Hydrogen electrochemical system, method for determining state of hydrogen cell stack, and program

Abstract

<sb / >In the present invention, a hydrogen electrochemical system 100 comprises an acquisition unit 541, a storage unit 53, and a determination unit 544. The acquisition unit 541 acquires at least distribution information Ds2, De2, and Dt2 indicating the distribution of a physical quantity Q in a hydrogen cell stack 1. In the hydrogen cell stack 1, water is generated by an electrochemical reaction in which hydrogen is used, or vice versa. The storage unit 53 stores correlation data Dc. The correlation data Dc indicates a correlation between the state of the hydrogen cell stack 1 and a feature amount φ that is converted from at least the information Ds2 and De2 regarding the distribution in the hydrogen cell stack 1 in a state including a normal state and one or more abnormal states. On the basis of the correlation information Dc, the determination unit 544 determines the state of the hydrogen cell stack 1 from at least the distribution information Dt2 of the hydrogen cell stack 1 in operation.

Description

Hydrogen electrochemical system, hydrogen cell stack state determination method, and program

[0001] The present invention relates to a hydrogen electrochemical system, a hydrogen cell stack state determination method, and a program.

[0002] Conventionally, as disclosed in Patent Document 1, a technique is known in which the alternating current impedance spectroscopy (EIS) is used to monitor the state of a fuel cell during power generation and detect abnormal states such as dryout.

[0003] JP 2023-154982 A

[0004] However, the EIS method takes time to output the detection result of the condition, making it difficult to detect an abnormal condition in real time.

[0005] In view of the above circumstances, an object of the present invention is to detect the state of a hydrogen cell stack in real time.

[0006] In order to achieve the above object, a hydrogen electrochemical system according to one aspect of the present invention comprises an acquisition unit, a storage unit, and a determination unit. The acquisition unit acquires at least distribution information indicating the distribution of physical quantities in a hydrogen cell stack. The hydrogen cell stack generates one of water and hydrogen through an electrochemical reaction using the other. The storage unit stores correlation information. The correlation information indicates a correlation between a feature quantity converted from at least the distribution information in the hydrogen cell stack in a normal state and a state including one or more abnormal states and the state of the hydrogen cell stack. The determination unit determines the state of the hydrogen cell stack during operation from at least the distribution information of the hydrogen cell stack based on the correlation information.

[0007] To achieve the above object, a hydrogen cell stack state determination method according to one aspect of the present invention includes an acquisition step and a determination step. The hydrogen cell stack generates water or hydrogen through an electrochemical reaction using the other. In the acquisition step, at least distribution information is acquired. The distribution information indicates a distribution of physical quantities in the hydrogen cell stack. In the determination step, the state of the hydrogen cell stack is determined from at least the distribution information of the operating hydrogen cell stack based on correlation information. The correlation information indicates a correlation between the state of the hydrogen cell stack and at least a feature converted from the distribution information in the hydrogen cell stack in a normal state and a state including one or more abnormal states.

[0008] To achieve the above object, one aspect of the present invention provides a program that causes a computer to execute a hydrogen cell stack state determination method. The program causes the computer to function as a means for executing an acquisition step and a determination step. In the acquisition step, at least distribution information is acquired. The distribution information indicates a distribution of physical quantities in the hydrogen cell stack. The hydrogen cell stack generates water or hydrogen through an electrochemical reaction using one of the two. In the determination step, the state of the hydrogen cell stack is determined from at least the distribution information of the operating hydrogen cell stack based on correlation information. The correlation information indicates a correlation between the state of the hydrogen cell stack and at least a feature converted from the distribution information in the hydrogen cell stack in a state including a normal state and one or more abnormal states.

[0009] Further features and advantages of the present invention will become more apparent from the following embodiments.

[0010] According to the present invention, the state of the hydrogen cell stack can be detected in real time.

