CN117874665A - SOFC system multi-fault diagnosis method and system - Google Patents

Abstract

The application relates to the technical field of fuel cells and provides a method and a system for diagnosing multiple faults of an SOFC system, wherein the method comprises the steps of obtaining multiple fault data in the SOFC system and forming a fault data set; preprocessing a fault data set by adopting a LightGBM algorithm, and generating a characteristic data set; adopting a multi-layer convolutional neural network to perform feature extraction on the feature data set, and converting the feature data set into a final feature data set; based on the final characteristic data set, decoupling classification is carried out on the multiple faults of the SOFC system through a Sigmoid function so as to formulate a decoupling classification diagnosis strategy and output fault types. According to the method, the Sigmoid function is used as a nonlinear activation function of the classifier, so that probability values of each class output can be guaranteed to be independent of each other, the decoupling classification function is achieved, and the problem that the following characteristic mixing is difficult to diagnose is solved.

Description

SOFC system multi-fault diagnosis method and system

Technical Field

The present disclosure relates to the field of fuel cell technologies, and in particular, to a method and a system for diagnosing multiple faults of an SOFC system.

Background

With the continuous development of society, energy sources are becoming more important for human survival and development. However, the progressive consumption of energy and the low efficiency of primary energy conversion to electrical energy present significant challenges for future sustainable development. A Fuel Cell (FC) is a converter capable of clean power generation, capable of converting chemical energy in Fuel into electric energy, and capable of reducing the generation of pollutants. In particular, solid oxide fuel cells (Solid Oxide Fuel Cell, SOFC) are attracting attention, which have advantages of high energy efficiency, low emission, low noise, and the like. At present, most of SOFC materials are perovskite (La_ (0.6) Sr_ (0.4) CoO_ (3-delta), LSC type and zirconia (Yttria-stabilized zirconia, YSZ) type, the internal working temperature is up to 600-1000 ℃, and serious consequences are caused once the SOFC breaks down, but the damage of the internal cell materials and the structure is difficult to directly observe from the outside due to the high tightness of the SOFC, and the important state of the inside of a cell stack affected by the fault cannot be directly measured, so that the research on SOFC fault diagnosis is severely restricted.

Meanwhile, the commercial application of the SOFC mainly faces the problems of high cost and short service life, and the fault diagnosis can timely discover and isolate the fault problem in the running of the system, so that the SOFC has important significance in improving the durability of the system and reducing the maintenance cost. However, SOFCs do not only suffer from single failure, but are more often prone to multiple failure simultaneous, coupled with multiple failure effects and mixed-signature problems, making failure diagnosis very challenging.

Disclosure of Invention

In order to avoid the defects of the prior art, the invention provides a multi-fault diagnosis method and system for an SOFC system, which are used for solving the problems that the SOFC system commonly generates multiple faults and is influenced by coupling of the multiple faults and mixed characteristics in the prior art.

The purpose of the application is realized by adopting the following technical scheme:

in a first aspect, the present application provides a method for diagnosing multiple faults of an SOFC system, including:

acquiring multi-fault data in the SOFC system and forming a fault data set;

preprocessing the fault data set by adopting a LightGBM algorithm, and generating a characteristic data set;

adopting a multi-layer convolutional neural network to perform feature extraction on the feature data set, and converting the feature data set into a final feature data set;

based on the final characteristic data set, decoupling classification is carried out on the multiple faults of the SOFC system through a Sigmoid function so as to formulate a decoupling classification diagnosis strategy and output fault types.

In one possible implementation, the acquiring the multiple failure data in the SOFC system and forming the failure data set includes:

acquiring data of a plurality of models to build an SOFC structure model;

and analyzing the failure occurrence mechanism of the SOFC system, and selecting a plurality of characteristic parameters to establish a failure data set of the SOFC.

In one possible implementation, a plurality of the models includes an air compressor model, a fuel and air heat exchanger model, a mixer model, a bypass valve model, a combustor model, and a galvanic pile model;

the plurality of characteristic parameters include characteristic data of voltage, connector, power and hydrogen flow.

In a possible implementation manner, the preprocessing the fault data set by adopting the LightGBM algorithm, and generating feature data, includes:

performing tree feature generation based on the LightGBM algorithm so that the plurality of feature parameters form first feature data;

and performing splicing processing on the first characteristic data and the fault data set to form the characteristic data set.

