prediction model is obtained by training a preset long short-term memory neural network model with training data, including: constructing a long short-term memory neural network model, wherein the long short-term memory neural network model includes an input layer, at least one long short-term memory neural network layer, and a fully connected layer, and each neuron of the long short-term memory neural network layer includes a forget gate, an input gate, and an output gate; and training the long short-term memory neural network model with training data to obtain the fuel cell life prediction model.

[0071] Traditional recurrent neural network (RNN) models struggle to learn dependencies with long time intervals. Long short-term memory (LSTM) neural network models use a "gating mechanism" to control the forgetting, updating, and output of cell state information.

[0072] In one example, the Long Short-Term Memory (LSTM) neural network model includes an input layer, two LSM layers, and a fully connected layer. The first LSM layer has 256 neurons, and the second LSM layer has 128 neurons; the training epochs are 100, the batch size is 64, the initial learning rate is 0.01, and the optimizer is set to Adaptive Moment Estimation (Adam).

[0073] Figure 4 This is a schematic diagram of the structure of neurons in a long short-term memory neural network layer in one embodiment of the method for obtaining a fuel cell lifetime prediction model according to this application. In some embodiments, the forgetting gate is configured to: obtain a forgetting factor by processing the hidden state at the previous time step and the model input data value at the current time step through a sigmoid activation function; and obtain filtered information based on the cell state at the previous time step and the forgetting factor.

[0074] The forget gate controls the cell state at the previous moment. Perform selective filtering. In one example, based on the hidden state of the previous time step. and the model input data value at the current time Through the sigmoid activation function Processing yields the forgetting factor. Based on the cell state at the previous moment and forgetting factor , and obtain the filtered information.

[0075] Forgetting factor The expression is: ; In the above formula, It is a sigmoid activation function; Here is the recursive weight matrix for the forget gate. Input the weight matrix into the forget gate; This is the forget gate bias vector.

[0076] In some embodiments, the input gate is configured to: obtain an input gate control signal by processing the hidden state of the previous time step and the model input data value of the current time step through a sigmoid activation function; obtain candidate memories by processing the hidden state of the previous time step and the model input data value of the current time step through a hyperbolic tangent (tanh) activation function; obtain information to be added based on the input gate control signal and the candidate memories; and add the information to be added to the filtered information to obtain the cell state of the current time step.

[0077] Allows information to be added to the cell state. It is determined by two parts, namely the input gate control signal. and candidate memories .

[0078] Based on the hidden state of the previous moment and the model input data value at the current time Through the sigmoid activation function Processing yields the input gate control signal. That is, the input gate control signal for: ; In the above formula, It is a sigmoid activation function; The input gate control signal recursive weight matrix, Input a weight matrix to the input gate control signal; This is the bias vector for the input gate control signal.

[0079] Based on the hidden state of the previous moment and the model input data value at the current time Candidate memories are obtained by processing them with the hyperbolic tangent activation function tanh. Candidate memories for: ; In the above formula, It is the hyperbolic tangent activation function; The candidate memory recursive weight matrix, Input a weight matrix for the candidate memory; is the candidate memory bias vector.

[0080] Therefore, based on the input gate control signal and candidate memory, the information to be added is obtained, which is: ; In the above formula, This is the Hadamard product (element-wise product).

[0081] Information to be added After adding the filtered information, the cell state at the current moment is obtained. The expression for updating cell state is: ; In the above formula, This represents the current state of the cell. This represents the cell state at the previous moment.

[0082] In some embodiments, the input gate is configured to: process the cell state at the current time step using a hyperbolic tangent activation function to obtain a candidate output; process the hidden state at the previous time step and the model input data value at the current time step using a sigmoid activation function to obtain an output gate control signal; and obtain the hidden state at the current time step based on the output gate control signal and the candidate output.

[0083] The cell state at the current time step is determined by the hyperbolic tangent activation function tanh. Processing yields candidate outputs. Candidate output for: .

[0084] Based on the hidden state of the previous moment and the model input data value at the current time Through the sigmoid activation function Processing yields the output gate control signal. That is, the output gate control signal for: ; In the above formula, It is a sigmoid activation function; The output gate recursive weight matrix, The weight matrix is ​​used as the input to the output gate; This is the output gate bias vector.

[0085] Therefore, based on the output gate control signal and candidate output Get the hidden state at the current moment. The hidden state at the current moment for: .

[0086] After step S140, the fuel cell life prediction model is obtained. Based on this fuel cell life prediction model...