[0011] Schematic diagram showing an example of the configuration of a hydrogen electrochemical system according to the first embodiment; Perspective view showing an example of the configuration of a hydrogen cell according to the first embodiment; Exploded perspective view of a hydrogen cell according to the first embodiment; Conceptual diagram showing an example of conversion from physical quantities to feature quantities; Flowchart for explaining an example of a method for determining the state of a hydrogen cell stack according to the first embodiment; Flowchart for explaining an example of a method for determining a conversion model by creating correlation information according to the first embodiment; Schematic diagram showing an example of conversion of feature quantities using distribution information according to the first embodiment; Graph showing an example of classifying each feature quantity into a normal state and L types of abnormal states using a virtual space according to the first embodiment; Flowchart for explaining an example of a method for determining the state of a hydrogen cell stack during operation based on correlation information according to the first embodiment; Schematic diagram showing an example of determining the state of a hydrogen cell stack during operation based on correlation information according to the first embodiment; Flowchart for explaining an example of a method for determining a conversion model by creating correlation information according to a first modified example of the first embodiment 1. Chart: A schematic diagram showing an example of converting feature quantities using distribution information in a first modified example of the first embodiment. A graph showing an example of classifying each feature quantity into a normal state and L types of abnormal states using a virtual space in a first modified example of the first embodiment. A flowchart for explaining an example of a method for determining the state of a hydrogen cell stack during operation based on correlation information in a first modified example of the first embodiment. A schematic diagram showing an example of determining the state of a hydrogen cell stack during operation based on correlation information in a first modified example of the first embodiment. A flowchart for explaining an example of a method for determining the state of a hydrogen cell stack during operation in a second modified example of the first embodiment. A schematic diagram showing an example of machine learning of a classifier using distribution information in a second modified example of the first embodiment. A schematic diagram showing an example of determining the state of a hydrogen cell stack during operation using a classifier in a second modified example of the first embodiment. A schematic diagram showing an example of the configuration of a hydrogen electrochemical system according to a second embodiment. A perspective view showing an example of the configuration of a hydrogen cell in a second embodiment. An exploded perspective view of a hydrogen cell in a second embodiment.

[0012] 100... Hydrogen electrochemical system, 101... Hydrogen cell stack unit, 1011... Fuel cell, 1012... Hydrogen generation device, 102... Control unit, 103... Material supply unit, 104... Supply fluid control unit, 105... Discharge fluid control unit, 106... Discharge unit, 1... Hydrogen cell stack, 10, 10a... Hydrogen cell, 11... Electrolyte membrane, 12, 12a... Anode electrode, 121, 121a... Anode catalyst layer, 122, 122a... Anode gas diffusion layer, 13, 13a... Cathode electrode, 131, 131a... Cathode catalyst layer, 132... Cathode gas diffusion layer, 132a... Cathode water transport layer, 14... Separator, 141... Anode side flow path, 142... Cathode side flow path, 2... End plate, 21, 23... Supply port, 211, 231... Supply channel, 22, 24...discharge port, 221, 241...discharge path, 3...sensor, 31...main sensor, 32...sub-sensor, 321...voltage sensor, 51...input unit, 52...alarm unit, 53...storage unit, 54...state determination unit, 541...acquisition unit, 542...conversion unit, 543...correlation information generation unit, 544...determination unit, 545...learning unit, Di...input information, Ds...simulation data, Ds1...operation information, Ds2...distribution information, Ds3...detection information, De...experimental data, De1...operation information, De2...distribution information, De3...detection information, Dt...monitoring data, Dt1...operation information, Dt2...distribution information, Dt3...detection information, Dc...correlation information, Q...physical quantity, Qm...main physical quantity, Qs...sub-physical quantity, Ic...classifier, S, Sa, Sb...stack direction

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0014] 1 is a schematic diagram showing an example of the configuration of a hydrogen electrochemical system 100 according to a first embodiment. The hydrogen electrochemical system 100 includes a hydrogen cell stack unit 101, a control unit 102, a material supply unit 103, a supply fluid control unit 104, a discharge fluid control unit 105, and a discharge unit 106.

[0015] The hydrogen cell stack unit 101 generates either water or hydrogen through an electrochemical reaction using the other. The hydrogen cell stack unit 101 according to this embodiment is a polymer electrolyte fuel cell (PEFC) 1011 that generates electricity by generating water through an electrochemical reaction using hydrogen (hydrogen gas) and outputs the electricity to the outside. The control unit 102 controls each component of the hydrogen electrochemical system 100 (e.g., the hydrogen cell stack unit 101, the material supply unit 103, the supply fluid control unit 104, the discharge fluid control unit 105, and the discharge unit 106). The control unit 102 also functions as a status determination device that determines the status of the hydrogen cell stack unit 101 in real time. The material supply unit 103 is connected to the hydrogen cell stack unit 101 via the supply fluid control unit 104 and supplies a material fluid to the hydrogen cell stack unit 101. The supplied fluid may be, for example, air (or oxygen gas), hydrogen gas, or water vapor. The supply fluid control unit 104 controls the pressure, flow rate, etc. of the fluids supplied from the material supply unit 103 to the hydrogen cell stack unit 101. The discharge fluid control unit 105 controls the pressure, flow rate, etc. of the fluids discharged from the hydrogen cell stack unit 101 to the discharge unit 106. The discharged fluids include, for example, liquid water, excess air (or oxygen gas), excess hydrogen gas, and water vapor. The discharge unit 106 is connected to the hydrogen cell stack unit 101 via the discharge fluid control unit 105 and processes the fluids discharged from the hydrogen cell stack unit 101. The discharge unit 106 may be a mechanism that discharges the fluid to the outside by, for example, being open to the atmosphere, or may be a mechanism that traps the fluid and stores it in a storage tank, etc.