In a possible implementation manner, the feature extraction of the feature data set by using the multi-layer convolutional neural network and conversion to a final feature data set includes:

performing feature extraction on the feature data set by adopting a convolution layer and a global pooling layer, and generating second feature data;

and performing splicing processing on the second characteristic data and the characteristic data set to form a final characteristic data set.

In a possible implementation manner, the feature extraction of the feature data set by using a convolution layer and a global pooling layer includes:

feature extraction is accomplished by alternating multiple of the convolution layer and global pooling layer.

In a possible implementation manner, the performing decoupling classification on the SOFC system multi-fault by using a Sigmoid function based on the final feature data set to formulate a decoupling classification diagnosis policy includes:

establishing a diagnostic classifier based on the final feature dataset;

and taking the Sigmoid function as a nonlinear activation function of the diagnosis classifier to output a probability prediction value of the corresponding class.

In a possible implementation manner, the performing decoupling classification on the SOFC system multi-fault by using a Sigmoid function based on the final feature data set to formulate a decoupling classification diagnosis policy, and further includes:

setting confidence thresholdDetermining the output of the diagnostic classifier, a loss function being the boundary loss by setting the confidence threshold +.>To output the corresponding fault type.

In one possible implementation, the multi-fault data includes stack faults, fan faults, condenser faults, fuel leakage faults, and reformer faults;

wherein the stack faults include electrode delamination faults, sealing faults, anode sulfur poisoning faults and cathode humidification faults.

In a second aspect, the present application further provides a diagnostic system for multiple failures of an SOFC system, including:

the data acquisition module is used for acquiring multi-fault data in the SOFC system and forming a fault data set;

the data processing module is used for preprocessing the fault data set by adopting a LightGBM algorithm and generating a characteristic data set;

the characteristic extraction module is used for carrying out characteristic extraction on the characteristic data set by adopting a multi-layer convolutional neural network and converting the characteristic data set into a final characteristic data set;

and the decoupling classification module is used for performing decoupling classification on the SOFC system multi-fault based on the final characteristic data set, and is used for preparing a decoupling classification diagnosis strategy and outputting fault types.

The technical scheme provided by the embodiment of the application can comprise the following beneficial effects:

the method adopts a mixed integrated model combining a tree-based method and a deep learning-based method for fault diagnosis of the solid oxide fuel cell, and uses a LightGBM algorithm for tree feature generation. Compared with the traditional algorithm, the proposed model has better prediction performance and stronger robustness.

In addition, the Sigmoid function is used as a nonlinear activation function of the classifier, so that the probability value of each output class can be ensured to be mutually independent, thereby playing a role in decoupling classification, and further solving the problem that the following feature mixing is difficult to diagnose when multiple faults occur simultaneously

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

Drawings

The present application is further described below with reference to the drawings and examples.

FIG. 1 is one of the flow schematic diagrams of a method for diagnosing multiple faults of a SOFC system in accordance with an embodiment of the present application;

FIG. 2 is a second flow chart of a method of diagnosing multiple faults in a SOFC system according to an embodiment of the present application;

FIG. 3 is a schematic diagram of a LightGBM algorithm according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a SOFC failure diagnosis network model in accordance with an embodiment of the present application;

FIG. 5 is a schematic diagram of a framework of a diagnostic system for multiple failures of a SOFC system in accordance with an embodiment of the application;

fig. 6 is a schematic structural view of a display device in an embodiment of the present invention;

fig. 7 is a schematic structural diagram of a program product for an interface synchronization display method according to an embodiment of the present invention.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. However, the exemplary embodiments may be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The described features may be combined in any suitable manner in one or more embodiments.

Furthermore, the drawings are only schematic illustrations of the present application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities.

In this exemplary embodiment, first, a method for diagnosing multiple faults of an SOFC system is provided, referring to fig. 1 and 2, the method includes:

step S101, multi-fault data in the SOFC system are obtained, and a fault data set is formed.

Step S102, preprocessing the fault data set by adopting a LightGBM algorithm, and generating a characteristic data set.

And step S103, adopting a multi-layer convolutional neural network, carrying out feature extraction on the feature data set through a Sigmoid function, and converting the feature data set into a final feature data set.

And step S104, decoupling classification is carried out on the multiple faults of the SOFC system based on the final characteristic data set so as to formulate a decoupling classification diagnosis strategy and output fault types.

In step S101, the solid oxide fuel cell, i.e. SOFC, has a plurality of fault types, and in one example, the multi-fault data includes stack fault, fan fault, condenser fault, fuel leakage fault, and reformer fault; the pile faults comprise electrode layering faults, sealing faults, anode sulfur poisoning faults and cathode humidifying faults.

