Acquisition method, device and equipment of electric pile state reference point and storage medium

Abstract

The invention discloses a method, a device and equipment for acquiring a pile state reference point, and a storage medium. The method comprises the following steps: acquiring operation data of a fuel cell pile under different working conditions; screening the operation data to obtain fuel cell stack state reference point data under different working conditions, and generating a reference point data set; constructing a nested neural network model based on a plurality of multilayer feedforward neural network sub-models connected in series; and training the nested neural network model by using the reference point data set, and outputting a predicted fuel cell stack state reference point through the trained nested neural network model. According to the method, the electric pile state reference point covering all working conditions can be predicted with high precision based on limited test data, a reliable basis is provided for fuel cell fault diagnosis, and the accuracy of state monitoring and the engineering application efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of hydrogen fuel cell technology, and in particular to a method, apparatus, device, and storage medium for obtaining a stack state reference point. Background Technology

[0002] Fuel cells are widely used in various fields due to their advantages such as being pollution-free, having stable performance, long service life, and high energy density. However, fuel cells are susceptible to changes in operating conditions during operation, which can lead to faults such as membrane drying, flooding, and low catalyst activity. Therefore, real-time monitoring and fault diagnosis of the fuel cell stack are necessary. Diagnosis requires comparing the fuel cell's monitoring points with baseline points to determine if a fault has occurred and to identify the type of fault. Thus, determining the fuel cell's baseline state is of great practical value for fuel cell fault diagnosis, thereby improving the fuel cell's durability and performance stability.

[0003] Currently, there are two main methods for determining the baseline state of fuel cell stacks: experimental and simulation methods. While experimental methods can obtain realistic data, they require a large amount of data and are susceptible to temperature fluctuations that prevent the acquisition of specific values ​​at particular temperatures. Furthermore, experimental methods are extremely costly and inefficient. Simulation methods, on the other hand, require the construction of complex fuel cell models with high accuracy requirements. These models also necessitate extensive experimental data for benchmarking to verify their feasibility.

[0004] Therefore, how to use a small amount of test data to obtain benchmark points, and then use the obtained benchmark points to complete the benchmark point MAP. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and storage medium for obtaining fuel cell stack condition reference points. This method can accurately predict fuel cell stack condition reference points covering all operating conditions based on limited experimental data, providing a reliable basis for fuel cell fault diagnosis and significantly improving the accuracy of condition monitoring and the efficiency of engineering applications.

[0006] In a first aspect, this application provides a method for obtaining a fuel cell stack state reference point, wherein the method includes the following steps: Obtain operating data of fuel cell stacks under different operating conditions; The operational data is filtered and processed to obtain fuel cell stack state reference point data under different operating conditions, and a reference point dataset is generated. A nested neural network model is constructed based on multiple cascaded multilayer feedforward neural network sub-models. The nested neural network model is trained using the reference point dataset, and the predicted fuel cell stack state reference points are output through the trained nested neural network model.

[0007] In conjunction with the first aspect mentioned above, as an optional implementation, the operating current and stack temperature in the reference point dataset are input into the optimized first feedforward neural network sub-model, which outputs the predicted fuel cell stack reference voltage data. The reference voltage data, operating current, and stack temperature are input into the optimized second feedforward neural network sub-model, which outputs the predicted imaginary part of the fuel cell stack reference AC impedance. The imaginary part of the AC impedance, the reference voltage, the operating current, and the stack temperature are input into the optimized third feedforward neural network sub-model, which outputs the predicted real part of the reference AC impedance of the fuel cell stack.

[0008] In conjunction with the first aspect mentioned above, as an optional implementation, the first feedforward neural network sub-model of the nested neural network model is trained using the operating current, stack temperature and stack voltage data in the reference point dataset, and the model is evaluated after training to obtain the optimized first feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, and imaginary part of AC impedance data in the reference point dataset, the second feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized second feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, imaginary part of AC impedance, and real part of AC impedance data from the reference point dataset, the third feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized third feedforward neural network sub-model.