[0016] In the hydrogen cell stack unit 101, as will be described later, hydrogen gas is supplied to the anode electrode 12, and air (or oxygen gas) is supplied to the cathode electrode 13. In a polymer electrolyte fuel cell such as that of this embodiment, a cation exchange membrane is used for the electrolyte membrane 11, and hydrogen ions move from the anode electrode 12 to the cathode electrode 13. The polymer electrolyte fuel cell generates water through an electrochemical reaction at the cathode electrode 13, and also generates electricity and outputs it to the outside.

[0017] However, without being limited to this example, the hydrogen cell stack unit 101 may be a fuel cell other than a polymer electrolyte fuel cell. For example, the hydrogen cell stack unit 101 may be a solid oxide fuel cell (SOFC), an alkaline fuel cell (AFC), a phosphoric acid fuel cell (PAFC), or a molten carbonate fuel cell (MCFC). In a solid oxide fuel cell, an oxygen ion conductive oxide such as YSZ (yttria stabilized zirconia) is used for the electrolyte membrane, and oxygen ions move from the cathode electrode to the anode electrode. In a solid oxide fuel cell, water is produced by an electrochemical reaction at the anode electrode, and electricity is generated and output to the outside. In an alkaline fuel cell, an aqueous potassium hydroxide (KOH) solution is used for the electrolyte layer, and hydroxide ions (OH － ) moves from the cathode to the anode. In alkaline fuel cells, water is produced by an electrochemical reaction at the anode, and electricity is generated and output to the outside. In phosphoric acid fuel cells, phosphoric acid (H ３ P.O. ４ A phosphoric acid fuel cell generates water through an electrochemical reaction at the cathode, and generates electricity to be output to the outside. In a molten carbonate fuel cell, the electrolyte layer is made of a porous ceramic (LiAlO ２The hydrogen cell stack unit 101 uses a molten carbonate held in a polymer electrolyte (such as a polymer electrolyte membrane) to transfer carbonate ions from the cathode to the anode. Molten carbonate is a liquid-phase electrolyte made of, for example, sodium (Na) carbonate and potassium (K) carbonate. Molten carbonate fuel cells produce water through an electrochemical reaction at the anode. The control unit 102 can determine the status of the hydrogen cell stack unit 101 in real time, even if the hydrogen cell stack unit 101 is a fuel cell other than the above-mentioned polymer electrolyte fuel cell.

[0018] <1-1. Hydrogen cell stack unit 101> The hydrogen cell stack unit 101 (fuel cell 1011) has a hydrogen cell stack 1, a set of end plates 2, and a sensor 3. The hydrogen cell stack 1 is a stack of multiple hydrogen cells 10 stacked in one direction S.

[0019] <1-1-1. Hydrogen cell 10> Fig. 2 is a perspective view showing an example of the configuration of the hydrogen cell 10. Fig. 3 is an exploded perspective view of the hydrogen cell 10. Note that Fig. 2 corresponds to the hydrogen cell 10 arranged in part II surrounded by the dashed line in Fig. 1. In this specification, the stacking direction of the multiple hydrogen cells 10 from one side of the paper, either the upper or lower side, to the other side in Figs. 2 and 3 is referred to as the "stacking direction S." Within the stacking direction S, the direction from the lower side to the upper side of the paper in Figs. 2 and 3 is referred to as the "stacking direction Sa," and the direction from the upper side to the lower side of the paper in Figs. 2 and 3 is referred to as the "stacking direction Sb."

[0020] 2 and 3 , the hydrogen cell 10 includes an electrolyte membrane 11, an anode electrode 12, a cathode electrode 13, and a separator 14. The anode electrode 12 includes an anode catalyst layer 121 and an anode gas diffusion layer 122. The cathode electrode 13 includes a cathode catalyst layer 131 and a cathode gas diffusion layer 132. In the stacking direction Sa, these are stacked in the following order: separator 14, anode gas diffusion layer 122, anode catalyst layer 121, electrolyte membrane 11, cathode catalyst layer 131, cathode gas diffusion layer 132, and separator 14.

[0021] <1-1-1-1. Electrolyte membrane 11> The electrolyte membrane 11 is an ion exchange membrane formed using a solid polymer material such as a fluorine-based resin material or a hydrocarbon-based resin material, and extends in a direction intersecting (for example, perpendicular to) the stacking direction Sa. In this embodiment, the electrolyte membrane 11 is a proton exchange membrane formed from Nafion (registered trademark), and exhibits good proton conductivity in a wet state.