In order to obtain multi-fault data in the SOFC system, the example may be obtained by building an SOFC structure test model, and specifically, in step S101, step S1011 is further included to obtain data of a plurality of models, so as to build an SOFC structure model; step S1012, analyzing the SOFC failure occurrence mechanism, and selecting a plurality of characteristic parameters to establish a failure data set of the SOFC.

In some embodiments, the plurality of models includes an air compressor model, a fuel and air heat exchanger model, a mixer model, a bypass valve model, a combustor model, and a galvanic pile model; the plurality of characteristic parameters includes characteristic data of voltage, connection, power and hydrogen flow in the SOFC.

It should be noted that, each subcomponent and connection mode of the SOFC structural model constructed in this example are consistent with the actual process flow. Characteristic data such as voltage, connector, power and hydrogen flow in the SOFC are obtained, an SOFC fault data set is established, faults are described by adopting the state quantities greatly influenced by the faults, and the maximum characteristic difference value among the faults and the maximum possible characteristic mixing avoidance can be ensured.

As shown in fig. 3, in step S102, the LightGBM is a model of multiple decision trees, which has a weak expression capability compared to a single tree model, and is insufficient to express multiple distinguishing feature combinations (e.g., a connection body versus voltage in a solid oxide fuel cell), while multiple trees have a stronger expression capability, a better residual learning capability, a smaller error, and can find effective features and feature combinations.

Step S102 includes step S1021, performing feature generation of a tree based on a LightGBM algorithm, so that a plurality of feature parameters form first feature data; step S1022, performing a stitching process on the first feature data and the fault data set to form a feature data set.

Specifically, the characteristics of the tree are generated by utilizing gradient single-side sampling, mutual exclusion characteristic binding, histogram and leaf growth of a LightGBM algorithm, in other words, a continuous characteristic value is divided into discrete blocks, then a decision tree growth strategy according to leaf growth is used, each independent tree is selected as a classification characteristic by selecting a leaf node with the largest increment loss, and the characteristics of voltage, connector, power, hydrogen flow and the like in the SOFC can be accurately fallen onto each leaf according to a leaf splitting method; the leaf number corresponding to the feature falling onto the leaf is subjected to feature coding to form new feature data (first feature data). Newly generated data (first characteristic data) and original characteristic data (failure data set)/(original characteristic data)>Stitching to form new features (feature data set）/>。

In step S103, the obtained data is processedDescribed as +.>Matrix of dimensions:

（1）

in the method, in the process of the invention,the feature class size is input for the corresponding fault diagnosis.

It will also be understood thatAnd inputting the high-dimensional sparse features into a convolution module, representing each classification feature as a low-dimensional vector, mapping the high-dimensional sparse features into a matrix embedding matrix (1) by using one-hot coding, and reconstructing the matrix embedding matrix into an input matrix of a first convolution layer.

Step S103 includes step S1031, performing feature extraction on the feature data set by using a convolution layer and a global pooling layer, and generating second feature data; step S1032, performing a stitching process on the second feature data and the feature data set to form a final feature data set.

Specifically, referring to fig. 4, the matrix (1) as the input of the convolution layer may be transformed into a feature map shown in the following formula after convolution calculation:

（2）

in the method, in the process of the invention,representing the ith convolution kernelMapping the out features; sign->Representing a convolution operation; />Denoted as the ith convolution kernel (>)；/>Bias terms for corresponding convolution kernels; />An input matrix representing a j-th convolution kernel; />Representing the characteristic size of the convolutional layer output.

In order to preserve the spatial dimension information of the input data, a mode of the Same convolution is adopted, namely, proper zero padding is added around the input data in the convolution process, the padding size P of the Same convolution can be automatically calculated according to the size and step size of a convolution kernel,the calculation can be performed according to the following formula, and the calculation can be performed downward when the calculation cannot be performed completely. The specific definition is as follows:

（3）

（4）

in the method, in the process of the invention,for the size of the input data +.>For the size of the convolution kernel, +.>To add the size of zero padding around the input data,/->Represented as convolution kernel sliding steps.

