JP7404196B2 - Fault diagnosis device, fault diagnosis system, and fault diagnosis program - Google Patents

Description

The present invention relates to a fault diagnosis device, a fault diagnosis system, and a fault diagnosis program, and more particularly to a fault diagnosis device, a fault diagnosis system, and a fault diagnosis program for diagnosing a fault in a fuel cell cogeneration system.

Conventional failure location diagnosis systems use a logistic regression model to determine, from multiple explanatory variables representing operating information of the power generation system, that a predetermined location out of multiple locations in the power generation system is faulty and other locations are not. An objective variable representing probability is specified (for example, see Patent Document 1).

Japanese Patent Application Publication No. 2020-030757

According to the technology described in Patent Document 1, failure diagnosis is performed using a logistic regression model, so if a highly accurate logistic regression model can be used, there may be no need to perform failure diagnosis work on site. be. However, since fault diagnosis is performed using one logistic regression model, the diagnostic accuracy cannot be said to be sufficient. Particularly in devices with a large number of parts, such as fuel cell systems, normally healthy parts may be mistakenly determined to be faulty, increasing the number of parts to be repaired and increasing the work load on the person in charge of the work.

The present invention has been made in view of the above-mentioned circumstances, and provides a failure diagnosis device and a failure diagnosis device capable of accurately diagnosing a fuel cell cogeneration system while reducing the on-site burden on workers. The purpose is to provide a diagnostic system and a fault diagnosis program.

In order to achieve the above object, a failure diagnosis device according to one aspect of the present invention provides, when a fuel cell cogeneration system detects an abnormality in operation, the fuel cell cogeneration system detects a malfunction corresponding to the abnormality. an acquisition unit that acquires operation history data regarding the operation history of the fuel cell cogeneration system; and a multivariate analysis or machine learning by associating a group of operation history data obtained in the past with a functional unit of the fuel cell cogeneration system. The operation history data or the relevant operation acquired by the acquisition unit for each of a plurality of types of failure diagnosis models that are generated by the above and determine the presence or absence of a failure for each functional unit of the fuel cell cogeneration system. Estimating the presence or absence of a failure for each functional part of the fuel cell cogeneration system by inputting data obtained by processing historical data and using judgment results output from each of the plurality of types of failure diagnosis models. It has a section and a.

Further, a fault diagnosis system according to one aspect of the present invention includes a fuel cell cogeneration system and a fault diagnosis device connected to the fuel cell cogeneration system, and the fuel cell cogeneration system When an abnormality in operation is detected, operation history data related to the operation history of the own system corresponding to the abnormality is transmitted to the failure diagnosis apparatus, and the failure diagnosis apparatus transmits the operation history data from the fuel cell cogeneration system. an acquisition unit that acquires the operation history data obtained in the past; , processing the operation history data acquired by the acquisition unit or the operation history data for each of the plurality of types of failure diagnosis models that determine the presence or absence of a failure for each functional unit of the fuel cell cogeneration system. an estimation unit that inputs the obtained data and estimates the presence or absence of a failure for each functional unit of the fuel cell cogeneration system using determination results output from each of the plurality of types of failure diagnosis models. There is.

Further, a fault diagnosis program according to one aspect of the present invention causes a computer to function as each part included in the above-mentioned fault diagnosis apparatus.

As described in detail above, according to the present invention, failure diagnosis of a fuel cell cogeneration system can be performed with high accuracy while reducing the burden on the person in charge of the work at the site.

1 is a diagram showing an example of a configuration of a failure diagnosis system according to a first embodiment. FIG. 2 is a block diagram showing an example of the configuration of a CGS according to the first embodiment. 3 is a flowchart illustrating an example of the flow of processing by a control program according to the first embodiment. 1 is a block diagram showing an example of an electrical configuration of a failure diagnosis device according to a first embodiment. FIG. 1 is a block diagram showing an example of a functional configuration of a failure diagnosis device according to a first embodiment; FIG. 3 is a flowchart illustrating an example of the flow of processing by the failure diagnosis program according to the first embodiment. (A) is a graph showing an example of the determination result of the first failure diagnosis model. (B) is a graph showing an example of the determination result of the second failure diagnosis model. (C) is a graph showing an example of determination results when using the first failure diagnosis model and the second failure diagnosis model. 7 is a flowchart illustrating an example of the flow of learning processing by the failure diagnosis program according to the first embodiment. FIG. 2 is a block diagram illustrating an example of a functional configuration of a failure diagnosis device according to a second embodiment. FIG. 7 is a diagram for explaining precision and recall according to the second embodiment. It is a graph which shows an example of the determination result of a 1st failure diagnosis model. FIG. 2 is a block diagram illustrating an example of a functional configuration of a failure diagnosis device according to a second embodiment. It is a figure showing an example of the 1st data table concerning an embodiment. It is a figure showing an example of the 2nd data table concerning an embodiment. 12 is a flowchart illustrating an example of the flow of processing by a failure diagnosis program according to a third embodiment. 12 is a flowchart illustrating an example of the process flow of route diagnosis A according to the third embodiment. 12 is a flowchart illustrating an example of the process flow of route diagnosis B according to the third embodiment. 12 is a flowchart illustrating an example of the process flow of route diagnosis C according to the third embodiment. 13 is a flowchart illustrating an example of the process flow of route diagnosis E according to the third embodiment.

EMBODIMENT OF THE INVENTION Hereinafter, with reference to drawings, an example of the form for implementing this invention is demonstrated in detail.

[First embodiment]
FIG. 1 is a diagram showing an example of the configuration of a failure diagnosis system 100 according to the first embodiment.

As shown in FIG. 1, a failure diagnosis system 100 according to the present embodiment includes a fuel cell cogeneration system (hereinafter simply referred to as "CGS") 40, a failure diagnosis device 10, and a maintenance terminal 70. ing. The failure diagnosis system 100 is configured as a system for diagnosing a failure of the CGS 40.

The CGS 40 is, for example, ENE-FARM (registered trademark), and is installed at the user's home 30. The failure diagnosis device 10 is a device for diagnosing a failure of the CGS 40, and is connected to the CGS 40 via a network N. For example, a general-purpose computer device such as a server computer or a personal computer (PC) is applied to the failure diagnosis device 10. For example, a network such as a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet is applied to the network N.

The maintenance terminal 70 is a terminal owned by a worker 60 who performs maintenance work on the CGS 40. The maintenance terminal 70 is connected to the failure diagnosis device 10 via the network N. The maintenance terminal 70 may be, for example, a portable tablet terminal, a smartphone, a notebook PC, or the like.

The failure diagnosis device 10 according to the present embodiment inputs operation history data at the time of error detection to multiple types of failure diagnosis models when the CGS 40 detects an abnormality in operation (hereinafter referred to as an "error"). and perform fault diagnosis.

Specifically, in FIG. 1, the CGS 40 issues an error notification when an error is detected. At this time, the CGS 40 automatically transmits the operation history data of its own system to the failure diagnosis device 10. The operation history data is, for example, data related to at least one operation history at the time of error detection, several seconds before error detection (for example, 10 seconds or less), and several seconds after error detection (for example, 10 seconds or less). The operation history data includes, for example, detection values of various sensors, data representing the operation status of auxiliary equipment (eg, valves, pumps, etc.), error codes, and the like.

The failure diagnosis device 10 stores in advance a plurality of types of failure diagnosis models for each of the plurality of functional units included in the CGS 40 in the storage unit. The failure diagnosis model is a mathematical model for diagnosing a failure for each functional unit of the CGS 40. A functional part includes at least one component (eg, a pump, a valve, a pipe, a tank, etc.). The failure diagnosis model of the CGS 40 is a model generated by performing multivariate analysis or machine learning on a group of operation history data obtained in the past in association with functional parts. Multivariate analysis includes, for example, logistic regression. Examples of machine learning include random forests, neural networks, support vector machines, and the like. The fault diagnosis model uses, for example, a two-class classification that determines the presence or absence of a fault for each functional unit. The fault diagnosis device 10 processes the operation history data acquired from the CGS 40 as is or into data that can be input into a fault diagnosis model, and inputs the processed data into each of a plurality of types of fault diagnosis models. Each of the plurality of types of fault diagnosis models outputs "fault present (fault)" or "no fault (healthy)" for each functional unit. Specifically, when the index value indicated by the data is greater than or equal to the threshold value, "failure present" is output, and when the index value indicated by the data is less than the threshold value, "no failure" is outputted. The failure diagnosis device 10 estimates the presence or absence of a failure for each functional unit using the determination results output from each of the plurality of types of failure diagnosis models, and holds the diagnosis results of the failure diagnosis obtained by the estimation.