[0022] <1-1-1-2. Anode Electrode 12> The anode 12 generates hydrogen ions and electrons e through a catalytic oxidation reaction of hydrogen gas and supplies the hydrogen ions to the electrolyte membrane 11. The hydrogen ions pass through the electrolyte membrane 11 and move from the anode 12 to the cathode 13. The electrons e generated at the same time move to the anode 12 of the adjacent hydrogen cell 10 adjacent in the stacking direction Sa or to the end plate 2 on the stacking direction Sa side. This generates a direct current flowing in the stacking direction Sb within the hydrogen cell stack unit 101.

[0023] The anode catalyst layer 121 of the anode electrode 12 is a porous layer in which a catalyst is supported on a porous body. The catalyst includes, for example, at least one of a precious metal such as Pt (platinum), Ru (ruthenium), or Ir (iridium), or an oxide of Ru or Ir. The porous body is made of a material such as carbon and extends in a direction intersecting (for example, perpendicular to) the stacking direction Sb. In this embodiment, the anode catalyst layer 121 is a porous layer in which a Pt catalyst is supported on a porous body made of carbon, and is disposed and stacked on the stacking direction Sb side of the electrolyte membrane 11.

[0024] The anode catalyst layer 121 preferably further contains an ionomer, more preferably an ionomer containing an organic fluorine compound-based electrolyte. The inclusion of an ionomer in the anode catalyst layer 121 makes it easier to arrange the anode catalyst layer 121 on the electrolyte membrane 11. The ionomer absorbs the moisture in the hydrogen gas, improving the conductivity of protons (i.e., hydrogen ions) in the anode catalyst layer 121. The remaining hydrogen gas refers to hydrogen gas that is not used in the electrochemical reaction. However, this example does not exclude a configuration in which the anode catalyst layer 121 does not contain an ionomer.

[0025] The anode gas diffusion layer 122 is a porous layer that is gas permeable and conductive. The anode gas diffusion layer 122 extends in a direction intersecting (for example, perpendicular to) the stacking direction Sb and is disposed on the stacking direction Sb side of the anode catalyst layer 121. The anode gas diffusion layer 122 is a porous body made of, for example, entangled metal fibers or a sintered body of metal particles. The metal is, for example, titanium (Ti).

[0026] <1-1-1-3. Cathode electrode 13> The cathode electrode 13 generates water through a reaction between hydrogen ions moving through the electrolyte membrane 11, electrons e moving from the stack direction Sb side, and oxygen (or oxygen gas) in the air. The water is discharged to the outside as water vapor together with the remaining air (or oxygen gas).

[0027] The cathode catalyst layer 131 of the cathode electrode 13 is a porous layer in which a catalyst is supported on a porous body. The catalyst includes, for example, at least one of platinum (Pt), platinum-coated titanium, platinum-supported carbon, palladium (Pd)-supported carbon, cobalt glyoxime, nickel glyoxime, and the like. The porous body is made of carbon or the like and extends in a direction intersecting (for example, perpendicular to) the stacking direction Sa. In this embodiment, the cathode catalyst layer 131 is a porous layer in which a platinum catalyst is supported on a porous body made of carbon, and is disposed and stacked on the stacking direction Sa side of the electrolyte membrane 11.

[0028] The cathode catalyst layer 131 preferably further contains an ionomer, more preferably an ionomer containing an organic fluorine compound-based electrolyte. The cathode catalyst layer 131 containing an ionomer facilitates placement of the cathode catalyst layer 131 on the electrolyte membrane 11. The ionomer absorbs liquid-phase water generated on the cathode electrode 13 side, improving the conductivity of protons (i.e., hydrogen ions) in the cathode catalyst layer 131. The remaining oxygen gas refers to oxygen gas that is not used in the electrochemical reaction. However, this example does not exclude a configuration in which the cathode catalyst layer 131 does not contain an ionomer.

[0029] The cathode gas diffusion layer 132 is a porous layer having gas permeability and electrical conductivity. The cathode gas diffusion layer 132 extends in a direction intersecting (for example, perpendicular to) the stacking direction Sa, and is disposed on the stacking direction Sa side of the cathode catalyst layer 131. The cathode gas diffusion layer 132 is, for example, a porous body made of entangled carbon fibers or carbon.

[0030] <1-1-1-4. Separator 14> The separator 14 extends in a direction intersecting (e.g., perpendicular to) the stacking direction S, and is arranged and stacked on the stacking direction Sa side and stacking direction Sb side of the hydrogen cell 10. The separator 14 is a gas-impermeable member formed using a conductive material. Examples of materials that can be used for the separator 14 include metal materials such as stainless steel and aluminum (Al), carbon materials, and carbon-containing composite resin materials.