[0087] In one example, at time t, the predicted value of the fuel cell output voltage output by the fuel cell lifetime prediction model is: ; In the above formula, This is the predicted output voltage of the fuel cell, output by the fuel cell lifetime prediction model. It is a sigmoid activation function; This is the weight matrix of the fully connected layer. This is the bias vector for the fully connected layer.

[0088] The remaining lifespan of a fuel cell can be obtained by relating the predicted output voltage of the fuel cell from the fuel cell lifespan prediction model to the failure time (the time of the failure voltage threshold).

[0089] According to the fuel cell life prediction model acquisition method of this application embodiment, P real operating data points corresponding to P fuel cells under test conditions are obtained, where P is a positive integer; a fuel cell stack simulation model is constructed, and Q simulated operating data points corresponding to Q fuel cells under simulated test conditions are obtained through the fuel cell stack simulation model, where Q is a positive integer; raw data is obtained based on the P real operating data points and the Q simulated operating data points, and training data is obtained based on the raw data points; a preset model is trained using the training data to obtain the fuel cell life prediction model. The obtained fuel cell life prediction model can be used to predict the life of fuel cells. In the above scheme, the raw data used to obtain the training data is obtained from real operating data and simulated operating data. Simulated operating data can replace a large amount of real operating data, thereby reducing the testing cost and testing time cost for fuel cell life prediction.

[0090] In some embodiments, the method for obtaining a fuel cell life prediction model further includes: after obtaining Q simulation operation data, determining whether the fuel cell stack simulation model is accurate based on at least one real operation data and at least one simulation operation data; if the fuel cell stack simulation model is determined to be accurate, retaining the Q simulation operation data; if the fuel cell stack simulation model is determined to be inaccurate, repeatedly performing the steps of adjusting the fuel cell stack simulation model, obtaining the Q simulation operation data, and determining whether the fuel cell stack simulation model is accurate, until the fuel cell stack simulation model is determined to be accurate. Using real operation data to verify the accuracy of the simulation operation data output by the fuel cell stack simulation model ensures the reliability of the simulation operation data, facilitates the large-scale substitution of real operation data by simulation operation data, and helps reduce the testing cost and testing time cost for fuel cell life prediction.

[0091] In some embodiments, denoising the original data to obtain denoised data includes: performing wavelet threshold denoising on the original data to obtain denoised data. Wavelet threshold denoising eliminates noise interference in the original data, facilitating the acquisition of clearer stack operating characteristics from the denoised data. Wavelet threshold denoising can decompose key signals such as voltage and current, shrinking high-frequency noise coefficients through threshold rules while preserving low-frequency degradation trends. This improves the training efficiency of subsequent long short-term memory neural network models, resulting in a fuel cell lifetime prediction model that can more accurately identify stack lifetime degradation characteristics and improve prediction accuracy.

[0092] In some embodiments, training a preset model using training data to obtain a fuel cell life prediction model includes: training a preset LSTM model using training data to obtain the fuel cell life prediction model. The fuel cell life prediction model is obtained by training a preset Long Short-Term Memory (LSTM) neural network model. The LSTM neural network model is highly sensitive to time series information, which helps improve the accuracy of fuel cell life prediction.

[0093] This application also provides a battery life prediction method, which includes: obtaining a fuel cell life prediction model by means of the fuel cell life prediction model acquisition method of any of the foregoing embodiments; and obtaining the remaining life information of the fuel cell by means of the fuel cell life prediction model.

[0094] In some embodiments, the remaining lifespan of a fuel cell can be obtained by relating the predicted time point of the fuel cell's output voltage output from the fuel cell lifespan prediction model to the failure time point (the time point of the failure voltage threshold).

[0095] The method for obtaining the aforementioned fuel cell life prediction model includes: acquiring P real operating data points corresponding to P fuel cells under test conditions, where P is a positive integer; constructing a fuel cell stack simulation model, and acquiring Q simulated operating data points corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; obtaining raw data based on the P real operating data and Q simulated operating data, and obtaining training data based on the raw data; training the preset model using the training data to obtain the fuel cell life prediction model. The obtained fuel cell life prediction model can be used to predict the life of fuel cells.

[0096] In the above scheme, the raw data used to obtain training data is obtained from real operating data and simulation operating data. Simulation operating data can replace a large amount of real operating data, thereby reducing the testing cost and testing time cost for fuel cell life prediction.

[0097] This application also provides a device for obtaining a fuel cell lifetime prediction model. Figure 5 This is a schematic diagram of one embodiment of the fuel cell lifetime prediction model acquisition device according to this application. The fuel cell lifetime prediction model acquisition device includes a sampling module 110, a simulation module 120, a data processing module 130, and a training module 140.

[0098] The sampling module 110 is used to acquire P real operating data corresponding to P fuel cells under test conditions, where P is a positive integer.