[0009] In conjunction with the first aspect mentioned above, as an optional implementation method, the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination of the model are calculated. If the mean absolute error, mean square error, mean absolute percentage error, and coefficient of determination are all less than the corresponding error values, then the model is judged to be the optimal model. If the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination are all greater than the corresponding error values, the hyperparameters of the model are adjusted until the model is determined to be optimal. The hyperparameters include the number of layers, the number of neurons, the activation function, and the learning rate.

[0010] In conjunction with the first aspect mentioned above, as an optional implementation method, according to the formula: Calculate the mean absolute error, where, For the sample true value, The predicted value of the model. The number of samples; According to the formula: Calculate the mean square error; According to the formula: Calculate the mean absolute percentage error; According to the formula: Calculate the coefficient of determination, where, This is the average of the true values ​​of the sample.

[0011] In conjunction with the first aspect mentioned above, as an optional implementation method, different operating conditions of temperature, current and humidity are set, and the operating conditions are cross-combined and tested in the order of temperature from high to low, current from low to high and humidity from low to high to obtain the operating data of the fuel cell stack under each operating condition. The operating data includes: operating current, stack temperature, stack voltage, real part of AC impedance and imaginary part of AC impedance.

[0012] In conjunction with the first aspect mentioned above, as an optional implementation method, the results with the highest voltage values ​​under different operating conditions and their corresponding operating current, stack temperature, real part of AC impedance, and imaginary part of AC impedance are selected from the operating data and used as the fuel cell stack state reference point data. Based on the obtained state reference point data, a reference point dataset is generated.

[0013] Secondly, this application provides an apparatus for obtaining a fuel cell stack state reference point, the apparatus comprising: The acquisition module is used to acquire operating data of the fuel cell stack under different operating conditions; The processing module is used to filter and process the operating data to obtain fuel cell stack state reference point data under different operating conditions and generate a reference point dataset. The building block is used to construct nested neural network models based on multiple cascaded multilayer feedforward neural network sub-models. The prediction module is used to train the nested neural network model using the benchmark dataset, and output the predicted fuel cell stack state benchmarks through the trained nested neural network model.

[0014] Thirdly, this application also provides an electronic device, the electronic device comprising: a processor; and a memory storing computer-readable instructions, which, when executed by the processor, implement the method described in any one of the first aspects.

[0015] Fourthly, this application also provides a computer-readable storage medium storing computer program instructions that, when executed by a computer, cause the computer to perform the method described in any of the first aspects.

[0016] This application provides a method, apparatus, device, and storage medium for obtaining fuel cell stack state reference points. The method includes the following steps: acquiring operating data of a fuel cell stack under different operating conditions; filtering the operating data to obtain fuel cell stack state reference point data under different operating conditions, and generating a reference point dataset; constructing a nested neural network model based on multiple cascaded multilayer feedforward neural network sub-models; training the nested neural network model using the reference point dataset, and outputting predicted fuel cell stack state reference points through the trained nested neural network model. This application can accurately predict fuel cell stack state reference points covering all operating conditions based on limited experimental data, providing a reliable basis for fuel cell fault diagnosis and significantly improving the accuracy of condition monitoring and the efficiency of engineering applications.

[0017] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit the invention. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0019] Figure 1 This is a flowchart of a method for obtaining a fuel cell stack state reference point provided in an embodiment of this application; Figure 2 This is a schematic diagram of a device for obtaining a fuel cell stack state reference point provided in an embodiment of this application; Figure 3 This is a schematic diagram illustrating the acquisition of the fuel cell stack state reference point provided in the embodiments of this application; Figure 4 This is a diagram of the nested neural network structure provided in the embodiments of this application; Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of this application; Figure 6 This is a schematic diagram of a computer-readable program medium provided in an embodiment of this application. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0021] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. Some of the block diagrams shown in the drawings represent functional entities and do not necessarily correspond to physically or logically independent entities.