[0031] An anode-side flow channel 141 recessed in the stacking direction Sb is disposed on the surface of the separator 14 that contacts the anode electrode 12. Hydrogen gas (and water seeping out from the electrolyte membrane 11) flows through the anode-side flow channel 141. The hydrogen gas diffuses from the anode-side flow channel 141 into the anode electrode 12. In this embodiment, there are multiple anode-side flow channels 141, which extend in one direction intersecting the stacking direction S and are arranged in parallel in a direction perpendicular to the one direction. However, the shape of the anode-side flow channel 141 is not limited to the above example.

[0032] Furthermore, a cathode-side flow channel 142 recessed in the stacking direction Sa is disposed on the surface of the separator 14 that contacts the cathode electrode 13. Air (or oxygen gas) and water produced in the hydrogen cell 10 flow through the cathode-side flow channel 142. The water is discharged from the cathode electrode 13 to the cathode-side flow channel 142 together with the air flow (or residual oxygen gas). In this embodiment, there are multiple cathode-side flow channels 142, which extend in one direction intersecting the stacking direction S and are arranged in parallel in a direction perpendicular to that direction. However, the shape of the cathode-side flow channel 142 is not limited to the example described above.

[0033] Preferably, the inner surfaces of the anode-side flow path 141 and the cathode-side flow path 142 are treated to be water-repellent. This prevents water vapor from liquefying and remaining in the anode-side flow path 141 and the cathode-side flow path 142. Therefore, the hydrogen electrochemical system 100 can prevent flooding from occurring in the hydrogen cell stack unit 101 (fuel cell 1011).

[0034] <1-1-2. End Plates 2> The end plates 2 are each plate-shaped and extend in a direction intersecting (for example, perpendicular to) the stacking direction S, and are arranged in contact with both ends of the hydrogen cell stack 1 in the stacking direction S. Each end plate 2 is made of a conductive material, such as a metal such as stainless steel or aluminum (Al), or an alloy thereof. A current terminal is arranged on each end plate 2. Electric power generated in the hydrogen cell stack unit 101 (fuel cell 1011) is output to the outside from the current terminal.

[0035] Further, a supply port 21 and a discharge port 22 are arranged on the end plate 2 on the stacking direction Sb side.

[0036] The hydrogen gas supply channel 211 is connected to the supply port 21 and is connected to the anode-side flow channel 141 of each separator 14 via the supply port 21 and through-holes (not shown) arranged in the hydrogen cell stack 1. As a result, hydrogen gas supplied from the supply channel 211 is supplied to the anode-side flow channel 141 of each hydrogen cell 10 and diffuses from the anode-side flow channel 141 into the anode electrode 12.

[0037] The discharge channel 221 is also connected to the discharge port 22, and is connected to the anode-side flow path 141 of each separator 14 via the discharge port 22 and other through-holes (not shown) arranged in the hydrogen cell stack 1. As a result, water leaking from the electrolyte membrane 11 in each hydrogen cell 10 and residual hydrogen gas are discharged from the anode electrode 12 to the anode-side flow path 141, and are also discharged from the discharge channel 221 via the discharge port 22 to the outside.

[0038] A supply port 23 and a discharge port 24 are arranged on the end plate 2 on the stacking direction Sa side.

[0039] The air (or oxygen gas) supply channel 231 is connected to the supply port 23 and is connected to the cathode-side flow channel 142 of each separator 14 via the supply port 23 and through-holes (not shown) arranged in the hydrogen cell stack 1. As a result, the air (or oxygen gas) supplied from the supply channel 231 is supplied to the cathode-side flow channel 142 of each hydrogen cell 10 and diffuses from the cathode-side flow channel 142 into the cathode electrode 13.

[0040] The discharge channel 241 is also connected to the discharge port 24, and is connected to the cathode-side flow path 142 of each separator 14 via the discharge port 24 and other through-holes (not shown) arranged in the hydrogen cell stack 1. As a result, water produced in each hydrogen cell 10 and the remaining air (or oxygen gas) are discharged from the cathode electrode 13 to the cathode-side flow path 142, and are also discharged from the discharge channel 241 via the discharge port 24 to the outside.