The worker (maintenance worker) 60 accesses the failure diagnosis device 10 from the maintenance terminal 70 and confirms the failure location indicated by the diagnosis result. Then, the worker 60 performs parts replacement work and the like at the site based on the diagnosis results obtained from the failure diagnosis device 10.

FIG. 2 is a block diagram showing an example of the configuration of the CGS 40 according to the first embodiment.

As shown in FIG. 2, the CGS 40 according to this embodiment is broadly divided into a control device 41 and a CGS main body 49. The control device 41 may be configured integrally with the CGS main body 49 or may be configured separately from the CGS main body 49.

The control device 41 functions as a controller for the CGS 40. The control device 41 includes a CPU (Central Processing Unit) 42, a ROM (Read Only Memory) 43, a RAM (Random Access Memory) 44, an input/output interface (I/O) 45, a storage section 46, and a communication section. 47, and a device interface (hereinafter referred to as "device I/F") 48.

The CPU 42, ROM 43, RAM 44, and I/O 45 are each connected via a bus. Each functional unit including a storage unit 46, a communication unit 47, and a device I/F 48 is connected to the I/O 45. Each of these functional units can communicate with the CPU 42 via the I/O 45.

As the storage unit 46, for example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), a flash memory, or the like is used. The storage unit 46 stores a control program 46A for controlling the operation of the CGS main unit 49. Note that this control program 46A may be stored in the ROM 43.

The control program 46A may be installed in the control device 41 in advance, for example. The control program 46A may be realized by being stored in a nonvolatile storage medium, or distributed via the network N, and installed in the control device 41 as appropriate. Note that examples of nonvolatile storage media include CD-ROM (Compact Disc Read Only Memory), magneto-optical disk, HDD, DVD-ROM (Digital Versatile Disc Read Only Memory), flash memory, memory card, etc. Ru.

The communication unit 47 is connected to a network N such as the Internet, LAN, WAN, etc., and is capable of communicating with the external failure diagnosis device 10 via the network N.

A CGS main body section 49 is connected to the device I/F 48. That is, various electrical components included in the CGS main body section 49 are communicably connected to the CPU 42 via the device I/F 48.

The CGS main body portion 49 has a known configuration. Specifically, the CGS main body section 49 includes a fuel cell unit 50. The fuel cell unit 50 is provided with a fuel cell module 51, a gas path 52 connected to the fuel cell module 51, a reformed water path 53, and an air path 54. The CGS main body section 49 includes a plurality of functional sections, as described above. This functional unit includes at least one component. This functional section may be, for example, a plurality of parts forming the gas path 52, a plurality of parts forming the reformed water path 53, or a plurality of parts forming the air path 54. This functional section may be an individual component constituting the CGS main body section 49. These parts include, for example, pumps, valves, piping, tanks, and the like.

The fuel cell module 51 includes a fuel cell stack 51A and a reformer 51B. The reformer 51B reforms gas (for example, city gas) containing a hydrocarbon raw material. In the fuel cell stack 51A, a plurality of fuel cells each having an electrolyte layer, a fuel electrode, and an air electrode are stacked. The fuel cell stack 51A is configured to generate electricity and water by causing hydrogen in the reformed gas reformed from city gas by the reformer 51B to react with oxygen in the air. The fuel cell unit 50 includes auxiliary equipment such as valves and pumps for adjusting the supply amount of reformed gas and air, and a heat transfer medium (for example, water) to transfer the heat generated by power generation by the fuel cell stack 51A. and a heat exchanger for exchanging heat between the two.

The CGS main body 49 includes a hot water storage tank. In the CGS main body section 49, when water is supplied to the hot water storage tank, the water is supplied from the hot water storage tank to the heat exchanger of the fuel cell unit 50, and the water is heated in this heat exchanger with the heat of the fuel cell stack 51A. When water is heated in a heat exchanger, it turns into hot water. This hot water is supplied to a hot water storage tank and stored in this hot water storage tank. The hot water storage tank includes auxiliary equipment such as a valve and a pump for transferring water and hot water to and from the fuel cell unit 50, and a valve and a pump for discharging hot water stored in the hot water storage tank.

Further, the gas route 52 is a route through which city gas flows. The gas path 52 includes auxiliary equipment such as valves and pumps for adjusting the amount of city gas supplied. The reformed water path 53 is a path through which reformed water flows. The reformed water path 53 includes auxiliary equipment such as valves and pumps for adjusting the amount of reformed water supplied. The air path 54 is a path through which air flows. The air path 54 includes auxiliary equipment such as valves and pumps for adjusting the amount of air supplied.

The control device 41 is a controller that controls auxiliary equipment provided in the CGS main body 49. The control device 41 outputs control signals for controlling the auxiliary equipment to the CGS main unit 49, respectively. Further, the CGS main body 49 outputs operational data representing the operating status of the CGS main body 49 to the control device 41. This operation data includes, as feature quantities, data representing the operating states of auxiliary equipment (for example, voltage values, current values, etc.), detection values of various sensors provided in the CGS main body 49, and the like. Specifically, the data representing the operating state of the auxiliary equipment is, for example, in the case of a pump, data representing the operating amount of the pump, and in the case of a valve, data representing the opening/closing state of the valve. Further, the detected values of various sensors are, for example, values representing the respective flow rates of city gas, reformed water, and air in the case of piping, and values representing the water level in the case of tanks.

The control device 41 stores the time-series operating data acquired from the CGS main unit 49 in the storage unit 46 . The control device 41 has a function of detecting errors in the CGS main body 49 based on operational data acquired from the CGS main body 49 . Note that error detection is performed, for example, by comparing a feature amount obtained from driving data with a corresponding threshold value.

Here, with reference to FIG. 3, error notification processing by the control device 41 according to the present embodiment will be specifically described. Note that this error reporting process is executed by a control program 46A included in the control device 41.

FIG. 3 is a flowchart showing an example of the flow of processing by the control program 46A according to the first embodiment.

First, the control program 46A stored in the storage unit 46 is activated by the CPU 42, and each step shown below is executed.

In step S101 of FIG. 3, the CPU 42 stores the operational data acquired in time series from the CGS main unit 49 in the storage unit 46, and determines whether an abnormal value indicating the occurrence of an error is detected based on the acquired operational data. Determine whether or not. If it is determined that an abnormal value has been detected (in the case of an affirmative determination), the process moves to step S102, and if it is determined that an abnormal value has not been detected (in the case of a negative determination), the process goes on standby in step S101.

In step S102, the CPU 42 selects, from among the time-series operation data stored in the storage unit 46, the time of error detection, several seconds before error detection (for example, 10 seconds or less), and several seconds after error detection (for example, 10 seconds or less). At least one operation data (seconds or less) is extracted as operation history data, and the extracted operation history data is held in the storage unit 46. As described above, this operation history data includes, for example, detection values of various sensors, data representing the operation status of auxiliary equipment, error codes, and the like. The operation history data held in the storage unit 46 is referred to as "data 1."

In step S103, the CPU 42 determines the error by retaining data 1 in step S102.

In step S104, the CPU 42 transmits the data 1 held in step S102 to the failure diagnosis device 10 via the communication unit 47, and ends the series of processing by the control program 46A.

FIG. 4 is a block diagram showing an example of the electrical configuration of the failure diagnosis device 10 according to the first embodiment.

As shown in FIG. 4, the failure diagnosis device 10 according to the present embodiment includes a CPU 11, a ROM 12, a RAM 13, an I/O 14, a storage section 15, a display section 16, an operation section 17, and a communication section 18. It is equipped with.

The CPU 11, ROM 12, RAM 13, and I/O 14 are each connected via a bus. Each functional section including a storage section 15, a display section 16, an operation section 17, and a communication section 18 is connected to the I/O 14. Each of these functional units can communicate with the CPU 11 via the I/O 14.

As the storage unit 15, for example, an HDD, SSD, flash memory, etc. are used. The storage unit 15 stores a failure diagnosis program 15A for diagnosing a failure of the CGS 40. Note that this failure diagnosis program 15A may be stored in the ROM 12.