[0022] It needs to be explained that a fuel cell stack is composed of multiple individual fuel cell cells stacked together.

[0023] Electrochemical impedance spectroscopy: When a small-amplitude AC disturbance signal of a different frequency is applied to an electrochemical system, the ratio of the AC signal voltage to the current is measured (i.e., impedance).

[0024] The real part of impedance is the portion of the impedance vector that intersects the real axis in a rectangular coordinate system. It represents the value of the resistance, that is, the part of the circuit's opposition to alternating current that is in the same direction as the current.

[0025] The imaginary part of impedance in AC circuits is the portion of the impedance that opposes changes in current perpendicular to the direction of the current. It includes capacitive reactance and inductive reactance.

[0026] Impedance parameters include the imaginary part of AC impedance and the real part of AC impedance.

[0027] Frequency: In AC circuits, frequency refers to the number of times an AC signal completes a periodic change per unit time.

[0028] Fuel cell stack standard state: The state in which the fuel cell stack is in good working condition and no reversible faults have occurred, abbreviated as standard state.

[0029] Neural network model: A machine learning model that makes decisions in a way similar to the human brain.

[0030] Benchmark point of fuel cell stack state: When the fuel cell stack is in its standard state, the operating current, temperature, stack voltage, real part of AC impedance, and imaginary part of AC impedance together constitute a benchmark point of fuel cell stack state, or simply benchmark point.

[0031] Fuel cell stack state check point: When the fuel cell stack is in any state, the operating current, temperature, stack voltage, real part of AC impedance, and imaginary part of AC impedance together constitute a fuel cell stack state check point, or simply check point.

[0032] The embodiments of this application will be further described in detail below with reference to the accompanying drawings.

[0033] Reference Figure 1 , Figure 1 The diagram shown is a flowchart of a method for obtaining a fuel cell stack state reference point provided by the present invention. Figure 1 As shown, the method includes the following steps: Step S101: Obtain operating data of the fuel cell stack under different operating conditions.

[0034] Specifically, different operating conditions of temperature, current and humidity are set, and the operating conditions are cross-combined and tested in the order of temperature from high to low, current from low to high and humidity from low to high to obtain the operating data of the fuel cell stack under each operating condition. The operating data includes: operating current, stack temperature, stack voltage, real part of AC impedance and imaginary part of AC impedance.

[0035] To illustrate this concept, to identify the standard operating state of the fuel cell stack in different temperature ranges, humidity sensitivity is tested within a specific temperature range, with the stack voltage used as the criterion for determining its condition. The stack temperature can be set within a specific range (50~90℃) as needed, with operating points taken at 5℃ intervals. Hydrogen-air intake humidity is measured at 20% intervals, specifically 10%, 30%, 50%, 70%, and 90%. Before testing, the lowest average single-cell voltage protection value on the fuel cell stack test bench is set to 0.6V. During polarization curve testing, operating conditions are changed sequentially according to the inlet temperature, intake humidity, and current. Current is tested from low to high, humidity from low to high, and temperature from high to low. During testing, if the lowest single-cell voltage falls below 0.6V, testing at higher currents under that operating temperature and humidity condition is terminated. By combining these conditions, the polarization curve is tested to obtain the stack voltage and its corresponding real and imaginary parts. At each operating point with a combination of temperature and current density, the result with the highest voltage value and its corresponding real and imaginary parts are selected as the baseline. That is, under a set of temperature and current conditions, the humidity condition with the highest voltage, i.e., the best performance, is selected as the baseline. For example, to determine the state baseline of the fuel cell stack under a specific operating condition of temperature = 80°C and current density = 0.5 A / cm², while keeping the temperature and current constant, the intake humidity of hydrogen and air is systematically changed, and the output voltage of the fuel cell stack is measured under each humidity level. For a detailed example, see Table 1.