[0041] <1-1-3. Sensor 3> The sensor 3 detects a physical quantity Q of the hydrogen cell stack unit 101 (fuel cell 1011). In this embodiment, the physical quantity Q includes a primary physical quantity Qm and a secondary physical quantity Qs. The primary physical quantity Qm is one or more types of physical quantities used to determine the state of the hydrogen cell stack unit 101, and is obtained, for example, by multi-point measurement, measurement in a plurality of different modes, or calculation. Each primary physical quantity Qm is the same type of physical quantity at each location distributed k-dimensionally in the hydrogen cell stack unit 101 (particularly the hydrogen cell stack 1). Note that k is 1, 2, or 3, and in this embodiment, k = 3. The secondary physical quantity Qs is one or more types of physical quantities that are used auxiliary when determining the state of the hydrogen cell stack unit 101. In this embodiment, each secondary physical quantity Qs is a physical quantity at one or more locations in the hydrogen cell stack unit 101. However, this example does not exclude a configuration in which at least one type of secondary physical quantity Qs is the same type of physical quantity at each location distributed k-dimensionally in the hydrogen cell stack unit 101 (particularly the hydrogen cell stack 1). Furthermore, the above example does not exclude a configuration in which the physical quantity Q includes the primary physical quantity Qm but does not include the secondary physical quantity Qs.

[0042] The sensor 3 includes a plurality of main sensors 31 and a sub-sensor 32 .

[0043] Each main sensor 31 is disposed on the side of and / or inside the hydrogen cell stack 1 (for example, at least one of the anode electrode 12 and the cathode electrode 13). Each main sensor 31 detects the main physical quantity Qm at a k-dimensional distribution position in the hydrogen cell stack 1 in real time, and outputs the detected main physical quantity Qm to the control unit 102. For example, in this embodiment, the multiple main sensors 31 detect the main physical quantity Qm that is distributed three-dimensionally in real time.

[0044] In the following description, unless otherwise specified, the main sensors 31 are described as being three-dimensionally distributed in the hydrogen cell stack 1. The term "three-dimensional distribution" refers to a distribution in a real space (three-dimensional) consisting of the stacking direction S and two directions perpendicular to the stacking direction S and orthogonal to each other. For example, in FIG. 3 , in each hydrogen cell 10, multiple main sensors 31 are arranged on the outer surface of the anode electrode 12 (particularly the anode gas diffusion layer 122), and multiple main sensors 31 are arranged two-dimensionally with respect to the end face of the anode electrode 12 (particularly the anode gas diffusion layer 122) on the stacking direction Sb side. Furthermore, multiple main sensors 31 are arranged on the outer surface of the cathode electrode 13 (particularly the cathode gas diffusion layer 132), and multiple main sensors 31 are arranged two-dimensionally with respect to the end face of the cathode electrode 13 (particularly the cathode gas diffusion layer 132) on the stacking direction Sa side. The distribution of the main sensors 31 is set so as not to affect the performance of the hydrogen cell stack 1. The same applies to the secondary sensor 32 disposed in the hydrogen cell stack 1 .

[0045] The main physical quantity Qm detected by the main sensor 31 includes at least one of a magnetic field, an electric field, temperature, humidity, etc. For example, in this embodiment, the main sensor 31 is a magnetic sensor element such as a Hall element, and detects the three-dimensional distribution of the magnetic field in the hydrogen cell stack 1.

[0046] The secondary sensor 32 is disposed, for example, in the hydrogen cell stack 1, the supply paths 211, 231, and the exhaust paths 221, 241, and detects the secondary physical quantity Qs of the hydrogen cell stack unit 101 in real time and outputs it to the control unit 102. For example, the secondary physical quantity Qs detected by the secondary sensor 32 includes at least one of the temperature, humidity, pressure, and flow rate of the gas supplied to and exhausted from the supply paths 211, 231 and the exhaust paths 221, 241.

[0047] In this embodiment, the secondary sensor 32 also includes a voltage sensor 321. The voltage sensor 321 detects in real time the voltage generated in the hydrogen cell stack unit 101 (fuel cell 1011) (in other words, the potential difference between a pair of end plates 2: see FIG. 1 , etc.), and outputs the voltage to the control unit 102. In other words, the secondary physical quantity Qs further includes the voltage detected by the voltage sensor 321 (in other words, the voltage generated in the hydrogen cell stack unit 101).

[0048] <1-2. Control Unit 102> Next, the configuration of the control unit 102 will be described. As shown in Fig. 1, the control unit 102 has an input unit 51, a notification unit 52, a storage unit 53, and a state determination unit 54, and is equipped with a computer including one or more arithmetic devices such as a processor or a microprocessor. The control unit 102 controls each component of the hydrogen electrochemical system 100 using the programs and information (data) stored in the storage unit 53.

[0049] The input unit 51 receives operation inputs from the user and outputs them to the storage unit 53 and the state determination unit 54 .

[0050] The notification unit 52 has a display, a speaker, etc., and notifies the outside by displaying on a screen, outputting sound, etc.

