Intelligent management method and system for fuel cell

Abstract

The invention discloses an intelligent management method and system for a fuel cell. The method comprises the following steps: constructing a fuel cell degradation empirical model and a fuel cell life prediction model; setting model weights of the fuel cell degradation empirical model and the fuel cell service life prediction model by using the comparison set, and superposing the fuel cell degradation empirical model and the fuel cell service life prediction model according to the model weights to obtain a cell service life prediction hybrid model; acquiring monitoring data of the fuel cell, and inputting the monitoring data into the cell life prediction hybrid model to obtain a residual life evaluation value of the fuel cell; generating a maintenance instruction of the fuel cell based on the residual life evaluation value; the accuracy and uncertainty of predicting the residual service life of the fuel cell in the prior art are balanced, and accurate prediction of the residual service life of the fuel cell is realized; based on the service life prediction result, intelligent task allocation and planned maintenance of the fuel cell equipment are realized, and the reliability and the service life of the fuel cell system are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of fuel cell health management, and specifically relates to an intelligent management method and an acceleration method for fuel cells. Background Art

[0002] As a clean energy technology, fuel cells have shown broad application prospects in engineering machinery and other fields. In actual dynamic operation, factors such as fluctuations in stack current / voltage, changes in operating temperature / pressure, and other unstable operating conditions will accelerate the degradation and failure rate of fuel cells, greatly reducing the remaining service life of fuel cells. Therefore, accurately predicting the remaining service life of fuel cells and timely maintaining fuel cells according to the remaining service life of fuel cells are of great significance for improving power generation efficiency, extending life cycle, and reducing operating costs.

[0003] At present, the existing methods for predicting the remaining service life of fuel cells include model-driven methods and data-driven methods. The model-driven method predicts the remaining service life based on the theoretical mathematical equation of the fuel cell degradation phenomenon, including mechanism models, semi-empirical models and empirical models. The data-driven method trains the machine learning framework through the monitoring data of the fuel cell and predicts the remaining service life through the trained machine learning framework.

[0004] Model-driven methods have limitations in model accuracy and applicability, and are difficult to cope with complex actual working conditions; data-driven methods rely on the quality and quantity of data, and are usually lacking in interpretability, making them difficult to adapt to changing working conditions. In addition, existing technologies do not involve intelligent control and planned maintenance of fuel cell equipment tasks based on life prediction, and are unable to fully utilize the application value of prediction results in actual operations. Summary of the invention

[0005] The present invention provides an intelligent management method and system for fuel cells, which balances the accuracy and uncertainty of the existing technology in predicting the remaining service life of fuel cells, and realizes the accurate prediction of the remaining service life of fuel cells; at the same time, based on the life prediction results, it realizes the intelligent task allocation and planned maintenance of fuel cell equipment, thereby improving the reliability and life of the fuel cell system.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] The first aspect of the present invention provides an intelligent management method for a fuel cell, comprising:

[0008] Acquire monitoring data of the fuel cell, input the monitoring data into a battery life prediction hybrid model to obtain a remaining life evaluation value of the fuel cell; generate a maintenance instruction for the fuel cell based on the remaining life evaluation value;

[0009] The construction process of the battery life prediction hybrid model includes:

[0010] Acquire historical operation data of the fuel cell; divide the historical operation data into a training set and a comparison set;

[0011] Establishing an initial model of fuel cell degradation; inputting historical operation data in the training set into the initial model of fuel cell degradation, and using a particle filter algorithm to estimate the parameters of the initial model of fuel cell degradation to obtain an empirical model of fuel cell degradation;

[0012] A fuel cell life prediction model is established based on a long short-term memory neural network, and the fuel cell life prediction model is trained using historical operation data in the training set. The trained fuel cell life prediction model is obtained by repeated iterations until the training termination condition is reached.

[0013] After setting the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model using the comparison set, the fuel cell degradation empirical model and the fuel cell life prediction model are superimposed according to the model weights to obtain a battery life prediction hybrid model.