[0036] Table 1

[0037] The fuel cell stack output voltage reaches its highest value of 0.84 V when the humidity is 50%. This indicates that under this specific temperature and current conditions, 50% humidity achieves the optimal balance between water and heat management within the fuel cell stack. The baseline is recorded as follows: Therefore, the baseline conditions for the operating condition (80°C, 0.5 A / cm²) are determined as follows: Temperature = 80°C, Current density = 0.5 A / cm², Voltage = 0.84 V, Imaginary part = (imaginary part value measured at 50% humidity), Real part = (real part value measured at 50% humidity).

[0038] Step S102: Filter the operating data to obtain fuel cell stack state reference point data under different operating conditions, and generate a reference point dataset.

[0039] Specifically, the highest voltage value under different operating conditions and its corresponding operating current, stack temperature, and impedance parameters are selected from the operational data. This includes the real and imaginary parts of the AC impedance, which serve as reference point data for the fuel cell stack state. Based on the obtained reference point data, a reference point dataset is generated. That is, the stack voltage and its corresponding real and imaginary parts are obtained by cross-combining polarization curves under different operating conditions. At each operating point with a combination of temperature and voltage, the highest voltage value and its corresponding real and imaginary parts are selected as reference points.

[0040] In one embodiment, after determining the standard state (i.e., determining the optimal humidity under fixed temperature and current), the polarization curve of the fuel cell is tested (i.e., different currents are applied in order from small to large to obtain voltage and other data). The operating data after the operation is stable is taken, and the voltage, imaginary part, and real part corresponding to the current and temperature are the reference points under the main operating conditions.

[0041] Step S103: Construct a nested neural network model based on multiple multi-layer feedforward neural network sub-models connected in series.

[0042] Specifically, nested neural networks establish multiple multi-layer feedforward neural network sub-models, using the final output of the previous model as one of the inputs to the next model, thus achieving the function of nested training of various sub-models. A multi-layer feedforward neural network consists of an input layer, a hidden layer, and an output layer, for a total of three layers.

[0043] Step S104: Train the nested neural network model using the reference point dataset, and output the predicted fuel cell stack state reference points through the trained nested neural network model.

[0044] Specifically, the first feedforward neural network sub-model of the nested neural network model is trained using the operating current, stack temperature and stack voltage data in the reference point dataset, and the model is evaluated after training to obtain the optimized first feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, and imaginary part of AC impedance data in the reference point dataset, the second feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized second feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, imaginary part of AC impedance, and real part of AC impedance data from the reference point dataset, the third feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized third feedforward neural network sub-model.

[0045] The dataset uses baseline data (current, temperature, voltage, imaginary part, and real part) selected through sensitivity testing, divided into training, testing, and validation sets in an 8:1:1 ratio. Three neural networks are built for nested training on the dataset. In each sample, current, temperature, and the output of the previous model are used as input variables, while voltage, imaginary part, and real part are used as output variables. After the previous model predicts a new result, it is substituted into the next model for retraining. In other words, the nested neural network achieves nested training by establishing multiple multi-layered feedforward neural network sub-models (three layers, consisting of three sub-models), using the final output of the previous model as one of the inputs to the next model.

[0046] During model training, the optimal model result is obtained through evaluation. Specifically, this involves evaluating the model after training, including: Calculate the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination of the model; If the mean absolute error, mean square error, mean absolute percentage error, and coefficient of determination are all less than the corresponding error values, then the model is judged to be the optimal model. If the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination are all greater than the corresponding error values, the hyperparameters of the model are adjusted until the model is determined to be optimal. The hyperparameters include the number of layers, the number of neurons, the activation function, and the learning rate.

[0047] According to the formula: Calculate the mean absolute error, where, For the sample true value, The predicted value of the model. The number of samples; According to the formula: Calculate the mean square error; According to the formula: Calculate the mean absolute percentage error; According to the formula: Calculate the coefficient of determination, where, This is the average of the true values ​​of the sample.