[0051] The storage unit 53 is a non-transitory storage medium that maintains its storage even when the power supply is stopped. The storage unit 53 stores, for example, input information Di, simulation data Ds, experimental data De, monitoring data Dt, and correlation information Dc. The storage unit 53 also stores various programs and data that can be read and executed by components of the hydrogen electrochemical system 100 (for example, a computer installed in the control unit 102). For example, the storage unit 53 has a program that causes the computer to function as a means for executing a state determination method for the hydrogen cell stack 1, which will be described later.

[0052] The input information Di is stored information based on an operation input received by the input unit 51 .

[0053] The simulation data Ds is the result of a computer simulation of the primary physical quantity Qm and the secondary physical quantity Qs exhibited by the hydrogen cell stack unit 101 in various states including a normal state and one or more abnormal states (i.e., a single abnormal state or multiple different abnormal states). The simulation data Ds may be the result of a simulation performed by the control unit 102 based on detection results acquired from the sensor 3 by an acquisition unit 541 (described later), or may be simulation data acquired by the acquisition unit 541 from an external source.

[0054] The simulation data Ds includes motion information Ds1, distribution information Ds2, and detection information Ds3.

[0055] The operation information Ds1 indicates the operating environment settings and conditions of the hydrogen cell stack unit 101 in a computer simulation, and the operation results. The operation results indicate whether the hydrogen cell stack unit 101 was in a normal state (operating normally), and whether an abnormal state occurred. There are L types of abnormal states, where L is an integer greater than or equal to 1. The L types of abnormal states include, for example, flooding, dryout, deterioration of the electrolyte membrane 11, catalyst poisoning, catalyst deterioration, and a lack of electrochemical reactants. In addition to these, the abnormal states may further include unknown abnormalities (for example, abnormalities of unknown cause).

[0056] Flooding is a phenomenon in which substances (for example, water) produced by an electrochemical reaction accumulate within the hydrogen cell stack unit 101 (particularly the hydrogen cell stack 1).

[0057] Dryout is a phenomenon in which the electrolyte membrane 11, the anode electrode 12 (particularly the ionomer contained therein), or the cathode electrode 13 (particularly the ionomer contained therein) dries out. Dryout reduces the conductivity of ions (such as hydrogen ions) within the electrolyte membrane 11 (and each of the electrodes 12 and 13).

[0058] Deterioration of the electrolyte membrane 11 is a phenomenon in which the electrolyte membrane 11 chemically or mechanically deteriorates. The deterioration of the electrolyte membrane 11 reversibly reduces the conductivity of ions (such as hydrogen ions) within the electrolyte membrane 11.

[0059] Catalyst poisoning is a phenomenon in which impurities (arsenic or sulfur and their compounds, carbon monoxide, etc.) contained in electrochemical reactants (e.g., hydrogen gas, air (or oxygen gas)) supplied to the hydrogen cell stack 1 reversibly or irreversibly reduce the activity of the catalysts contained in the anode catalyst layer 121 and the cathode catalyst layer 131.

[0060] Catalyst degradation is a phenomenon in which the activity of the catalyst decreases or becomes inactive due to changes in the catalyst properties caused by chemical reactions in the anode catalyst layer 121 and the cathode catalyst layer 131, damage or peeling of the catalyst carrier, and abnormal temperature rise in the anode catalyst layer 121 and the cathode catalyst layer 131.

[0061] The shortage of electrochemical reactants is a phenomenon in which there is a shortage of electrochemically reactive substances in the hydrogen cell stack 1, and is caused by an insufficient supply of reactants. The reactants are, for example, hydrogen gas or air (or oxygen gas).

[0062] The distribution information Ds2 indicates the detection amount of the main sensor 31 at each distribution position. In detail, the distribution information Ds2 indicates, for example, the main physical quantities Qm (magnetic field, electric field, temperature, humidity, etc.) of the hydrogen cell stack 1 and the k-dimensional distribution (e.g., three-dimensional distribution) of each main physical quantity Qm within the hydrogen cell stack 1.

[0063] The detection information Ds3 indicates the amount detected by the secondary sensor 32. The detection information Ds3 indicates, for example, the secondary physical quantity Qs in the hydrogen cell stack 1, such as the temperature, humidity, pressure, and flow rate of the gas supplied to and exhausted from the supply paths 211, 231 and the exhaust paths 221, 241, and the voltage value (or the potential difference between a pair of end plates) of the hydrogen cell stack 1. Note that if the secondary physical quantity Qs includes a physical quantity that is distributed three-dimensionally, the detection information Ds3 may further indicate the three-dimensional distribution of the physical quantity.