When convolutional layer output mapping featuresAfter that, the problem of redundancy of information features still exists, and the feature data is not obvious, so that the pooling layer is utilized to reduce the dimension of the output of the upper layer, and the main features are selected. The traditional feature extraction method is that the convolution layer carries out dimension reduction treatment through a plurality of full-connection layers after passing through the pooling layer, so that the full-connection layers have a plurality of parameters, the feature extraction speed is reduced, and the fitting condition is easy to occur. Therefore, the global average pooling (Global Average Pooling) method is adopted, network parameters are greatly reduced, the dimension reduction is realized with small calculation cost, and the formula is shown as follows:

（5）

in the method, in the process of the invention,representing a global average pooled output value mapped to a kth feature; />Representing that the kth feature is mapped in the region +.>Middle is located at->An output at; />Representing the number of all elements of the kth feature map.

Through convolution and global poolAfter the chemical operation, new important characteristics (second characteristic data) are generated. The convolved features can learn not only the relation between adjacent features, but also the relation between non-adjacent features (for example, the relation between the stack temperature and the hydrogen flow in a solid oxide fuel cell), and the newly generated data (second feature data) is spliced with the data (feature data set) generated by the LightGBM to form final data (final feature data set)。

In one example, feature extraction is accomplished by alternating multiple of the convolutional layers and global pooling layers. Specifically, feature extraction is needed to be completed through mutual alternation of multiple convolution layers and pooling layers when feature extraction is performed on SOFC fault data, and internal parameters can be automatically obtained according to back propagation through a loss function in a training process, so that the number of convolution kernels of different layers is only needed to be set.

In step S104, after the SOFC failure data is extracted by multi-layer rolling and pooling, the final feature data set obtained by the last layer from which the features are extracted is obtainedDescribed as->Is expanded into an m-dimensional vector by dimensional transformation

As shown in fig. 4, step S104 includes step S1041 of building a diagnostic classifier based on the final feature dataset; step S1042, using Sigmoid function as nonlinear activation function of the diagnostic classifier to output probability prediction value of corresponding class; step S1043, setting a confidence thresholdDetermining the output of the diagnostic classifier, the loss function being a boundary loss by setting theConfidence threshold->To output the corresponding fault type.

In particular, the fault diagnosis of SOFCs corresponds to classification problems, but in actual situations, a plurality of faults often occur in a combined manner. Therefore, the non-linear activation function of the classification layer needs to be reselected to meet that the last layer can output multiple fault class labels for the composite fault. The Sigmoid function is a function related to logistic regression, when the Sigmoid function is used as a nonlinear activation function of a diagnosis classifier, the output value of the feature map output by the global pooling layer after the Sigmoid classifier function of the classification layer is within [0,1], and the probability value output by each neuron of the classification layer is mutually independent, so that the Sigmoid function can be used as an activation function of a compound fault classifier.

（6）

In the method, in the process of the invention,outputting a probability prediction value of the k-th class of the fault by the classification layer; />Representing the last layer of kth output vector values extracted for the fault signature.

At this time, a confidence threshold value needs to be setTo determine the output of the diagnostic classifier, the loss function is a boundary loss by setting a confidence threshold +.>To output the corresponding fault type.

By setting confidence thresholdOutputting the fault class corresponding to the fault label. The relation between the corresponding prediction output label and the threshold value can be obtainedDescribed by the following formula, the corresponding classification diagnostic strategy is as follows:

（7）

in the method, in the process of the invention,a predictive tag of a kth class; />I.e. as the kth category exists.

Meanwhile, when decoupling and classifying SOFC composite faults, boundary loss functions are adopted as optimization objective functions, and the definition of the corresponding boundary loss functions is as follows:

（8）

in the method, in the process of the invention,is a corresponding indicator function, which is expressed as +.f when the prediction sample belongs to class k>And when the prediction sample does not belong to category k +.>The value of (2) is 0; />When the fault category belongs to k, outputting a lower bound of a prediction probability value by the classifier; />Outputting an upper bound of the prediction probability value by the classifier when the corresponding prediction fault category does not belong to the category k; />Penalty coefficients are weights.

Thus, the construction of the convolutional neural network SOFC system fault structure diagnosis model is completed.

The optimization of the above-described composite fault diagnosis model may be performed by minimizing the boundary loss function. The boundary loss function not only can expand the differences among classes based on the Euclidean distance, but also can reduce the differences in the classes, thereby being beneficial to optimizing the network model and improving the accuracy of the diagnosis model.