[0048] After model training, the operating current and stack temperature in the benchmark dataset are input into the optimized first feedforward neural network sub-model, which outputs the predicted fuel cell stack benchmark voltage data. The reference voltage data, operating current, and stack temperature are input into the optimized second feedforward neural network sub-model, which outputs the predicted imaginary part of the fuel cell stack reference AC impedance. The imaginary part of the AC impedance, the reference voltage, the operating current, and the stack temperature are input into the optimized third feedforward neural network sub-model, which outputs the predicted real part of the reference AC impedance of the fuel cell stack.

[0049] In summary, this application obtains fuel cell stack operating data (current, temperature, voltage, imaginary part, and real part) under different operating conditions (different temperatures, humidity, and current densities) through fuel cell experiments; filters and processes the operating data to obtain state reference points; and trains the experimental data based on a nested neural network model to predict the reference point MAP within the main operating condition range. Temperature and current are the headers of the MAP. After determining the temperature and current, the voltage, real part, and imaginary part can be obtained from the MAP table, i.e., the reference state (voltage, real part, and imaginary part) of the fuel cell under different operating conditions (temperature and current).

[0050] Reference Figure 2 , Figure 2 The diagram shown is a schematic of a device for obtaining a fuel cell stack state reference point provided by the present invention. Figure 2 As shown, the device includes: Acquisition module 201: It is used to acquire the operating data of the fuel cell stack under different operating conditions.

[0051] Processing module 202: It is used to filter and process the operating data to obtain fuel cell stack state reference point data under different operating conditions, and generate a reference point dataset.

[0052] Module 203: It is used to construct nested neural network models based on multiple multilayer feedforward neural network sub-models in series.

[0053] Prediction module 204: It is used to train the nested neural network model using the reference point dataset, and output the predicted fuel cell stack state reference points through the trained nested neural network model.

[0054] Furthermore, in one possible implementation, the prediction module is also used to input the operating current and stack temperature in the reference point dataset into the optimized first feedforward neural network sub-model, and output the predicted fuel cell stack reference voltage data. The reference voltage data, operating current, and stack temperature are input into the optimized second feedforward neural network sub-model, which outputs the predicted imaginary part of the fuel cell stack reference AC impedance. The imaginary part of the AC impedance, the reference voltage, the operating current, and the stack temperature are input into the optimized third feedforward neural network sub-model, which outputs the predicted real part of the reference AC impedance of the fuel cell stack.

[0055] Furthermore, in one possible implementation, the prediction module is also used to train the first feedforward neural network sub-model of the nested neural network model using the operating current, stack temperature and stack voltage data in the reference point dataset, and to evaluate the model after training to obtain the optimized first feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, and imaginary part of AC impedance data in the reference point dataset, the second feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized second feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, imaginary part of AC impedance, and real part of AC impedance data from the reference point dataset, the third feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized third feedforward neural network sub-model.

[0056] Furthermore, in one possible implementation, a prediction module is used to calculate the model's mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination. If the mean absolute error, mean square error, mean absolute percentage error, and coefficient of determination are all less than the corresponding error values, then the model is judged to be the optimal model. If the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination are all greater than the corresponding error values, the hyperparameters of the model are adjusted until the model is determined to be optimal. The hyperparameters include the number of layers, the number of neurons, the activation function, and the learning rate.

[0057] Furthermore, in one possible implementation, the prediction module is also used to predict according to the formula: Calculate the mean absolute error, where, For the sample true value, The predicted value of the model. The number of samples; According to the formula: Calculate the mean square error; According to the formula: Calculate the mean absolute percentage error; According to the formula: Calculate the coefficient of determination, where, This is the average of the true values ​​of the sample.

[0058] Furthermore, in one possible implementation, the acquisition module is also used to set different temperature, current and humidity operating conditions, and to perform cross-combination tests on the operating conditions in the order of temperature from high to low, current from low to high and humidity from low to high, so as to obtain the operating data of the fuel cell stack under each operating condition. The operating data includes: operating current, stack temperature, stack voltage, real part of AC impedance and imaginary part of AC impedance.