[0014] Furthermore, the process of establishing an initial fuel cell degradation model includes:

[0015] The actual output voltage of the fuel cell is used to represent the remaining service life of the fuel cell, and the initial fuel cell degradation model is constructed; the expression formula is: ; ; ; ; ;

[0016] In the formula, is the actual output voltage of the fuel cell; is the Nernst voltage of the fuel cell; is the activation loss of the fuel cell; is the ohmic loss of the fuel cell; is the concentration loss of the fuel cell; is the change in Gibbs free energy before and after the fuel cell reaction; is the Faraday constant; is the universal gas constant; is the entropy change before and after the fuel cell reaction; is the fuel cell operating temperature; is the partial pressure of hydrogen; is the partial pressure of oxygen; is the thermodynamic standard state pressure; is the thermodynamic standard state temperature; , , , is the fitting parameter; is the current density per unit electrode area; is the electrode area; is the charge transfer resistance per unit area; is the limiting current density.

[0017] Furthermore, the historical operation data in the training set is input into the initial model of fuel cell degradation, and the particle filter algorithm is used to estimate the parameters of the initial model of fuel cell degradation to obtain the empirical model of fuel cell degradation. The process includes:

[0018] The initial measurement value of the remaining service life of the fuel cell is obtained from the historical operation data in the training set, and the measurement noise is set to conform to the normal distribution; the initial state value of the remaining service life of the fuel cell is obtained based on the initial measurement value and the measurement noise;

[0019] Generate a set number based on the initial state value of the remaining service life of the fuel cell Particles; set the initial weight of the particles, the expression formula is: ; is the initial weight of the i-th particle;

[0020] The initial fuel cell degradation model is used as a state transition model; the particles are input into the state transition model to obtain the predicted value of the fuel cell usage state at the next moment, and the measured value of the remaining service life of the fuel cell at the next moment is obtained from the historical operation data in the training set. The particle weight is updated according to the predicted value and measured value of the remaining service life of the fuel cell, and the particle weight is normalized;

[0021] The particles are resampled according to the particle weights, and the parameters of the initial fuel cell degradation model are estimated by combining the resampled particle set to obtain the fuel cell degradation empirical model.

[0022] Furthermore, the particle weights are updated according to the predicted value and measured value of the remaining service life of the fuel cell, including: ;

[0023] In the formula, is the standard deviation of the measurement noise; is the updated weight of the particle; is a measure of the remaining useful life of the fuel cell; is the predicted value of the remaining service life of the fuel cell; is the ratio of pi; is the time step index.

[0024] Furthermore, the weights of the particles are normalized, including: ;

[0025] In the formula, For the The initial weight of each particle; is a measure of the remaining useful life of the fuel cell; For the The predicted value of the remaining service life of the fuel cell corresponding to each particle; For the The predicted value of the remaining service life of the fuel cell corresponding to each particle; is the parameter vector of the state transition model.

[0026] Furthermore, the fuel cell life prediction model is trained using the historical operation data in the training set, and the iteration is repeated until the training termination condition is reached to obtain the trained fuel cell life prediction model. The process includes:

[0027] The historical operation data in the training set is used as training samples, and the training samples include working time, fuel cell voltage, current, single chip voltage, intake pressure, exhaust pressure, intake metering ratio, inlet and outlet cooling water temperature, and actual working life of the fuel cell; the actual working life of the fuel cell is used as the true label;

[0028] The historical operation data in the training set is input into the fuel cell life prediction model to obtain the battery life training value; the training loss value is calculated based on the battery life training value and the true label; the Adam optimizer is used to optimize the parameters of the fuel cell life prediction model according to the training loss value, and the training process of the fuel cell life prediction model is repeated until the number of iterations reaches the iteration threshold to output the trained fuel cell life prediction model.