The diagnostic mode of the application can be briefly summarized as that firstly, the LightGBM algorithm is used for generating the characteristics, SOFC operation parameters are used as initial input data according to the characteristics of the SOFC according to the leaf splitting, nodes falling on the leaves are used as new characteristics to be encoded so as to enhance the characteristic correlation, and then the nodes are used as the input of a subsequent model. And then, the strong feature extraction capability of the convolutional neural network is utilized to perform feature extraction on data with high fault sensitivity, deeper feature interaction is generated, and feature sparsity is reduced after convolution and global average pooling, so that non-adjacent features are deeply connected. And finally, constructing a SOFC system fault decoupling diagnosis model by utilizing the decoupling characteristic activated by the Sigmoid function.

The method is used for fault diagnosis of the SOFC by adopting a hybrid integrated model combining a tree-based method and a deep learning-based method, and the model uses a LightGBM algorithm for tree feature generation. Compared with the traditional algorithm, the proposed model has better prediction performance and stronger robustness.

In addition, the Sigmoid function is used as a nonlinear activation function of the classifier, so that probability values of each output class can be ensured to be independent, the decoupling classification function is realized, and the problem that the following feature mixing is difficult to diagnose is solved.

Also provided in this example embodiment is a diagnostic system for multiple failures of an SOFC system, please refer to fig. 5, the diagnostic system comprising: the data acquisition module 201, the data processing module 202, the feature extraction module 203 and the decoupling classification module 204.

The data acquisition module 201 is configured to acquire multiple fault data in the SOFC system, and form a fault data set; the data processing module 202 is configured to pre-process the fault data set by adopting a LightGBM algorithm, and generate a feature data set; the feature extraction module 203 is configured to perform feature extraction on the feature data set by using a multi-layer convolutional neural network, and convert the feature data set into a final feature data set; the decoupling classification module 204 performs decoupling classification on the SOFC system multiple faults based on the final feature data set, and is used for preparing a decoupling classification diagnosis strategy and outputting fault types.

It should be noted that, the principle of the diagnostic system provided in this example may be understood by referring to the above diagnostic method, and will not be described herein.

It should be noted that although the various steps of the methods herein are depicted in the accompanying drawings in a particular order, this does not require or imply that the steps be performed in that particular order, or that all illustrated steps be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step to perform, and/or one step decomposed into multiple steps to perform, etc. In addition, it is also readily understood that these steps may be performed synchronously or asynchronously, for example, in a plurality of modules/processes/threads.

It should be noted that although several units and modules of the system for action execution are mentioned in the above detailed description, this division is not mandatory. Indeed, the features and functions of two or more units or modules described above may be embodied in one unit or module, in accordance with embodiments of the present application. Conversely, the features and functions of one unit or module described above may be further divided into a plurality of units or modules to be embodied. The components shown as units or modules may or may not be physical units, may be located in one place, or may be distributed across multiple network elements. Some or all of the units or modules may be selected according to actual needs to achieve the purposes of the present application. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

Referring to fig. 6, an embodiment of the present invention also provides a display device 300, the display device 300 including at least one memory 310, at least one processor 320, and a bus 330 connecting different platform systems.

Memory 310 may include readable media in the form of volatile memory, such as Random Access Memory (RAM) 311 and/or cache memory 312, and may further include Read Only Memory (ROM) 313.

The memory 310 further stores a computer program, where the computer program may be executed by the processor 320, so that the processor 320 executes the steps of the interface synchronization display method in any embodiment of the present invention, and a specific implementation manner of the computer program is consistent with the implementation manner and the achieved technical effect described in the embodiment of the interface synchronization display method, and some contents are not repeated.

Memory 310 may also include utility 314 having at least one program module 315, such program modules 315 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of which may include an implementation of a network environment.

Accordingly, processor 320 may execute the computer programs described above, as well as may execute utility 314.

Bus 330 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

The display device 300 may also communicate with one or more external devices 340, such as a keyboard, pointing device, bluetooth device, etc., as well as with one or more devices capable of interacting with the display device 300, and/or with any device (e.g., router, modem, etc.) that enables the display device 300 to communicate with one or more other computing devices. Such communication may occur through input-output interface 350. Also, the display device 300 may communicate with one or more networks, such as a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network, such as the Internet, through a network adapter 360. The network adapter 360 may communicate with other modules of the display device 300 via the bus 330. It should be appreciated that although not shown, other hardware and/or software modules may be used in connection with display device 300, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, data backup storage platforms, and the like.

The embodiment of the invention also provides a computer readable storage medium, which is used for storing a computer program, the computer program is executed to realize the steps of the interface synchronous display method in the embodiment of the invention, the specific implementation manner is consistent with the implementation manner and the achieved technical effect recorded in the embodiment of the interface synchronous display method, and part of contents are not repeated.