[0059] Furthermore, in one possible implementation, the processing module is also used to filter out the results with the highest voltage values ​​under different operating conditions from the operating data, as well as their corresponding operating current, stack temperature, real part of AC impedance, and imaginary part of AC impedance, as fuel cell stack state reference point data. Based on the obtained state reference point data, a reference point dataset is generated.

[0060] Reference Figure 3 , Figure 3 The diagram shown illustrates the acquisition of the fuel cell stack state reference point provided by this invention. Figure 3 As shown: Humidity sensitivity was tested within a specific temperature range, with the stack voltage used as the criterion for determining its condition. The stack temperature could be set within a specific range (50~90℃) as needed, with operating points taken at 5℃ intervals. Hydrogen-air inlet humidity was measured at 20% intervals, with 10%, 30%, 50%, 70%, and 90% recommended. Before testing, the lowest average single-cell voltage protection value on the stack test bench was set to 0.6V. During polarization curve testing, operating conditions were changed sequentially according to inlet temperature, inlet humidity, and current. Current was tested from low to high, humidity from low to high, and temperature from high to low. During testing, if the lowest single-cell voltage fell below 0.6V, the test at higher currents under that operating temperature and humidity condition was terminated. The polarization curve was tested using the above conditions in combination to obtain the stack voltage and its corresponding real and imaginary parts. At each operating point of temperature and voltage combination, the result with the highest voltage value and its corresponding real and imaginary parts were selected as the benchmark point.

[0061] For any state reference point composed of temperature and current within the main operating condition range, it can be obtained through interpolation methods. Specifically, nested neural network interpolation can be used.

[0062] Nested neural networks achieve this by establishing multiple multi-layer feedforward neural network sub-models, using the final output of the previous model as one of the inputs to the next model, thus enabling nested training of the sub-models. A multi-layer feedforward neural network has a three-layer structure: an input layer, a hidden layer, and an output layer.

[0063] The steps for predicting the state baseline (MAP) of a fuel cell using nested neural networks are as follows: First, construct the dataset: Use the baseline data (current, temperature, voltage, imaginary part, real part) selected through sensitivity testing as the dataset, and divide it into training, test, and validation sets in an 8:1:1 ratio. Second, build three nested neural networks to train on the dataset. In each sample, use current, temperature, and the output of the previous model as input variables, and voltage, imaginary part, and real part as output variables. After the previous model predicts a new result, substitute it into the next model for retraining.

[0064] Building Neural Networks: Design the hyperparameters of each neural network (number of layers, number of neurons, activation function, learning rate, etc.), construct a nested neural network model, and use current, temperature, and voltage as training data for model training. After building the model, evaluation parameters are needed to assess its performance. By appropriately adjusting the hyperparameters within a reasonable range, an attempt is made to find a better-performing model. The main metrics for evaluating model performance are mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination. The mean absolute error is the average of the absolute differences between the true values ​​and the model's predicted values, calculated using the following formula: .

[0065] The mean squared error is the squared average of the squared differences between the true values ​​and the model's predicted values. The formula for calculating this error is: .

[0066] The mean absolute percentage error (MAE) is the average of the absolute distances between the true values ​​and the model predictions divided by the true values ​​of the sample. The formula for calculating MAE is: .

[0067] The coefficient of determination is the ratio of the sum of squared differences between the true values ​​and the model's predicted values ​​of the sample to the sum of squared differences between the mean of the true values ​​and the sum of squared differences of the actual observed values. It is calculated as follows: .

[0068] The smaller the three types of errors mentioned above, the closer the coefficient of determination is to 1, indicating better model performance. Therefore, finding a better model means finding a model with smaller errors and better fitting performance.