[0029] Furthermore, the comparison set is used to set the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model, and the process includes:

[0030] Inputting the comparison set into the fuel cell degradation empirical model to obtain the model driving function of the fuel cell remaining useful life probabilities distribution;

[0031] The comparison set is input into a fuel cell life prediction model to obtain a data-driven function of the fuel cell remaining service life probabilities distribution;

[0032] Establishing the actual distribution function of the service life of the fuel cell based on the actual working life of the fuel cell;

[0033] The model weights of the fuel cell degradation empirical model and the fuel cell life prediction model are calculated based on the model driving function, data driving function and actual service life distribution function. The expression formula is: ;

[0034] In the formula, is the actual distribution function of service life; A model-driven function for the distribution of the remaining useful life of a fuel cell; A data driven function for the fuel cell remaining useful life probabilities distribution; is the model weight of the fuel cell degradation empirical model; is the model weight of the fuel cell life prediction model; To ensure the accuracy of the prediction results;

[0035] When accuracy When it reaches 1, the model weight of the fuel cell degradation empirical model is obtained. And the model weight of the fuel cell life prediction model .

[0036] Further, a maintenance instruction for the fuel cell is generated based on the remaining life evaluation value, and the process includes:

[0037] The life factor of the fuel cell is calculated based on the remaining life evaluation value, and the expression formula is: ;

[0038] The formula is, is the lifespan factor; It is the estimated value of the remaining life of the fuel cell; The service life of the fuel cell;

[0039] When the life factor is less than the life threshold, a maintenance instruction for the fuel cell is generated.

[0040] A second aspect of the present invention provides an intelligent management system for a fuel cell, comprising:

[0041] A monitoring module, used to obtain monitoring data of the fuel cell, input the monitoring data into a battery life prediction hybrid model to obtain a remaining life evaluation value of the fuel cell; and generate a maintenance instruction for the fuel cell based on the remaining life evaluation value;

[0042] An acquisition module is used to acquire historical operation data of the fuel cell; and divide the historical operation data into a training set and a comparison set respectively;

[0043] The first building module is to establish an initial model of fuel cell degradation; input the historical operation data in the training set into the initial model of fuel cell degradation, and use a particle filter algorithm to estimate the parameters of the initial model of fuel cell degradation to obtain an empirical model of fuel cell degradation;

[0044] The second building block is to establish a fuel cell life prediction model based on a long short-term memory neural network, use the historical operation data in the training set to train the fuel cell life prediction model, and repeat the iteration until the training termination condition is reached to obtain the trained fuel cell life prediction model;

[0045] The fitting module uses the comparison set to set the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model, and then superimposes the fuel cell degradation empirical model and the fuel cell life prediction model according to the model weights to obtain a battery life prediction hybrid model.

[0046] The third aspect of the present invention provides an electronic device, comprising a storage medium and a processor; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the intelligent management method for fuel cells described in the first aspect.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention obtains monitoring data of fuel cells, inputs the monitoring data into a battery life prediction hybrid model to obtain a remaining life evaluation value of the fuel cell; generates maintenance instructions for the fuel cell based on the remaining life evaluation value; and through planned maintenance and intelligent task allocation, can extend the service life of the fuel cell, reduce replacement frequency, and reduce long-term operating costs.

[0049] The present invention constructs a fuel cell degradation empirical model and a fuel cell life prediction model. After using a comparison set to set the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model, the fuel cell degradation empirical model and the fuel cell life prediction model are superimposed according to the model weights to obtain a battery life prediction hybrid model. The battery life prediction hybrid model balances the accuracy and uncertainty of the two models and can more accurately predict the remaining service life of the fuel cell. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A flowchart of an intelligent management method for a fuel cell provided in Example 1 of the present invention;

[0051] Figure 2 A structural diagram of a fuel cell provided in Example 1 of the present invention;

[0052] Figure 3 This is a flow chart of the hybrid model for battery life prediction provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0053] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0054] Example 1

[0055] like Figure 1 As shown, this embodiment provides an intelligent management method for a fuel cell, comprising:

[0056] like Figure 2 As shown, the fuel cell includes a fuel cell stack; the fuel cell stack is provided with a hydrogen supply circuit, an air supply circuit and a cooling system.