Fig. 7 shows a program product 400 provided in this embodiment for implementing the above-described interface synchronization display method, which may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product 400 of the present invention is not limited thereto, and in the present invention, the readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. Program product 400 may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The computer readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, with readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A readable storage medium may also be any readable medium that can transmit, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing. Program code for carrying out operations of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++ or the like and conventional procedural programming languages, such as the C programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of remote computing devices, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., connected via the Internet using an Internet service provider).

The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the claims, which are to be protected by the present invention.

Claims (10)

1. A method for diagnosing multiple faults in an SOFC system, comprising:

acquiring multi-fault data in the SOFC system and forming a fault data set;

preprocessing the fault data set by adopting a LightGBM algorithm, and generating a characteristic data set;

adopting a multi-layer convolutional neural network to perform feature extraction on the feature data set, and converting the feature data set into a final feature data set;

based on the final characteristic data set, decoupling classification is carried out on the multiple faults of the SOFC system through a Sigmoid function so as to formulate a decoupling classification diagnosis strategy and output fault types.

2. The method for diagnosing multiple faults in an SOFC system of claim 1, wherein the acquiring multiple fault data in the SOFC system and forming the fault data set includes:

acquiring data of a plurality of models to build an SOFC structure model;

and analyzing a fault occurrence mechanism of the SOFC system, and selecting a plurality of characteristic parameters to establish the fault data set.

3. The SOFC system multi-fault diagnostic method of claim 2, wherein a plurality of the models includes an air compressor model, a fuel and air heat exchanger model, a mixer model, a bypass valve model, a combustor model, and a galvanic pile model;

the plurality of characteristic parameters includes characteristic data of voltage, connector, power and hydrogen flow.

4. The SOFC system multi-fault diagnosis method of claim 2, wherein the preprocessing the fault dataset with the LightGBM algorithm and generating feature data includes:

performing tree feature generation based on the LightGBM algorithm so that the plurality of feature parameters form first feature data;

and performing splicing processing on the first characteristic data and the fault data set to form the characteristic data set.

5. The SOFC system multi-fault diagnosis method of claim 4, wherein the feature extraction and conversion of the feature data set into a final feature data set using a multi-layer convolutional neural network comprises:

performing feature extraction on the feature data set by adopting a convolution layer and a global pooling layer, and generating second feature data;

and performing splicing processing on the second characteristic data and the characteristic data set to form a final characteristic data set.

6. The SOFC system multi-fault diagnosis method of claim 5, wherein the feature extraction of the feature dataset with a convolution layer and a global pooling layer comprises:

feature extraction is accomplished by alternating multiple of the convolution layer and global pooling layer.

7. The SOFC system multi-fault diagnosis method of claim 5, wherein the decoupling classification of the SOFC system multi-fault by a Sigmoid function based on the final feature data set to formulate a decoupling classification diagnosis strategy comprises:

establishing a diagnostic classifier based on the final feature dataset;

and taking the Sigmoid function as a nonlinear activation function of the diagnosis classifier to output a probability prediction value of the corresponding class.

8. The SOFC system multi-fault diagnosis method of claim 7, wherein the decoupling classification of the SOFC system multi-fault by a Sigmoid function based on the final feature dataset to formulate a decoupling classification diagnosis policy further comprises:

setting confidence thresholdDetermining the output of the diagnostic classifier, a loss function being the boundary loss by setting the confidence threshold +.>To output the corresponding fault type.

9. The SOFC system multi-fault diagnostic method of any one of claims 1-8, wherein the multi-fault data includes stack fault, fan fault, condenser fault, fuel leakage fault, and reformer fault;

wherein the stack faults include electrode delamination faults, sealing faults, anode sulfur poisoning faults and cathode humidification faults.

10. A diagnostic system for multiple failures of an SOFC system, comprising:

the data acquisition module is used for acquiring multi-fault data in the SOFC system and forming a fault data set;

the data processing module is used for preprocessing the fault data set by adopting a LightGBM algorithm and generating a characteristic data set;

the characteristic extraction module is used for carrying out characteristic extraction on the characteristic data set by adopting a multi-layer convolutional neural network and converting the characteristic data set into a final characteristic data set;

and the decoupling classification module is used for performing decoupling classification on the SOFC system multi-faults through a Sigmoid function based on the final characteristic data set, and is used for preparing a decoupling classification diagnosis strategy and outputting fault types.