[0069] After finding the optimal model (the optimized first feedforward neural network sub-model), the operating current and stack temperature in the reference point dataset are substituted into the model to output the predicted reference voltage value, thereby obtaining the voltage at various currents and temperatures within the main operating conditions (the main operating conditions are custom operating conditions), thus obtaining the reference voltage MAP.

[0070] Similarly, using the operating current, stack temperature, stack voltage, and AC impedance imaginary part data from the reference point dataset, the second feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized second feedforward neural network sub-model. The reference voltage data, operating current, and stack temperature are substituted into the model to output the predicted imaginary part value, thereby obtaining the imaginary part of each current and temperature within the main operating conditions range, thus obtaining the reference imaginary part MAP.

[0071] Using the operating current, stack temperature, stack voltage, imaginary part of AC impedance, and real part of AC impedance data from the reference point dataset, the third feedforward neural network sub-model of the nested neural network model is trained. After training, the model is evaluated to obtain the optimized third feedforward neural network sub-model. The test set is substituted into the model to output the predicted real part value, thereby obtaining the real part of each current and temperature within the main operating conditions, thus obtaining the reference real part MAP.

[0072] In summary, this application uses a small amount of experimental data (selected benchmark points) and a nested model to output benchmark points to complete the benchmark point MAP under different working conditions.

[0073] Reference Figure 4 , Figure 4 The diagram shown is a nested neural network structure provided by the present invention. Figure 4 As shown: Based on the generated benchmark dataset, a nested neural network model is trained. This model consists of three feedforward neural network sub-models. The first sub-model takes the operating current and stack temperature as input and outputs the predicted voltage. The second sub-model takes the predicted voltage, operating current, and stack temperature as input and outputs the predicted first impedance parameter, i.e., the imaginary part of the AC impedance. The stack temperature, operating current, the predicted imaginary part, and the voltage are used as inputs to the third sub-model, which outputs the predicted second impedance parameter, i.e., the real part of the AC impedance. In other words, the first layer takes current and temperature as input and outputs voltage; the second layer takes current, temperature, and predicted voltage as input and outputs the imaginary part; and the third layer takes current, temperature, predicted voltage, and the predicted imaginary part as input and outputs the real part.

[0074] The following reference Figure 5 To describe an electronic device 500 according to this embodiment of the present invention. Figure 5 The electronic device 500 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0075] like Figure 5As shown, the electronic device 500 is manifested in the form of a general-purpose computing device. The components of the electronic device 500 may include, but are not limited to: at least one processing unit 510, at least one storage unit 520, and a bus 530 connecting different system components (including storage unit 520 and processing unit 510).

[0076] The storage unit stores program code that can be executed by the processing unit 510, causing the processing unit 510 to perform the steps described in the "Embodiment Methods" section of this specification according to various exemplary embodiments of the present invention.

[0077] Storage unit 520 may include readable media in the form of volatile storage units, such as random access memory (RAM) 521 and / or cache memory 522, and may further include read-only memory (ROM) 523.

[0078] Storage unit 520 may also include a program / utility 524 having a set (at least one) program module 525, such program module 525 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0079] Bus 530 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0080] Electronic device 500 can also communicate with one or more external devices (e.g., keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with electronic device 500, and / or any device that enables electronic device 500 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 550. Furthermore, electronic device 500 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 560. As shown, network adapter 560 communicates with other modules of electronic device 500 via bus 530. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 500, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0081] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0082] According to the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the methods described above is stored. In some possible embodiments, various aspects of the present invention can also be implemented as a program product comprising program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps of the various exemplary embodiments of the present invention described in the "Exemplary Methods" section above.

[0083] refer to Figure 6 As shown, a program product 600 for implementing the above-described method according to an embodiment of the present invention is described. It may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0084] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0085] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0086] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0087] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0088] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of the present invention, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0089] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

[0090] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

Claims

1. A method for obtaining a fuel cell stack state reference point, characterized in that, include: Obtain operating data of fuel cell stacks under different operating conditions; The operational data is filtered and processed to obtain fuel cell stack state reference point data under different operating conditions, and a reference point dataset is generated. A nested neural network model is constructed based on multiple cascaded multilayer feedforward neural network sub-models. The nested neural network model is trained using the reference point dataset, and the predicted fuel cell stack state reference points are output through the trained nested neural network model.