[0057] like Figure 3 As shown in Figure 1, the process of building a hybrid model for battery life prediction includes:

[0058] Acquire historical operation data of the fuel cell; divide the historical operation data into a training set, a prediction set and a comparison set;

[0059] The process of developing an initial fuel cell degradation model includes:

[0060] The actual output voltage of the fuel cell is used to represent the remaining service life of the fuel cell, and the initial fuel cell degradation model is constructed; the expression formula is: ; ; ; ; ;

[0061] In the formula, is the actual output voltage of the fuel cell; is the Nernst voltage of the fuel cell; is the activation loss of the fuel cell; is the ohmic loss of the fuel cell; is the concentration loss of the fuel cell; is the change in Gibbs free energy before and after the fuel cell reaction; is the Faraday constant; is the universal gas constant; is the entropy change before and after the fuel cell reaction; is the fuel cell operating temperature; is the partial pressure of hydrogen; is the partial pressure of oxygen; is the thermodynamic standard state pressure; is the thermodynamic standard state temperature; , , , is the fitting parameter; is the current density per unit electrode area; is the electrode area; is the charge transfer resistance per unit area; is the limiting current density.

[0062] The historical operation data in the training set is input into the initial model of fuel cell degradation, and the particle filter algorithm is used to estimate the parameters of the initial model of fuel cell degradation to obtain the empirical model of fuel cell degradation. The process includes:

[0063] The initial measurement value of the remaining service life of the fuel cell is obtained from the historical operation data in the training set, and the measurement noise is set to conform to the normal distribution; the initial state value of the remaining service life of the fuel cell is obtained based on the initial measurement value and the measurement noise;

[0064] Generate a set number based on the initial state value of the remaining service life of the fuel cell Particles; set the initial weight of the particles, the expression formula is: ; is the initial weight of the i-th particle;

[0065] The initial model of fuel cell degradation is used as a state transfer model; particles are input into the state transfer model to obtain the predicted value of the fuel cell usage state at the next moment, and the measured value of the remaining service life of the fuel cell at the next moment is obtained from the historical operation data in the training set.

[0066] Update particle weights based on predicted and measured values ​​of the remaining useful life of the fuel cell, including: ;

[0067] In the formula, is the standard deviation of the measurement noise; is the updated weight of the particle; is a measure of the remaining useful life of the fuel cell; is the predicted value of the remaining service life of the fuel cell; is the ratio of pi; is the time step index.

[0068] Normalize the particle weights, including: ;

[0069] In the formula, For the The initial weight of each particle; is a measure of the remaining useful life of the fuel cell; For the The predicted value of the remaining service life of the fuel cell corresponding to each particle; For the The predicted value of the remaining service life of the fuel cell corresponding to each particle; is the parameter vector of the state transition model.

[0070] The particles are resampled according to the particle weights, and the parameters of the initial fuel cell degradation model are estimated by combining the resampled particle set to obtain the fuel cell degradation empirical model.

[0071] A fuel cell life prediction model is established based on a long short-term memory neural network. The fuel cell life prediction model is trained using historical operation data in the training set. The trained fuel cell life prediction model is obtained by repeated iterations until the training termination condition is reached. The process includes:

[0072] The historical operation data in the training set is used as training samples, and the training samples include working time, fuel cell voltage, current, single chip voltage, intake pressure, exhaust pressure, intake metering ratio, inlet and outlet cooling water temperature, and actual working life of the fuel cell; the actual working life of the fuel cell is used as the true label;

[0073] The historical operation data in the training set is input into the fuel cell life prediction model to obtain the battery life training value; the training loss value is calculated based on the battery life training value and the true label; the Adam optimizer is used to optimize the parameters of the fuel cell life prediction model according to the training loss value, and the training process of the fuel cell life prediction model is repeated until the number of iterations reaches the iteration threshold to output the trained fuel cell life prediction model.

[0074] The fuel cell degradation empirical model and the fuel cell life prediction model are tested using the prediction set. When the accuracy of the fuel cell degradation empirical model and the fuel cell life prediction model is less than the set accuracy threshold, the historical operation data is retrieved to optimize the fuel cell degradation empirical model and the fuel cell life prediction model.