The failure diagnosis program 15A may be installed in the failure diagnosis apparatus 10 in advance, for example. The fault diagnosis program 15A may be realized by being stored in a nonvolatile storage medium, or distributed via the network N, and installed in the fault diagnosis apparatus 10 as appropriate. Note that examples of nonvolatile storage media include CD-ROMs, magneto-optical disks, HDDs, DVD-ROMs, flash memories, and memory cards.

For example, a liquid crystal display (LCD), an organic EL (electro luminescence) display, or the like is used for the display unit 16. The display unit 16 may integrally include a touch panel. The operation unit 17 is provided with an operation input device such as a keyboard and a mouse. The display unit 16 and the operation unit 17 receive various instructions from the user of the failure diagnosis device 10. The display unit 16 displays various information such as the results of processing executed in response to instructions received from the user and notifications regarding the processing.

The communication unit 18 is connected to a network N such as the Internet, LAN, WAN, etc., and can communicate with the CGS 40 and the maintenance terminal 70 via the network N.

The CPU 11 of the failure diagnosis device 10 according to this embodiment functions as each unit shown in FIG. 5 by writing the failure diagnosis program 15A stored in the storage unit 15 into the RAM 13 and executing it.

FIG. 5 is a block diagram showing an example of the functional configuration of the failure diagnosis device 10 according to the first embodiment.

As shown in FIG. 5, the CPU 11 of the failure diagnosis device 10 according to this embodiment functions as an acquisition section 11A, an estimation section 11B, a learning section 11C, and an output section 11D.

Furthermore, as described above, the storage unit 15 stores in advance a plurality of types of first failure diagnosis models 15B and second failure diagnosis models 15C for each functional unit. In other words, a plurality of types of first failure diagnosis models 15B and second failure diagnosis models 15C correspond to one functional unit (or component). As described above, each of the first failure diagnosis model 15B and the second failure diagnosis model 15C is a mathematical model for diagnosing a failure for each functional unit of the CGS 40, which is obtained by multivariate analysis or machine learning. The first fault diagnosis model 15B and the second fault diagnosis model 15C have different model generation methods, that is, different generation logics. The first fault diagnosis model 15B and the second fault diagnosis model 15C are, for example, a combination of any two of a neural network, a random forest, and a support vector machine. The first fault diagnosis model 15B is, for example, a neural network, and the second fault diagnosis model 15C is, for example, a random forest. Note that three or more different failure diagnosis models may correspond to one functional unit (or component).

The acquisition unit 11A acquires the operation history data transmitted from the CGS 40 described above. For example, data 1 is acquired as operation history data.

The estimation unit 11B processes the operation history data acquired by the acquisition unit 11A into data that can be input to each of the first failure diagnosis model 15B and the second failure diagnosis model 15C, and uses the processed data to It is input to each of the first fault diagnosis model 15B and the second fault diagnosis model 15C. Note that the operation history data may be input as is to each of the first failure diagnosis model 15B and the second failure diagnosis model 15C without being processed. Then, the estimation unit 11B estimates the presence or absence of a failure for each functional unit of the CGS 40 using the determination results output from each of the first failure diagnosis model 15B and the second failure diagnosis model 15C. Note that each of the first failure diagnosis model 15B and the second failure diagnosis model 15C is a model that performs two-class classification that outputs failure or sound as a determination result, for example.

Specifically, when the determination results of the first fault diagnosis model 15B and the second fault diagnosis model 15C for the input data are both faulty, the estimating unit 11B determines whether the first fault diagnosis model 15B or the second fault diagnosis model 15B is faulty. The functional unit associated with the diagnostic model 15C is presumed to be at fault. Furthermore, when the determination results of the first failure diagnosis model 15B and the second failure diagnosis model 15C for the input data are both healthy, the estimation unit 11B estimates the first failure diagnosis model 15B and the second failure diagnosis model 15C. The functional unit associated with 15C is presumed to be healthy. Furthermore, when the determination result of either the first failure diagnosis model 15B or the second failure diagnosis model 15C for the input data is failure and the determination result of the other is healthy, the estimation unit 11B performs the first failure diagnosis. The functional units associated with the model 15B and the second fault diagnosis model 15C are presumed to be faulty.

The learning unit 11C generates each of the first fault diagnosis model 15B and the second fault diagnosis model 15C by performing machine learning using a group of operation history data obtained in the past as learning data. The first fault diagnosis model 15B and the second fault diagnosis model 15C generated by the learning section 11C are stored in the storage section 15 via the output section 11D. Note that the storage location of the first failure diagnosis model 15B and the second failure diagnosis model 15C is not limited to the storage unit 15. For example, it may be stored in an external storage device. Further, in this embodiment, the configuration is such that the machine learning is performed by the own device, but the configuration may be such that the machine learning is performed by an external device. Further, instead of machine learning, each of the first fault diagnosis model 15B and the second fault diagnosis model 15C may be generated using multivariate analysis.

The output unit 11D outputs and holds the final diagnosis result for each functional unit estimated by the estimation unit 11B to, for example, the storage unit 15. Further, the output unit 11D transmits the final diagnosis result held in the storage unit 15 to the maintenance terminal 70 in response to the access from the maintenance terminal 70. Note that this final diagnosis result includes the diagnosis result (failure, soundness, suspected failure) and actions to be taken for each functional unit (or each part). For example, if the determination result of data 1 is "failure", the action that the worker 60 should take is "part replacement". Further, if the determination result of data 1 is "suspected failure", the action that the worker 60 should take is "confirm component operation". Further, if the determination result of data 1 is "healthy", the action that the worker 60 should take is "no".

Next, with reference to FIG. 6, the operation of the failure diagnosis device 10 according to the first embodiment will be described.

FIG. 6 is a flowchart showing an example of the flow of processing by the failure diagnosis program 15A according to the first embodiment. Note that the functional unit in this embodiment is represented as one component.

First, the fault diagnosis program 15A stored in the storage unit 15 is activated by the CPU 11, and each step shown below is executed.

In step S111 of FIG. 6, the CPU 11 acquires operation history data (here, data 1) transmitted from the control device 41 of the CGS 40.

In step S112, the CPU 11 processes the data 1 acquired in step S111 into data that can be input to each of the first failure diagnosis model 15B and the second failure diagnosis model 15C for each functional unit (here, for each part). Processing here means deleting part of the data, extracting part of the data, adding new data, etc. to the data 1. The processing also includes processing the data format of the data 1 into a format suitable for each fault diagnosis model. Note that if the operation history data is used as is, the processing is omitted.

In step S113, the CPU 11 inputs the data 1 processed in step S112 to the first failure diagnosis model 15B.

In step S114, the CPU 11 stores the determination result output from the first failure diagnosis model 15B in the storage unit 15.

In step S115, the CPU 11 inputs the data 1 processed in step S112 to the second failure diagnosis model 15C.

In step S116, the CPU 11 stores the determination result output from the second failure diagnosis model 15C in the storage unit 15.

In step S117, the CPU 11 stores the final diagnosis result, for example, based on the determination result of the first failure diagnosis model 15B held in step S114 and the determination result of the second failure diagnosis model 15C held in step S116. 15, and the series of processing by this failure diagnosis program 15A is completed. Note that the failure diagnosis process for data 1 is repeatedly performed for a predetermined number of functional units.

Next, with reference to FIGS. 7(A) to 7(C), failure diagnosis processing using the first failure diagnosis model 15B and the second failure diagnosis model 15C according to the first embodiment will be specifically explained. do.

FIG. 7(A) is a graph showing an example of the determination result of the first failure diagnosis model 15B. FIG. 7(B) is a graph showing an example of the determination result of the second failure diagnosis model 15C. FIG. 7(C) is a graph showing an example of the determination result when the first failure diagnosis model 15B and the second failure diagnosis model 15C are used.

In FIGS. 7(A) to 7(C), black circles represent data when a certain component A actually fails, and white circles represent data when that component A does not actually fail. The vertical and horizontal axes indicate each data item (temperature, voltage, etc.), and although it is actually multidimensional, it is schematically shown as a two-dimensional graph to simplify the explanation.