[0099] The simulation module 120 is used to construct a fuel cell stack simulation model and to obtain Q simulation operation data corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer.

[0100] The data processing module 130 is used to obtain raw data based on P real running data and Q simulated running data, and to obtain training data based on the raw data.

[0101] The training module 140 is used to train a preset long short-term memory neural network model using training data to obtain a fuel cell life prediction model.

[0102] According to the fuel cell life prediction model acquisition method of this application embodiment, the sampling module 110 is used to acquire P real operating data corresponding to P fuel cells under test conditions. The simulation module 120 is used to construct a fuel cell stack simulation model and to acquire Q simulated operating data corresponding to Q fuel cells under simulated test conditions through the stack simulation model. The data processing module 130 is used to obtain raw data based on the P real operating data and Q simulated operating data, and to obtain training data based on the raw data. The training module 140 is used to train a preset long short-term memory neural network model using the training data to obtain a fuel cell life prediction model. In the above scheme, the raw data used to obtain the training data is obtained by coupling real operating data and simulated operating data. The simulated operating data can replace a large amount of real operating data, thereby reducing the testing cost and testing time cost for fuel cell life prediction.

[0103] This application also provides an electronic device. Figure 6 This is a schematic diagram of the hardware structure of an embodiment of the electronic device according to this application. The electronic device includes a memory 910 and a processor 920. The memory 910 and the processor 920 are communicatively connected. The memory 910 stores instructions, and the processor 920 calls the instructions in the memory 910 to cause the electronic device to execute the fuel cell lifetime prediction model acquisition method according to any of the foregoing embodiments of this application.

[0104] The method for obtaining the fuel cell life prediction model includes: obtaining P real operating data points corresponding to P fuel cells under test conditions, where P is a positive integer; constructing a fuel cell stack simulation model, and obtaining Q simulated operating data points corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; obtaining raw data based on the P real operating data points and Q simulated operating data points, and obtaining training data based on the raw data points; and training the preset model using the training data to obtain the fuel cell life prediction model.

[0105] Specifically, the processor 920 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0106] Memory 910 may include a large-capacity memory for data or instructions. For example, and not limitingly, memory 910 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 910 may include removable or non-removable (or fixed) media. Where appropriate, memory 910 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 910 may be the non-volatile memory described above. In a particular embodiment, memory 910 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0107] In one example, the electronic device may also include a communication interface 930 and a bus 940. The processor 920, memory 910, and communication interface 930 are connected via the bus 940 and communicate with each other.

[0108] The communication interface 930 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0109] Bus 940 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 940 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0110] Furthermore, in conjunction with the fuel cell lifetime prediction model acquisition method in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores instructions that, when executed by a processor, implement the fuel cell lifetime prediction model acquisition method of any of the foregoing embodiments of this application.

[0111] The method for obtaining the fuel cell life prediction model includes: obtaining P real operating data points corresponding to P fuel cells under test conditions, where P is a positive integer; constructing a fuel cell stack simulation model, and obtaining Q simulated operating data points corresponding to Q fuel cells under simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; obtaining raw data based on the P real operating data points and Q simulated operating data points, and obtaining training data based on the raw data points; and training the preset model using the training data to obtain the fuel cell life prediction model.

[0112] This application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0113] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0114] It should be noted that the technical solutions or features described in the above embodiments can be combined or supplemented with each other without conflict. The scope of protection of this application is not limited to the precise structures described in the above embodiments and shown in the accompanying drawings; all modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for obtaining a fuel cell lifetime prediction model, characterized in that, include: Obtain P real operating data points for P fuel cells under test conditions, where P is a positive integer; Construct a fuel cell stack simulation model, and obtain Q simulation operation data of the fuel cell under the simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; The original data is obtained based on P real running data and Q simulation running data, and the training data is obtained based on the original data. The preset model is trained using the training data to obtain a fuel cell life prediction model.

2. The method for obtaining the fuel cell lifetime prediction model according to claim 1, characterized in that, The acquisition of P real operating data corresponding to P fuel cells under test conditions includes: The real values ​​of preset operating parameters of each fuel cell at multiple time points are collected at a preset sampling frequency to obtain the real operating data of each fuel cell in a preset time series; The step of obtaining Q simulation operation data points for the fuel cell under the simulated test conditions through the fuel cell stack simulation model includes: The simulation operation data is obtained by acquiring the predicted values ​​of the preset operating parameters of each fuel cell on the preset time series through the fuel cell stack simulation model.