[0075] When the accuracy of the fuel cell degradation empirical model and the fuel cell life prediction model reaches a set accuracy threshold, the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model are set using the comparison set, and the process includes:

[0076] Inputting the comparison set into the fuel cell degradation empirical model to obtain the model driving function of the fuel cell remaining useful life probabilities distribution;

[0077] The comparison set is input into a fuel cell life prediction model to obtain a data-driven function of the fuel cell remaining service life probabilities distribution;

[0078] Establishing the actual distribution function of the service life of the fuel cell based on the actual working life of the fuel cell;

[0079] The model weights of the fuel cell degradation empirical model and the fuel cell life prediction model are calculated based on the model driving function, data driving function and actual service life distribution function. The expression formula is: ;

[0080] In the formula, is the actual distribution function of service life; A model-driven function for the fuel cell remaining useful life probabilities distribution; A data driven function for the distribution of the remaining useful life of a fuel cell; is the model weight of the fuel cell degradation empirical model; is the model weight of the fuel cell life prediction model; To ensure the accuracy of the prediction results;

[0081] When accuracy When it reaches 1, the model weight of the fuel cell degradation empirical model is obtained. And the model weight of the fuel cell life prediction model .

[0082] The fuel cell degradation empirical model and the fuel cell life prediction model are superimposed according to the model weight to obtain a battery life prediction hybrid model; by balancing the accuracy and uncertainty of the fuel cell degradation empirical model and the fuel cell life prediction model, the battery life prediction hybrid model can more accurately predict the remaining service life of the fuel cell.

[0083] Acquire monitoring data of the fuel cell, input the monitoring data into the battery life prediction hybrid model to obtain the remaining life evaluation value of the fuel cell; generate maintenance instructions for the fuel cell based on the remaining life evaluation value, the process includes:

[0084] The life factor of the fuel cell is calculated based on the remaining life evaluation value, and the expression formula is: ;

[0085] The formula is, is the lifespan factor; It is the estimated value of the remaining life of the fuel cell; The service life of the fuel cell;

[0086] When the life factor is less than the life threshold, a maintenance instruction for the fuel cell is generated; warnings can be issued in time and planned repairs and maintenance can be carried out to reduce the failure rate of the fuel cell system and improve the reliability and stability of the system; the intelligent management system can reduce the need for manual monitoring and realize automated and intelligent maintenance plans, thereby reducing overall operating costs.

[0087] Example 2

[0088] This embodiment provides an intelligent management system for a fuel cell, the intelligent management system is used to execute the intelligent management method described in Embodiment 1, and the intelligent management system includes:

[0089] A monitoring module, used to obtain monitoring data of the fuel cell, input the monitoring data into a battery life prediction hybrid model to obtain a remaining life evaluation value of the fuel cell; and generate a maintenance instruction for the fuel cell based on the remaining life evaluation value;

[0090] An acquisition module is used to acquire historical operation data of the fuel cell; and divide the historical operation data into a training set and a comparison set respectively;

[0091] The first building module is to establish an initial model of fuel cell degradation; input the historical operation data in the training set into the initial model of fuel cell degradation, and use a particle filter algorithm to estimate the parameters of the initial model of fuel cell degradation to obtain an empirical model of fuel cell degradation;

[0092] The second building block is to establish a fuel cell life prediction model based on a long short-term memory neural network, use the historical operation data in the training set to train the fuel cell life prediction model, and repeat the iteration until the training termination condition is reached to obtain the trained fuel cell life prediction model;

[0093] The fitting module uses the comparison set to set the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model, and then superimposes the fuel cell degradation empirical model and the fuel cell life prediction model according to the model weights to obtain a battery life prediction hybrid model.

[0094] The intelligent management system is equipped with a monitoring screen; the monitoring screen provides multi-dimensional data statistical analysis results such as equipment macro data, single machine monitoring, and fault statistics to help operators better understand equipment status and performance

[0095] The monitoring module is integrated into the data cloud platform, realizing the integrated management of equipment monitoring, data processing, data storage and data forwarding, and improving the efficiency and accuracy of data processing.