As shown in FIG. 7A, the boundary line TH1 represents the threshold of the first fault diagnosis model 15B (for example, neural network) generated using arbitrary generation logic. The first fault diagnosis model 15B determines the presence or absence of a fault based on which side of the boundary line TH1 the data is plotted on. For example, region R1 indicates a region determined to have a failure, and region R2 indicates a region determined to have no failure (healthy). Looking at the plot in FIG. 7A, it is possible to roughly classify whether there is a failure or not based on the vicinity of the boundary line TH1. However, there is an ambiguous region near the boundary line TH1 where both plots coexist, and it is difficult to make an accurate diagnosis based on data alone. For example, it is possible to draw a complex boundary line TH1 by increasing the parameters when generating a machine learning model. However, drawing a complicated boundary line TH1 requires a huge amount of data, and it can be said that it is difficult to accurately diagnose ambiguous areas, especially in the CGS 40, which has a limited number of collected data.

As shown in FIG. 7(B), a second fault diagnosis model 15C (for example, random forest) is prepared using a generation logic different from that of the first fault diagnosis model 15B shown in FIG. 7(A). The boundary line TH2 represents the threshold of the second fault diagnosis model 15C. For example, region R3 indicates a region determined to have a failure, and region R4 indicates a region determined to have no failure (healthy). As described above, the second failure diagnosis model 15C cannot perform accurate diagnosis like the first failure diagnosis model 15B. However, since the method of creating the boundary line TH2 is different from that of the first fault diagnosis model 15B, different boundary lines can be created with similar accuracy.

As shown in FIG. 7(C), in order to improve diagnosis accuracy, both the boundary line TH1 of the first failure diagnosis model 15B and the boundary line TH2 of the second failure diagnosis model 15C are used to determine the failure. conduct. For example, region R5 indicates a region determined to have a failure, region R6 indicates a region determined to have no failure (healthy), and region R7 indicates a region determined to have a suspected failure. That is, the determination results are classified into the following (1) to (3).

(1) When both the first failure diagnosis model 15B and the second failure diagnosis model 15C determine that there is a failure → Part A is determined to be failure, and part A is replaced.
(2) If the judgment result of the first failure diagnosis model 15B and the judgment result of the second failure diagnosis model 15C are different, → Part A is considered to be a failure, and part A is checked on-site.
(3) When both the first failure diagnosis model 15B and the second failure diagnosis model 15C determine that there is no failure (healthy) -> Part A is determined to have no failure (healthy).

As shown in FIG. 7(A), when only the first failure diagnosis model 15B is used, eight plots are erroneously determined. Further, as shown in FIG. 7(B), when only the second failure diagnosis model 15C is used, nine plots are incorrectly determined. On the other hand, as shown in FIG. 7C, in the plurality of types of first fault diagnosis models 15B and second fault diagnosis models 15C, ambiguous areas (Fig. In example 7(C), area R7) is determined as a suspected failure, and failure diagnosis work is performed on-site for this suspected failure. Therefore, the number of false determinations is reduced to 5 plots.

Next, with reference to FIG. 8, learning processing by the failure diagnosis apparatus 10 according to the first embodiment will be specifically described.

FIG. 8 is a flowchart showing an example of the flow of learning processing by the failure diagnosis program 15A according to the first embodiment.

First, the fault diagnosis program 15A stored in the storage unit 15 is activated by the CPU 11, and each step shown below is executed.

In step S121 of FIG. 8, the CPU 11 acquires past operation history data as learning data.

In step S122, the CPU 11 processes the operation history data acquired as learning data in step S121 into data that can be input to the learning model. Note that if the operation history data is used as is, the processing is omitted.

In step S123, the CPU 11 inputs the data processed in step S122 together with correct data (for example, faulty, healthy) to a plurality of types of learning models for each functional unit.

In step S124, the CPU 11 determines whether machine learning has been completed for a predetermined number of data. If it is determined that the machine learning has ended (in the case of an affirmative determination), the process moves to step S125, and if it is determined that the machine learning has not ended (in the case of a negative determination), the process returns to step S121 and repeats the process.

In step S125, the CPU 11 stores the plurality of types of learned models generated by machine learning in the storage unit 15 as the plurality of types of first fault diagnosis model 15B and second fault diagnosis model 15C. Finish the series of processing.

Note that, in this embodiment, a case has been described in which the failure diagnosis device 10 is implemented as an external server device, but the present invention is not limited to this. The failure diagnosis device 10 may be realized as the control device 41 of the CGS 40.

As described above, according to the present embodiment, when a CGS error is detected, failure diagnosis for each functional unit of the CGS is performed using a plurality of types of failure diagnosis models. Therefore, erroneous determinations in fault diagnosis can be suppressed and fault diagnosis can be performed with high accuracy.

[Second embodiment]
In the first embodiment described above, a mode in which failure diagnosis is performed using a plurality of types of failure diagnosis models has been described. In the second embodiment, a mode will be described in which an ambiguous region is determined by selectively using any one of a plurality of types of fault diagnosis models.

FIG. 9 is a block diagram showing an example of the functional configuration of a failure diagnosis device 10A according to the second embodiment.

As shown in FIG. 9, the CPU 11 of the failure diagnosis device 10A according to the present embodiment writes the failure diagnosis program 15A stored in the storage unit 15 into the RAM 13 and executes the acquisition unit 11A, the estimation unit 11E, It functions as a learning section 11C and an output section 11D. Note that the same components as those of the failure diagnosis apparatus 10 described in the first embodiment are given the same reference numerals, and repeated explanation thereof will be omitted.

The estimation unit 11E selects any one of the plurality of types of first failure diagnosis models 15B and second failure diagnosis models 15C, and selects the selected failure diagnosis model (for example, the first failure diagnosis model 15B). Set multiple different thresholds for . Then, the estimating unit 11E estimates that the functional unit (or component) associated with the selected first failure diagnosis model 15B is faulty, healthy, or suspected to be faulty, based on the plurality of different set thresholds and input data. Here, the plurality of different threshold values include, for example, a threshold value with a precision of 90% or more and 100% or less, and a threshold with a recall rate of 90% or more and 100% or less. Precision rate is the ratio of the number of data that is actually a failure to the total number of data that is predicted to be a failure, and recall is the ratio of the number of data that is predicted to be a failure to the total number of data that is actually a failure. It is a percentage. Note that when the precision rate is 100%, it is a threshold value for not determining that the presence of a failure is healthy, and when the recall rate is 100%, it is a threshold value for not overlooking the presence of a failure.

FIG. 10 is a diagram for explaining precision and recall according to the second embodiment.

The table shown in FIG. 10 shows the correspondence between predictions and actual conditions regarding the failure diagnosis model for component A. The number inside the frame indicates the number of data. In other words, the number of data when the prediction is sound and the actual situation is sound is 28 (=A). The number of data when the prediction is failure and the actual condition is healthy is 1 (=B). The number of data when the prediction is healthy and the actual situation is a failure is 12 (=C). When the prediction is a failure and the actual condition is a failure, the number of data is 9 (=D).

In this case, the precision rate (at the time of failure determination)=D/(B+D)=9/10=90% is obtained. Further, the recall rate represents the rate of finding failures, and is calculated as recall rate=D/(C+D)=9/21≈43%. From this, the threshold value may be adjusted so that the precision rate is 90% or more and 100% or less, and the threshold value is adjusted so that the recall rate is 90% or more and 100% or less.

FIG. 11 is a graph showing an example of the determination result of the first failure diagnosis model 15B.

In FIG. 11, black circles are data when component A actually fails, and white circles are data when component A does not actually fail. The vertical and horizontal axes indicate each data item (temperature, voltage, etc.), and although it is actually multidimensional, it is schematically shown as a two-dimensional graph to simplify the explanation.

As shown in FIG. 11, a boundary line TH3 represents a threshold value at which the precision of the first fault diagnosis model 15B is, for example, 100%, and a boundary line TH4 represents a threshold value at which the recall rate of the first fault diagnosis model 15B is, for example, 100%. % represents the threshold value. For example, region R8 indicates a region determined to have a failure, region R9 indicates a region determined to have no failure (healthy), and region R10 indicates a region determined to have a suspected failure. That is, the determination results are classified into the following (1) to (3).