3. The method for obtaining the fuel cell lifetime prediction model according to claim 1, characterized in that, The preset operating parameters of the actual operating data and the simulated operating data include: output voltage, output current, output current density, relative humidity of the anode and cathode intake air, and operating temperature.

4. The method for obtaining the fuel cell lifetime prediction model according to claim 1, characterized in that, Also includes: After obtaining Q sets of simulation running data, the accuracy of the fuel cell stack simulation model is determined based on at least one set of real running data and at least one set of simulation running data. If the fuel cell stack simulation model is determined to be accurate, retain Q sets of the simulation operation data. If the fuel cell stack simulation model is determined to be inaccurate, the steps of adjusting the fuel cell stack simulation model, obtaining Q sets of simulation running data, and determining whether the fuel cell stack simulation model is accurate are repeated until the fuel cell stack simulation model is determined to be accurate.

5. The method for obtaining the fuel cell lifetime prediction model according to claim 4, characterized in that, The step of determining whether the fuel cell stack simulation model is accurate based on at least one of the real operating data and at least one of the simulation operating data includes: Obtain error information of one of the simulation running data relative to one of the real running data; If the error information meets the preset error requirements, the fuel cell stack simulation model is determined to be accurate. If the error information does not meet the preset error requirements, the fuel cell stack simulation model is determined to be inaccurate.

6. The method for obtaining the fuel cell lifetime prediction model according to claim 5, characterized in that, The actual operating data includes the actual value of the output voltage of the fuel cell over a preset time series, and the simulated operating data includes the predicted value of the output voltage of the fuel cell over the preset time series. The step of obtaining error information of one of the simulation running data relative to one of the real running data includes: The actual lifespan of the fuel cell is obtained from the true value of the output voltage based on the real operating data, and the predicted lifespan of the fuel cell is obtained from the predicted value of the output voltage based on the simulated operating data. The prediction error is obtained based on the predicted service life and the actual service life. The error information meets the preset error requirements, including: the prediction error is less than a first threshold.

7. The method for obtaining the fuel cell lifetime prediction model according to claim 6, characterized in that, The step of obtaining error information of one of the simulation running data relative to one of the real running data further includes: Calculate the root mean square error and coefficient of determination of the true value of the output voltage of the real operating data and the predicted value of the output voltage of the simulated operating data; The error information meets the preset error requirements, including: the prediction error is less than a first threshold, the root mean square error is less than a second threshold, and the coefficient of determination is greater than a third threshold.

8. The method for obtaining the fuel cell lifetime prediction model according to claim 1, characterized in that, The ratio of P to Q is between 2 / 8 and 4 / 6.

9. The method for obtaining the fuel cell lifetime prediction model according to claim 1, characterized in that, The process of obtaining training data based on the original data includes: The original data is subjected to noise reduction processing to obtain the noise-reduced data; The denoised data is divided into a training set, a validation set, and a test set; The training set, the validation set, and the test set are subjected to data standardization to obtain training data.

10. The method for obtaining the fuel cell lifetime prediction model according to claim 9, characterized in that, The noise reduction process on the original data to obtain the noise-reduced data includes: The original data is subjected to wavelet threshold denoising to obtain the denoised data.

11. The method for obtaining the fuel cell lifetime prediction model according to claim 1, characterized in that, The step of training a preset model using the training data to obtain a fuel cell life prediction model includes: The preset long short-term memory neural network model is trained using the training data to obtain a fuel cell life prediction model.

12. A method for predicting battery life, characterized in that, include: A fuel cell life prediction model is obtained by the method for obtaining the fuel cell life prediction model according to any one of claims 1 to 11; The remaining lifespan information of the fuel cell is obtained through the fuel cell lifespan prediction model.

13. A device for acquiring a fuel cell lifetime prediction model, characterized in that, include: The sampling module is used to acquire P real operating data points corresponding to P fuel cells under test conditions, where P is a positive integer; The simulation module is used to construct a fuel cell stack simulation model and to obtain Q simulation operation data of the fuel cell under the simulated test conditions through the fuel cell stack simulation model, where Q is a positive integer; The data processing module is used to obtain raw data based on P real running data and Q simulated running data, and to obtain training data based on the raw data; The training module is used to train a preset long short-term memory neural network model using the training data to obtain a fuel cell life prediction model.

14. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory being communicatively connected to the processor, and the memory storing instructions. The processor invokes the instructions in the memory to cause the electronic device to execute the method for obtaining the fuel cell lifetime prediction model according to any one of claims 1 to 11.

15. A computer-readable storage medium storing instructions thereon, characterized in that, When the instruction is executed by the processor, it implements the method for obtaining the fuel cell lifetime prediction model according to any one of claims 1 to 11.

Images