[0096] Example 3

[0097] This embodiment provides an electronic device, including a storage medium and a processor; the storage medium is used to store instructions; the processor is used to operate according to the instructions to execute the intelligent management method for fuel cells described in Example 1.

[0098] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0099] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0100] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0101] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0102] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intelligent management method for a fuel cell, characterized in that: include: Acquire monitoring data of the fuel cell, and input the monitoring data into a battery life prediction hybrid model to obtain a remaining life evaluation value of the fuel cell; generating a maintenance instruction for the fuel cell based on the remaining life estimate; The construction process of the battery life prediction hybrid model includes: Acquire historical operation data of the fuel cell; divide the historical operation data into a training set and a comparison set; Establishing an initial model of fuel cell degradation; inputting historical operation data in the training set into the initial model of fuel cell degradation, and using a particle filter algorithm to estimate the parameters of the initial model of fuel cell degradation to obtain an empirical model of fuel cell degradation; A fuel cell life prediction model is established based on a long short-term memory neural network, and the fuel cell life prediction model is trained using historical operation data in the training set. The trained fuel cell life prediction model is obtained by repeated iterations until the training termination condition is reached. After setting the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model using the comparison set, the fuel cell degradation empirical model and the fuel cell life prediction model are superimposed according to the model weights to obtain a battery life prediction hybrid model.

2. The intelligent management method for fuel cells according to claim 1, characterized in that: The process of developing an initial fuel cell degradation model includes: The actual output voltage of the fuel cell is used to represent the remaining service life of the fuel cell, and the initial fuel cell degradation model is constructed; the expression formula is: ； ； ； ； ； In the formula, is the actual output voltage of the fuel cell; is the Nernst voltage of the fuel cell; is the activation loss of the fuel cell; is the ohmic loss of the fuel cell; is the concentration loss of the fuel cell; is the change in Gibbs free energy before and after the fuel cell reaction; is the Faraday constant; is the universal gas constant; is the entropy change before and after the fuel cell reaction; is the fuel cell operating temperature; is the partial pressure of hydrogen; is the partial pressure of oxygen; is the thermodynamic standard state pressure; is the thermodynamic standard state temperature; , , , is the fitting parameter; is the current density per unit electrode area; is the electrode area; is the charge transfer resistance per unit area; is the limiting current density.

3. The intelligent management method for fuel cells according to claim 2, characterized in that: The historical operation data in the training set is input into the initial model of fuel cell degradation, and the particle filter algorithm is used to estimate the parameters of the initial model of fuel cell degradation to obtain the empirical model of fuel cell degradation. The process includes: The initial measurement value of the remaining service life of the fuel cell is obtained from the historical operation data in the training set, and the measurement noise is set to conform to the normal distribution; the initial state value of the remaining service life of the fuel cell is obtained based on the initial measurement value and the measurement noise; Generate a set number based on the initial state value of the remaining service life of the fuel cell Particles; set the initial weight of the particles, the expression formula is: ; is the initial weight of the i-th particle; The initial fuel cell degradation model is used as a state transition model; the particles are input into the state transition model to obtain the predicted value of the fuel cell usage state at the next moment, and the measured value of the remaining service life of the fuel cell at the next moment is obtained from the historical operation data in the training set. The particle weight is updated according to the predicted value and measured value of the remaining service life of the fuel cell, and the particle weight is normalized; The particles are resampled according to the particle weights, and the parameters of the initial fuel cell degradation model are estimated by combining the resampled particle set to obtain the fuel cell degradation empirical model.

4. The intelligent management method for fuel cells according to claim 3, characterized in that: Update particle weights based on predicted and measured values ​​of the remaining useful life of the fuel cell, including: ； In the formula, is the standard deviation of the measurement noise; is the updated weight of the particle; is a measure of the remaining useful life of the fuel cell; is the predicted value of the remaining service life of the fuel cell; is the ratio of pi; is the time step index.

5. The intelligent management method for fuel cells according to claim 3, characterized in that: Normalize the particle weights, including: ； In the formula, For the The initial weight of each particle; is a measure of the remaining useful life of the fuel cell; For the The predicted value of the remaining service life of the fuel cell corresponding to each particle; For the The predicted value of the remaining service life of the fuel cell corresponding to each particle; is the parameter vector of the state transition model.