(1) When it is determined that there is a failure based on both the boundary line TH3 and the boundary line TH4 → Part A is determined to be failure, and part A is replaced.
(2) When the judgment result at boundary line TH3 and the judgment result at boundary line TH4 are different → Part A is suspected to be malfunctioning, and part A is inspected on-site.
(3) When it is determined that there is no failure (healthy) based on both boundary line TH3 and boundary line TH4 → Part A is determined to be no failure (healthy).

In this embodiment, the case where the failure diagnosis device 10A is implemented as an external server device has been described, but the present invention is not limited to this. The failure diagnosis device 10A may be realized as the control device 41 of the CGS 40.

As described above, according to the present embodiment, when a CGS error is detected, failure diagnosis for each functional unit of the CGS is performed using a plurality of different threshold values of one failure diagnosis model. Therefore, erroneous determinations in fault diagnosis can be suppressed and fault diagnosis can be performed with high accuracy.

[Third embodiment]
In the third embodiment, route diagnosis according to the type of error in the CGS 40 is performed on at least one of the gas route 52, the reformed water route 53, and the air route 54, and a route with a possibility of failure is detected. An embodiment in which multiple types of fault diagnosis models related to the narrowed-down routes are selectively used will be described.

FIG. 12 is a block diagram showing an example of a functional configuration of a failure diagnosis device 10B according to the third embodiment.

As shown in FIG. 12, the CPU 11 of the failure diagnosis device 10B according to the present embodiment writes the failure diagnosis program 15A stored in the storage unit 15 into the RAM 13 and executes the acquisition unit 11A, the estimation unit 11B, It functions as a learning section 11C, an output section 11D, and a route diagnosis section 11F. Note that the same components as those of the failure diagnosis apparatus 10 described in the first embodiment are given the same reference numerals, and repeated explanation thereof will be omitted.

When detecting an error in the fuel cell unit 50, the route diagnosis section 11F specifies route diagnosis according to the error code included in the operation history data from the CGS 40. Then, the route diagnosis unit 11F performs route diagnosis on at least one of the gas route 52, the reformed water route 53, and the air route 54 according to the identified error code, and identifies routes that may have a failure. Refine. Note that the gas path 52, reformed water path 53, and air path 54 are paths provided in the fuel cell unit 50 and connected to the fuel cell module 51, as described above.

The estimation unit 11B selects a plurality of types of failure diagnosis models for the functional unit (for example, for the functional unit A) that are specified by the route narrowed down by the route diagnosis unit 11F from among the plurality of failure diagnosis models stored in the storage unit 15. The first fault diagnosis model 15B and the second fault diagnosis model 15C) are selected. Then, the estimation unit 11B processes the operation history data or the operation history data acquired by the acquisition unit 11A for each of the selected plurality of types of first failure diagnosis models 15B and second failure diagnosis models 15C. Enter the data shown. Then, the estimation unit 11B estimates the presence or absence of a failure for each functional unit of the CGS 40 using the determination results output from each of the first failure diagnosis model 15B and the second failure diagnosis model 15C.

Further, the route diagnosis unit 11F may perform control to notify the result of route narrowing by route diagnosis together with an error code indicating the type of error. The destination of the error code is, for example, the maintenance terminal 70 held by the worker 60.

Further, the route diagnosis section 11F performs route diagnosis for at least one of the gas route 52, the reformed water route 53, and the air route 54 using a method according to the temperature range of the fuel cell unit 50.

Here, the storage unit 15 according to this embodiment stores a first data table 15D and a second data table 15E. These first data table 15D and second data table 15E can be referenced by the route diagnosis section 11F.

FIG. 13 is a diagram showing an example of the first data table 15D according to this embodiment.

The first data table 15D shown in FIG. 13 is a data table in which routes with a possibility of failure are specified and classified in advance for each error type (error codes A to H). Error code A is a code that targets the gas path 52, reformed water path 53, and air path 54, and is an example of the first error code. Error code B is a code that targets the gas path 52 and reformed water path 53, and is an example of a second error code. Error code C is a code that targets the gas path 52 and air path 54, and is an example of a third error code. Error code E is a code that targets the reformed water path 53 and air path 54, and is an example of the fourth error code. These error codes A, B, C, and E all target two or more routes. These error codes A, B, C, and E are associated with a route diagnosis method for narrowing down the route. Specifically, error code A is associated with route diagnosis A, error code B is associated with route diagnosis B, error code C is associated with route diagnosis C, and error code E is associated with route diagnosis. E is associated. Note that specific route diagnosis methods for these route diagnoses A, B, C, and E will be described later.

Note that since error code D targets only the gas path 52, the cause of the error is identified as the gas path 52. Similarly, error code F targets only the reformed water route 53, so the cause of the error is identified as the reformed water route 53, and error code G targets only the air route 54, so the error The cause is identified as the air path 54. Further, error code H is an error unrelated to each path of gas, reformed water, and air. In the example of FIG. 13, route diagnosis is not performed for error codes D, F, G, and H, but route diagnosis may be performed for error codes D, F, and G.

FIG. 14 is a diagram showing an example of the second data table 15E according to this embodiment.

The second data table 15E shown in FIG. 14 is a data table that defines the classification of temperature zones and the route diagnosis method for reformed water, gas, and air in each temperature zone. The temperature zones of the fuel cell unit 50 are a temperature zone [A] which is an example of the first temperature zone, a temperature zone [B] which is an example of the second temperature zone, and a temperature zone [C] which is an example of the third temperature zone. , and a temperature zone [d] which is an example of the fourth temperature zone.

The maximum temperature of the fuel cell module 51 is defined as the maximum temperature T1 [°C], and the temperature at which carbon precipitation occurs (for example, 500°C) when gas is introduced into the reformer 51B (reforming catalyst) is defined as the carbon deposition temperature T2 [°C]. , the temperature at which the fuel cell is reoxidized by introducing air into the fuel cell stack 51A (for example, 300°C) is the reoxidation temperature T3 [°C], and the temperature at which the reformed water introduced into the reformer 51B is vaporized. (For example, 100° C.) is the vaporization temperature T4 [° C.], the temperature zone [a] is higher than the carbon precipitation temperature T2 and below the maximum temperature T1. The temperature zone [b] is a temperature zone higher than the reoxidation temperature T3 and lower than the carbon precipitation temperature T2. Temperature zone [c] is a temperature zone higher than vaporization temperature T4 and lower than reoxidation temperature T3. The temperature zone [d] is a temperature zone below the vaporization temperature T4.

In the gas route diagnosis method, when the temperature of the fuel cell unit 50 is in the temperature range [A], reformed water and gas are introduced, and the temperature of the fuel cell unit 50 is in the temperature range [B] or the temperature range [C]. Alternatively, there is a method of introducing gas when the temperature is in the temperature range [d]. The reason why reformed water is introduced in temperature range [A] is to prevent carbon precipitation. Note that in the temperature zone [B] and temperature zone [C], reformed water and gas may be introduced. Note that, as the temperature of the fuel cell unit 50, at least one of the stack temperature, which is the temperature of the fuel cell stack 51A, and the reforming temperature, which is the temperature of the reformer 51B, is used. For example, the higher temperature of the stack temperature and the reforming temperature may be set as the temperature of the fuel cell unit 50, or the average temperature of the stack temperature and the reforming temperature may be set as the temperature of the fuel cell unit 50. Further, temperature information of the fuel cell unit 50 is acquired from the control device 41 of the CGS 40. The temperature information of the fuel cell unit 50 may be included in the operation history data transmitted from the control device 41 to the failure diagnosis device 10B, for example.

The method for diagnosing the reformed water path is to introduce reformed water and check the temperature of the fuel cell unit 50 when the temperature of the fuel cell unit 50 is in the temperature range [A], the temperature range [B], or the temperature range [C]. This is a method that makes diagnosis impossible when the temperature is in the temperature range [d].

In the air path diagnosis method, when the temperature of the fuel cell unit 50 is in the temperature range [A] or the temperature range [B], reformed water, gas, and air are introduced, and the temperature of the fuel cell unit 50 is in the temperature range [B]. This is a method of introducing air when the temperature is in [C] or [D]. The reason why reformed water, gas, and air are introduced in temperature zone [a] or temperature zone [b] is to prevent reoxidation of the cell. Note that in temperature zone [A], reformed water and air may be introduced. In the temperature zone [B], reformed water and air, or gas and air may be introduced. In temperature zone [C], reformed water, gas, and air, reformed water and air, or gas and air may be introduced.