2. The method according to claim 1, characterized in that, The fuel cell stack state reference point predicted by the trained nested neural network model includes: The operating current and stack temperature from the benchmark dataset are input into the optimized first feedforward neural network sub-model, which outputs the predicted fuel cell stack benchmark voltage data. The reference voltage data, operating current, and stack temperature are input into the optimized second feedforward neural network sub-model, which outputs the predicted imaginary part of the fuel cell stack reference AC impedance. The imaginary part of the AC impedance, the reference voltage, the operating current, and the stack temperature are input into the optimized third feedforward neural network sub-model, which outputs the predicted real part of the reference AC impedance of the fuel cell stack.

3. The method according to claim 1, characterized in that, The step of training the nested neural network model using the benchmark dataset includes: The first feedforward neural network sub-model of the nested neural network model is trained using the operating current, stack temperature and stack voltage data in the reference point dataset, and the model is evaluated after training to obtain the optimized first feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, and imaginary part of AC impedance data in the reference point dataset, the second feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized second feedforward neural network sub-model. Using the operating current, stack temperature, stack voltage, imaginary part of AC impedance, and real part of AC impedance data from the reference point dataset, the third feedforward neural network sub-model of the nested neural network model is trained, and the model is evaluated after training to obtain the optimized third feedforward neural network sub-model.

4. The method according to claim 3, characterized in that, The evaluation of the model after training includes: Calculate the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination of the model; If the mean absolute error, mean square error, mean absolute percentage error, and coefficient of determination are all less than the corresponding error values, then the model is judged to be the optimal model. If the mean absolute error, mean squared error, mean absolute percentage error, and coefficient of determination are all greater than the corresponding error values, the hyperparameters of the model are adjusted until the model is determined to be optimal. The hyperparameters include the number of layers, the number of neurons, the activation function, and the learning rate.

5. The method according to claim 4, characterized in that, include: According to the formula: Calculate the mean absolute error, where, For the sample true value, The predicted value of the model. The number of samples; According to the formula: Calculate the mean square error; According to the formula: Calculate the mean absolute percentage error; According to the formula: Calculate the coefficient of determination, where, This is the average of the true values ​​of the sample.

6. The method according to claim 1, characterized in that, The acquisition of operating data of the fuel cell stack under different operating conditions includes: Different operating conditions of temperature, current and humidity are set, and the operating conditions are cross-combined and tested in the order of temperature from high to low, current from low to high and humidity from low to high to obtain the operating data of fuel cell stack under each operating condition. The operating data includes: operating current, stack temperature, stack voltage, real part of AC impedance and imaginary part of AC impedance.

7. The method according to claim 1, characterized in that, The process of filtering the operational data to obtain fuel cell stack state reference point data under different operating conditions and generating a reference point dataset includes: The highest voltage values ​​under different operating conditions and their corresponding operating current, stack temperature, real part of AC impedance, and imaginary part of AC impedance are selected from the operating data and used as the fuel cell stack state reference point data. Based on the obtained state reference point data, a reference point dataset is generated.

8. A device for obtaining a fuel cell stack state reference point, characterized in that, include: The acquisition module is used to acquire operating data of the fuel cell stack under different operating conditions; The processing module is used to filter and process the operating data to obtain fuel cell stack state reference point data under different operating conditions and generate a reference point dataset. The building block is used to construct nested neural network models based on multiple cascaded multilayer feedforward neural network sub-models. The prediction module is used to train the nested neural network model using the benchmark dataset, and output the predicted fuel cell stack state benchmarks through the trained nested neural network model.

9. An electronic device, characterized in that, The electronic device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer program instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 7.

Images