6. The intelligent management method for fuel cells according to claim 1, characterized in that: The fuel cell life prediction model is trained using historical operating data in the training set, and repeated iterations are performed until the training termination condition is reached to obtain the trained fuel cell life prediction model. The process includes: The historical operation data in the training set is used as training samples, and the training samples include working time, fuel cell voltage, current, single chip voltage, intake pressure, exhaust pressure, intake metering ratio, inlet and outlet cooling water temperature, and actual working life of the fuel cell; the actual working life of the fuel cell is used as the true label; The historical operation data in the training set is input into the fuel cell life prediction model to obtain the battery life training value; the training loss value is calculated based on the battery life training value and the true label; the Adam optimizer is used to optimize the parameters of the fuel cell life prediction model according to the training loss value, and the training process of the fuel cell life prediction model is repeated until the number of iterations reaches the iteration threshold to output the trained fuel cell life prediction model.

7. The intelligent management method for fuel cells according to claim 1, characterized in that: The model weights of the fuel cell degradation empirical model and the fuel cell life prediction model are set using the comparison set. The process includes: Inputting the comparison set into the fuel cell degradation empirical model to obtain the model driving function of the fuel cell remaining useful life probabilities distribution; The comparison set is input into a fuel cell life prediction model to obtain a data-driven function of the fuel cell remaining service life probabilities distribution; Establishing the actual distribution function of the service life of the fuel cell based on the actual working life of the fuel cell; The model weights of the fuel cell degradation empirical model and the fuel cell life prediction model are calculated based on the model driving function, data driving function and actual service life distribution function. The expression formula is: ； In the formula, is the actual distribution function of service life; A model-driven function for the distribution of the remaining useful life of a fuel cell; A data driven function for the fuel cell remaining useful life probabilities distribution; is the model weight of the fuel cell degradation empirical model; is the model weight of the fuel cell life prediction model; To ensure the accuracy of the prediction results; When accuracy When it reaches 1, the model weight of the fuel cell degradation empirical model is obtained. And the model weight of the fuel cell life prediction model .

8. The intelligent management method for fuel cells according to claim 1, characterized in that: Generate fuel cell maintenance instructions based on the remaining life assessment value, the process includes: The life factor of the fuel cell is calculated based on the remaining life evaluation value, and the expression formula is: ； The formula is, is the lifespan factor; It is the estimated value of the remaining life of the fuel cell; The service life of the fuel cell; When the life factor is less than the life threshold, a maintenance instruction for the fuel cell is generated.

9. An intelligent management system for a fuel cell, characterized in that: include: A monitoring module is used to obtain monitoring data of the fuel cell and input the monitoring data into a battery life prediction hybrid model to obtain a remaining life evaluation value of the fuel cell; generating a maintenance instruction for the fuel cell based on the remaining life estimate; An acquisition module is used to acquire historical operation data of the fuel cell; and divide the historical operation data into a training set and a comparison set respectively; The first building module is to establish an initial model of fuel cell degradation; input the historical operation data in the training set into the initial model of fuel cell degradation, and use a particle filter algorithm to estimate the parameters of the initial model of fuel cell degradation to obtain an empirical model of fuel cell degradation; The second building block is to establish a fuel cell life prediction model based on a long short-term memory neural network, use the historical operation data in the training set to train the fuel cell life prediction model, and repeat the iteration until the training termination condition is reached to obtain the trained fuel cell life prediction model; The fitting module uses the comparison set to set the model weights of the fuel cell degradation empirical model and the fuel cell life prediction model, and then superimposes the fuel cell degradation empirical model and the fuel cell life prediction model according to the model weights to obtain a battery life prediction hybrid model.

10. An electronic device comprising a storage medium and a processor; the storage medium is used to store instructions; characterized in that: The processor is used to operate according to the instructions to execute the intelligent management method for a fuel cell according to any one of claims 1 to 8.

Images