Next, with reference to FIG. 15, the operation of the failure diagnosis device 10B according to the third embodiment will be described.

FIG. 15 is a flowchart showing an example of the flow of processing by the failure diagnosis program 15A according to the third embodiment.

First, the fault diagnosis program 15A stored in the storage unit 15 is activated by the CPU 11, and each step shown below is executed.

In step S131 of FIG. 15, the CPU 11 acquires operation history data (here, data 1) transmitted from the control device 41 of the CGS 40.

In step S132, the CPU 11 extracts an error code from the data 1 acquired in step S131.

In step S133, the CPU 11 determines whether the error code extracted in step S132 is error code A. If it is determined that the error code is A (in the case of an affirmative determination), the process moves to step S134, and if it is determined that the error code is not the error code A (in the case of a negative determination), the process moves to step S135.

In step S134, the CPU 11, for example, refers to the first data table 15D shown in FIG. Then, the process moves to step S141. Note that a specific method for this route diagnosis A will be described later.

On the other hand, in step S135, the CPU 11 determines whether the error code extracted in step S132 is error code B. If it is determined that the error code is B (in the case of an affirmative determination), the process proceeds to step S136, and if it is determined that the error code is not B (in the case of a negative determination), the process proceeds to step S137.

In step S136, the CPU 11 refers to the first data table 15D shown in FIG. 13 described above, performs route diagnosis B associated with error code B, and proceeds to step S141. Note that a specific method for this route diagnosis B will be described later.

On the other hand, in step S137, the CPU 11 determines whether the error code extracted in step S132 is error code C. If it is determined that the error code is C (in the case of an affirmative determination), the process proceeds to step S138, and if it is determined that the error code is not C (in the case of a negative determination), the process proceeds to step S139.

In step S138, the CPU 11 refers to, for example, the first data table 15D shown in FIG. 13, performs route diagnosis C associated with error code C, and proceeds to step S141. Note that a specific method for this route diagnosis C will be described later.

On the other hand, in step S139, the CPU 11 determines whether the error code extracted in step S132 is error code E. If it is determined that the error code is E (in the case of a positive determination), the process moves to step S140, and if it is determined that the error code is not E (in the case of a negative determination), an error other than error code A, B, C, or E is detected. The process moves to step S141 without performing route diagnosis for codes D, F, G, and H.

In step S140, the CPU 11 refers to the first data table 15D shown in FIG. 13 described above, performs route diagnosis E associated with the error code E, and proceeds to step S141. Note that a specific method for this route diagnosis E will be described later.

In step S141, the CPU 11 identifies a failure route that is the result of narrowing down by the route diagnosis in step S134, step S136, step S138, or step S140, or identifies an error code D other than error code A, B, C, or E. Identify the failure paths corresponding to F, G, and H.

In step S142, the CPU 11 selects a plurality of types of failure diagnosis models for the functional unit specified by the failure path specified in step S141 from among the plurality of failure diagnosis models stored in the storage unit 15.

In step S143, the CPU 11 performs fault diagnosis using only the plurality of types of fault diagnosis models selected in step S142, and ends the series of processing by the fault diagnosis program 15A.

Next, with reference to FIG. 16, a specific method of route diagnosis A related to step S134 in FIG. 15 will be described.

FIG. 16 is a flowchart illustrating an example of the process flow of route diagnosis A according to the third embodiment. Note that in the route diagnosis A according to the present embodiment, route diagnosis is performed in the order of the reformed water route 53, the gas route 52, and the air route 54, as an example.

In step S151 of FIG. 16, the CPU 11 refers to the second data table 15E shown in FIG. Determine whether it is [d]. Note that, as described above, at least one of the stack temperature and the reforming temperature is used as the temperature of the fuel cell unit 50. If it is determined that the temperature of the fuel cell unit 50 is in the temperature zone [A], the temperature zone [B], or the temperature zone [C] (in the case of A, B, or C), the process moves to step S152, and the temperature of the fuel cell unit 50 is If it is determined that the temperature is in the temperature range [d] (in the case of d), the process moves to step S153.

In step S152, the CPU 11 controls the CGS 40 to introduce reforming water into the reforming water path 53, and proceeds to step S154.

On the other hand, in step S153, the CPU 11 adds a route determination impossible tag to the reformed water route 53, and proceeds to step S159.

In step S154, the CPU 11 determines whether the reformed water path 53 is normal according to a predetermined criterion. If it is determined that the reformed water path 53 is normal (in the case of a positive determination), the process moves to step S155, and if it is determined that the reformed water path 53 is not normal, that is, abnormal (in the case of a negative determination) ), the process moves to step S156.

In step S155, the CPU 11 refers to the second data table 15E shown in FIG. Determine which of the following. If it is determined that the temperature of the fuel cell unit 50 is in the temperature range [A] (in the case of A), the process moves to step S158, and the temperature of the fuel cell unit 50 is in the temperature range [B] or in the temperature range [C] or in the temperature range [C]. If it is determined that it is band [D] (in the case of B, C, or D), the process moves to step S159.

On the other hand, in step S156, the CPU 11 adds a route abnormality tag to the reformed water route 53.

In step S157, the CPU 11 waits until the temperature of the fuel cell unit 50 falls below the temperature range [b], and then proceeds to step S159.

In step S158, the CPU 11 controls introducing reformed water into the reformed water path 53 and gas into the gas path 52, and proceeds to step S160.

On the other hand, in step S159, the CPU 11 performs control to introduce gas into the gas path 52, and proceeds to step S160.

In step S160, the CPU 11 determines whether the gas path 52 is normal according to a predetermined criterion. If it is determined that the gas path 52 is normal (in the case of a positive determination), the process moves to step S161, and if it is determined that the gas path 52 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S165. to move to.

In step S161, the CPU 11 determines whether the reformed water path 53 is normal according to a predetermined criterion. If it is determined that the reformed water path 53 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S162, and if it is determined that the reformed water path 53 is normal (in the case of an affirmative determination) ), the process moves to step S164.

In step S162, the CPU 11 waits until the temperature of the fuel cell unit 50 falls below the temperature range [b].

In step S163, the CPU 11 performs control to introduce gas into the gas path 52 and air into the air path 54, and proceeds to step S170.

On the other hand, in step S164, the CPU 11 refers to the second data table 15E shown in FIG. (d)]. If it is determined that the temperature of the fuel cell unit 50 is in the temperature range [A] or [B] (in the case of A or B), the process moves to step S168, and the temperature of the fuel cell unit 50 is in the temperature range [C]. Alternatively, if it is determined that the temperature is in the temperature range [d] (in the case of c or d), the process moves to step S169.

On the other hand, in step S165, the CPU 11 adds a route abnormality tag to the gas route 52.

In step S166, the CPU 11 determines whether the reformed water path 53 is normal according to a predetermined criterion. If it is determined that the reformed water path 53 is normal (in the case of a positive determination), the process moves to step S164, and if it is determined that the reformed water path 53 is not normal, that is, abnormal (in the case of a negative determination) ), the process moves to step S167.

In step S167, the CPU 11 waits until the temperature of the fuel cell unit 50 falls below the temperature range [c], and then proceeds to step S169.

In step S168, the CPU 11 controls introducing reformed water into the reformed water path 53, gas into the gas path 52, and air into the air path 54, and proceeds to step S170.

On the other hand, in step S169, the CPU 11 performs control to introduce air into the air path 54, and proceeds to step S170.

In step S170, the CPU 11 determines whether the air path 54 is normal according to a predetermined criterion. If it is determined that the air path 54 is normal (in the case of a positive determination), the process moves to step S172, and if it is determined that the air path 54 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S171. to move to.

In step S171, the CPU 11 adds a route abnormality tag to the air path 54, and proceeds to step S172.

In step S172, the CPU 11 determines the error, that is, determines the route to which the route abnormality tag is attached among the reformed water route 53, the gas route 52, and the air route 54 as a route with a possibility of failure. , the process returns to step S141 in FIG.

Next, with reference to FIG. 17, a specific method of route diagnosis B according to step S136 in FIG. 15 will be described.

FIG. 17 is a flowchart illustrating an example of the process flow of route diagnosis B according to the third embodiment. Note that in route diagnosis B according to the present embodiment, route diagnosis is performed in the order of the reformed water route 53 and the gas route 52, as an example.

In step S181 of FIG. 17, the CPU 11 refers to the second data table 15E shown in FIG. Determine whether it is [d]. If it is determined that the temperature of the fuel cell unit 50 is in the temperature zone [A], the temperature zone [B], or the temperature zone [C] (in the case of A, B, or C), the process moves to step S182, and the temperature of the fuel cell unit 50 is If it is determined that the temperature is in the temperature range [d] (in the case of d), the process moves to step S183.

In step S182, the CPU 11 controls introducing reformed water into the reformed water path 53, and proceeds to step S184.

On the other hand, in step S183, the CPU 11 adds a route determination impossible tag to the reformed water route 53, and proceeds to step S189.

In step S184, the CPU 11 determines whether the reformed water path 53 is normal according to a predetermined criterion. If it is determined that the reformed water path 53 is normal (in the case of a positive determination), the process moves to step S185, and if it is determined that the reformed water path 53 is not normal, that is, abnormal (in the case of a negative determination) ), the process moves to step S186.

In step S185, the CPU 11 refers to the second data table 15E shown in FIG. Determine which of the following. If it is determined that the temperature of the fuel cell unit 50 is in the temperature range [A] (in the case of A), the process moves to step S188, and the temperature of the fuel cell unit 50 is in the temperature range [B] or in the temperature range [C] or in the temperature range [C]. If it is determined that it is band [D] (in the case of B, C, or D), the process moves to step S189.

On the other hand, in step S186, the CPU 11 adds a route abnormality tag to the reformed water route 53.

In step S187, the CPU 11 waits until the temperature of the fuel cell unit 50 falls below the temperature range [b], and then proceeds to step S189.

In step S188, the CPU 11 controls introducing reformed water into the reformed water path 53 and gas into the gas path 52, and proceeds to step S190.

On the other hand, in step S189, the CPU 11 performs control to introduce gas into the gas path 52, and proceeds to step S190.

In step S190, the CPU 11 determines whether the gas path 52 is normal according to a predetermined criterion. If it is determined that the gas path 52 is normal (in the case of a positive determination), the process moves to step S192, and if it is determined that the gas path 52 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S191. to move to.

In step S191, the CPU 11 attaches a route abnormality tag to the gas route 52, and proceeds to step S192.

In step S192, the CPU 11 determines the error, that is, determines the route to which the route abnormality tag is attached among the reformed water route 53 and the gas route 52 as a route with a possibility of failure, and steps The process returns to S141.

Next, with reference to FIG. 18, a specific method of route diagnosis C according to step S138 in FIG. 15 will be described.

FIG. 18 is a flowchart showing an example of the process flow of route diagnosis C according to the third embodiment. Note that in the route diagnosis C according to the present embodiment, route diagnosis is performed in the order of the gas route 52 and the air route 54, as an example.

In step S201 of FIG. 18, the CPU 11 refers to the second data table 15E shown in FIG. Determine whether it is [d]. If it is determined that the temperature of the fuel cell unit 50 is in the temperature range [A] (in the case of A), the process moves to step S202, and the temperature of the fuel cell unit 50 is in the temperature range [B] or in the temperature range [C] or in the temperature range [C]. If it is determined that it is band [d] (in the case of b, c, or d), the process moves to step S203.

In step S202, the CPU 11 controls introducing reformed water into the reformed water path 53 and gas into the gas path 52, and proceeds to step S204.

On the other hand, in step S203, the CPU 11 performs control to introduce gas into the gas path 52, and proceeds to step S204.

In step S204, the CPU 11 determines whether the gas path 52 is normal according to a predetermined criterion. If it is determined that the gas path 52 is normal (in the case of a positive determination), the process moves to step S205, and if it is determined that the gas path 52 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S206. to move to.

In step S205, the CPU 11 refers to the second data table 15E shown in FIG. Determine which of the following. If it is determined that the temperature of the fuel cell unit 50 is in the temperature range [A] or [B] (in the case of A or B), the process moves to step S207, and the temperature of the fuel cell unit 50 is in the temperature range [C]. Alternatively, if it is determined that the temperature is in the temperature range [d] (in the case of c or d), the process moves to step S208.

On the other hand, in step S206, the CPU 11 attaches a route abnormality tag to the gas route 52, and proceeds to step S205.

In step S207, the CPU 11 controls introducing reformed water into the reformed water path 53, gas into the gas path 52, and air into the air path 54, and proceeds to step S209.

On the other hand, in step S208, the CPU 11 performs control to introduce air into the air path 54, and proceeds to step S209.

In step S209, the CPU 11 determines whether the air path 54 is normal according to a predetermined criterion. If it is determined that the air path 54 is normal (in the case of a positive determination), the process moves to step S211, and if it is determined that the air path 54 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S210. to move to.

In step S210, the CPU 11 adds a route abnormality tag to the air path 54, and proceeds to step S211.

In step S211, the CPU 11 determines the error, that is, determines the route to which the route abnormality tag is attached among the gas route 52 and the air route 54 as a route with a possibility of failure, and proceeds to step S141 in FIG. Return.

Next, with reference to FIG. 19, a specific method of route diagnosis E according to step S140 in FIG. 15 will be described.

FIG. 19 is a flowchart illustrating an example of the process flow of route diagnosis E according to the third embodiment. In the route diagnosis E according to the present embodiment, the route diagnosis is performed in the order of the reformed water route 53 and the air route 54, as an example.

In step S221 in FIG. 19, the CPU 11 refers to the second data table 15E shown in FIG. Determine whether it is [d]. If it is determined that the temperature of the fuel cell unit 50 is in the temperature zone [A], the temperature zone [B], or the temperature zone [C] (in the case of A, B, or C), the process moves to step S222, and the temperature of the fuel cell unit 50 is If it is determined that the temperature is in the temperature range [d] (in the case of d), the process moves to step S223.

In step S222, the CPU 11 controls introducing reformed water into the reformed water path 53, and proceeds to step S224.

On the other hand, in step S223, the CPU 11 adds a route determination impossible tag to the reformed water route 53, and proceeds to step S228.

In step S224, the CPU 11 determines whether the reformed water path 53 is normal according to a predetermined criterion. If it is determined that the reformed water path 53 is normal (in the case of a positive determination), the process moves to step S225, and if it is determined that the reformed water path 53 is not normal, that is, abnormal (in the case of a negative determination) ), the process moves to step S226.

In step S225, the CPU 11 refers to the second data table 15E shown in FIG. Determine which of the following. If it is determined that the temperature of the fuel cell unit 50 is in the temperature range [A] or [B] (in the case of A or B), the process moves to step S227, and the temperature of the fuel cell unit 50 is in the temperature range [C]. Alternatively, if it is determined that the temperature is in the temperature range [d] (in the case of c or d), the process moves to step S228.

On the other hand, in step S226, the CPU 11 adds a route abnormality tag to the reformed water route 53, and proceeds to step S228.

In step S227, the CPU 11 controls introducing reformed water into the reformed water path 53, gas into the gas path 52, and air into the air path 54, and proceeds to step S229.

On the other hand, in step S228, the CPU 11 performs control to introduce air into the air path 54, and proceeds to step S229.

In step S229, the CPU 11 determines whether the air path 54 is normal according to a predetermined criterion. If it is determined that the air path 54 is normal (in the case of a positive determination), the process moves to step S231, and if it is determined that the air path 54 is not normal, that is, abnormal (in the case of a negative determination), the process moves to step S230. to move to.

In step S230, the CPU 11 adds a route abnormality tag to the air path 54, and proceeds to step S231.

In step S231, the CPU 11 determines the error, that is, determines the route to which the route abnormality tag is attached among the reformed water route 53 and the air route 54 as a route with a possibility of failure, and steps The process returns to S141.

In this embodiment, the case where the failure diagnosis device 10B is implemented as an external server device has been described, but the present invention is not limited to this. The failure diagnosis device 10B may be realized as the control device 41 of the CGS 40.

As described above, according to the present embodiment, failure paths are narrowed down based on the error type, and failure diagnosis is performed by selectively using only the plurality of types of failure diagnosis models for functional units specified by the narrowed down failure paths. Therefore, failure diagnosis can be performed efficiently and accurately.

The above embodiments have been described by exemplifying a fault diagnosis device and a fault diagnosis system, but the embodiments may also be in the form of a program for causing a computer to execute the functions of each part included in the fault diagnosis device. Embodiments may be in the form of a computer readable storage medium storing this program.

In addition, the configurations of the failure diagnosis device and failure diagnosis system described in the above embodiments are merely examples, and may be changed according to the situation without departing from the spirit of the invention.

Furthermore, the process flow of the program described in the above embodiment is only an example, and unnecessary steps may be deleted, new steps may be added, or the processing order may be changed without departing from the main purpose. good.

Further, in the above embodiment, a case has been described in which the processing according to the embodiment is realized by a software configuration using a computer by executing a program, but the present invention is not limited to this. The embodiments may be realized by, for example, a hardware configuration or a combination of a hardware configuration and a software configuration.

10, 10A, 10B Failure diagnosis device 11, 42 CPU
11A Acquisition unit 11B, 11E Estimation unit 11C Learning unit 11D Output unit 11F Route diagnosis unit 12, 43 ROM
13, 44 RAM
14, 45 I/O
15, 46 Storage section 15A Failure diagnosis program 15B First failure diagnosis model 15C Second failure diagnosis model 15D First data table 15E Second data table 16 Display section 17 Operation section 18, 47 Communication section 30 User's home 40 CGS
41 Control device 46A Control program 48 Device I/F
49 CGS main body 50 Fuel cell unit 51 Fuel cell module 51A Fuel cell stack 51B Reformer 52 Gas path 53 Reformed water path 54 Air path 60 Worker 70 Maintenance terminal 100 Failure diagnosis system

Claims (11)

an acquisition unit that acquires operation history data regarding the operation history of the fuel cell cogeneration system corresponding to the abnormality from the fuel cell cogeneration system when the fuel cell cogeneration system detects an abnormality in operation;
It is generated by multivariate analysis or machine learning by associating a group of operation history data obtained in the past with the functional parts of the fuel cell cogeneration system, and the failure occurs for each functional part of the fuel cell cogeneration system. The operation history data acquired by the acquisition unit or the data obtained by processing the operation history data is input to each of the plurality of types of failure diagnosis models that determine the presence or absence of the failure diagnosis of the plurality of types. an estimation unit that estimates the presence or absence of a failure for each functional unit of the fuel cell cogeneration system using determination results output from each of the models;
A fault diagnosis device equipped with The plurality of types of fault diagnosis models include a first fault diagnosis model and a second fault diagnosis model whose generation method is different from that of the first fault diagnosis model,
The estimating unit is configured to calculate the first fault diagnosis model and the second fault diagnosis model when the determination results of the first fault diagnosis model and the second fault diagnosis model for the input data are both faulty. If the functional unit associated with the input data is estimated to be faulty, and the determination results of the first fault diagnosis model and the second fault diagnosis model for the input data are both healthy, the first fault diagnosis is performed. The functional unit associated with the model and the second fault diagnosis model is presumed to be healthy, and the judgment result of either the first fault diagnosis model or the second fault diagnosis model with respect to the input data is a failure. The fault diagnosis device according to claim 1, wherein when the other determination result is healthy, the functional units associated with the first fault diagnosis model and the second fault diagnosis model are presumed to be faulty. The fault diagnosis device according to claim 2, wherein the first fault diagnosis model and the second fault diagnosis model are a combination of any two of a neural network, a random forest, and a support vector machine. Instead of using the determination results output from each of the plurality of types of failure diagnosis models, the estimating unit selects any one of the plurality of types of failure diagnosis models. A plurality of different thresholds are set for the fault diagnosis model, and the functional unit associated with the selected fault diagnosis model is determined to be faulty, healthy, or suspected to be faulty based on the set plurality of different thresholds and the input data. The failure diagnosis device according to any one of claims 1 to 3, wherein the failure diagnosis device estimates. The plurality of different threshold values include a threshold at which the precision rate is 90% or more and 100% or less, which is the ratio of the number of data that is actually a failure to the total number of data predicted to be a failure, and a threshold that is a ratio of the number of data that is actually a failure to the total number of data that is predicted to be a failure, and a threshold that is a ratio of 90% to 100%. The failure diagnosis device according to claim 4, further comprising: a threshold value at which a recall rate, which is a ratio of the number of data predicted to be a failure to the number of data, is 90% or more and 100% or less. The fuel cell cogeneration system includes a fuel cell unit that is provided with a fuel cell module that generates electricity, and that is provided with a gas path, a reformed water path, and an air path that are connected to the fuel cell module,
When an abnormality in the operation of the fuel cell unit is detected, path diagnosis according to the type of abnormality is performed on at least one of the gas path, the reformed water path, and the air path, and the malfunction is detected. It is further equipped with a route diagnosis section that narrows down possible routes.
The estimation unit selectively uses only the plurality of types of failure diagnosis models for the functional units specified by the route narrowed down by the route diagnosis unit to determine the presence or absence of a failure for each functional unit of the fuel cell cogeneration system. The failure diagnosis device according to any one of claims 1 to 3, wherein the failure diagnosis device estimates. The operation history data includes an error code representing the type of abnormality,
The error code includes a first error code that targets the gas route, the reformed water route, and the air route, a second error code that targets the gas route and the reformed water route, and a second error code that targets the gas route and the reformed water route. The failure diagnosis device according to claim 6, wherein the error code is one of a third error code targeted at the air path and a fourth error code targeted at the reformed water path and the air path. The route diagnosis section performs the route diagnosis for at least one of the gas route, the reformed water route, and the air route using a method according to a temperature range of the fuel cell unit. The failure diagnosis device according to item 7. The fuel cell module includes a fuel cell stack in which a plurality of fuel cells having an electrolyte layer, a fuel electrode, and an air electrode are stacked, and a reformer for reforming a gas containing a hydrocarbon raw material,
The temperature zone of the fuel cell unit is a first temperature zone that is higher than the carbon deposition temperature, which is the temperature at which carbon deposition occurs when gas is introduced into the reformer, and is lower than or equal to the maximum temperature of the fuel cell module, and the fuel cell stack. a second temperature zone higher than the reoxidation temperature, which is the temperature at which reoxidation of the fuel cell occurs by introducing air into the fuel cell, and lower than the carbon deposition temperature; and a second temperature zone at which the reformed water introduced into the reformer vaporizes. a third temperature zone higher than a certain vaporization temperature and below the reoxidation temperature, and a fourth temperature zone below the vaporization temperature,
The route diagnosis method for the gas route includes introducing reformed water and gas when the temperature of the fuel cell unit is in the first temperature range, and introducing reformed water and gas when the temperature of the fuel cell unit is in the second temperature range or A method of introducing gas when the temperature is in the third temperature zone or the fourth temperature zone,
The route diagnosis method for the reformed water route includes introducing reformed water when the temperature of the fuel cell unit is in the first temperature zone, the second temperature zone, or the third temperature zone; A method of disabling diagnosis when the temperature of the fuel cell unit is in the fourth temperature range,
The route diagnosis method for the air route includes introducing at least reformed water and air when the temperature of the fuel cell unit is in the first temperature range, and introducing the reformed water and air when the temperature of the fuel cell unit is in the second temperature range. In this case, at least one of reformed water and gas and air are introduced, and the air is introduced when the temperature of the fuel cell unit is in the third temperature zone or the fourth temperature zone. The failure diagnosis device according to claim 8. fuel cell cogeneration system,
a fault diagnosis device connected to the fuel cell cogeneration system;
Equipped with
The fuel cell cogeneration system includes:
When an abnormality in the operation of the own system is detected, transmitting operation history data regarding the operation history of the own system corresponding to the abnormality to the failure diagnosis device,
The failure diagnosis device includes:
an acquisition unit that acquires the operation history data transmitted from the fuel cell cogeneration system;
It is generated by multivariate analysis or machine learning by associating a group of operation history data obtained in the past with the functional parts of the fuel cell cogeneration system, and the failure occurs for each functional part of the fuel cell cogeneration system. The operation history data acquired by the acquisition unit or the data obtained by processing the operation history data is input to each of the plurality of types of failure diagnosis models that determine the presence or absence of the failure diagnosis of the plurality of types. an estimation unit that estimates the presence or absence of a failure for each functional unit of the fuel cell cogeneration system using determination results output from each of the models;
A fault diagnosis system equipped with A fault diagnosis program for causing a computer to function as each part included in the fault diagnosis apparatus according to any one of claims 1 to 9.