[0064] Next, the experimental data De indicates real-time experimental results of the primary physical quantity Qm and secondary physical quantity Qs detected from the hydrogen cell stack unit 101 in various operating states, including a normal state and one or more abnormal states (i.e., a single abnormal state or multiple different abnormal states). The experimental data De may be experimental results acquired by the acquisition unit 541 from the sensor 3 in the hydrogen electrochemical system 100, or experimental result data acquired by the acquisition unit 541 from an external source. The experimental data De includes operation information De1, distribution information De2, detection information De3, etc. The operation information De1 indicates the operating environment settings and conditions of the hydrogen cell stack unit 101 and the operation results when the experimental results are acquired. The distribution information De2 indicates real-time detection results of the primary physical quantity Qm output from the primary sensor 31 of the hydrogen cell stack unit 101 and the three-dimensional distribution of each primary physical quantity Qm when the experimental results are acquired. The detection information De3 indicates the real-time detection result of the secondary physical quantity Qs output from the secondary sensor 32 of the hydrogen cell stack unit 101 when the experimental results are obtained. Note that if the secondary physical quantity Qs includes a physical quantity that is distributed three-dimensionally, the detection information De3 may further indicate the three-dimensional distribution of the physical quantity.

[0065] The monitoring data Dt indicates real-time detection results of the primary physical quantity Qm and secondary physical quantity Qs detected from the object whose operating state is to be determined (i.e., the hydrogen cell stack unit 101 in operation). The monitoring data Dt is real-time detection results acquired by the acquisition unit 541 from the sensor 3 in the hydrogen electrochemical system 100. The hydrogen electrochemical system 100 determines the state of the hydrogen cell stack 1 from the monitoring data Dt, which is real-time detection results of each physical quantity Q (such as the primary physical quantity Qm and secondary physical quantity Qs) in the hydrogen cell stack 1 in operation. The monitoring data Dt includes operation information Dt1, distribution information Dt2, detection information Dt3, etc. The operation information Dt1 indicates the operating environment settings and conditions of the hydrogen cell stack unit 101 in operation and the operation results. The distribution information Dt2 indicates real-time detection results of the primary physical quantity Qm output from the main sensor 31 of the hydrogen cell stack unit 101 in operation and the three-dimensional distribution of each primary physical quantity Qm. The detection information Dt3 indicates the real-time detection result of the secondary physical quantity Qs output from the secondary sensor 32 of the operating hydrogen cell stack unit 101. Note that if the secondary physical quantity Qs includes a physical quantity that is distributed three-dimensionally, the detection information Dt3 may further indicate the three-dimensional distribution of each of the physical quantities.

[0066] The correlation information Dc indicates the correlation between each of the feature quantities φ(0,1,1) to φ(L,N,M) converted from each of the main physical quantities Qm and the operating state of the hydrogen cell stack 1 (see FIG. 7 , which will be described later). Hereinafter, the feature quantities (0,1,1) to φ(L,N,M) may be referred to as "feature quantities φ."

[0067] As mentioned above, "F" is the number of states of the hydrogen cell stack 1. Specifically, the F states include one normal state and L different abnormal states. In other words, F = (1 + f). "F" is an integer equal to or greater than 2. "L" is an integer equal to or greater than 1. Furthermore, "f" is an ordinal number for individually identifying one normal state and L different abnormal states, and is expressed as an integer equal to or greater than 0 and equal to or less than F.

[0068] N is the total number of main physical quantities Qm of the same type obtained from the hydrogen cell stack 1 in the same state, and is an integer equal to or greater than 2. Specifically, N is the sum of the individual numbers n0, n1, ..., nL of main physical quantities Qm of the same type obtained from the hydrogen cell stack 1 in each state. That is, N = Σnf = n0 + n1 + ... + nL. n0 is the number of main physical quantities Qm of the same type obtained from the hydrogen cell stack 1 in the normal state. Similarly, n2 to nL are the numbers of main physical quantities Qm of the same type obtained from the hydrogen cell stack 1 in the first to Lth abnormal states, respectively. Note that in this specification, n0, n1, ..., nL are sometimes collectively referred to as "nf," and are represented by an integer equal to or greater than 1 and equal to or less than N.

[0069] Furthermore, in this embodiment, the numbers nf of different types of main physical quantities Qm obtained from the same hydrogen cell stack 1 are all equal. For example, the number n0 of each type of main physical quantity Qm in a hydrogen cell stack 1 in a normal state is all equal. Similarly, the numbers n1 to nL of each type of main physical quantity Qm in a hydrogen cell stack 1 in the first to Lth abnormal states are all equal.

[0070] However, the above example is not limiting, and does not exclude a configuration in which the number nf of some types of main physical quantities Qm is different from the number nf of other types of main physical quantities Qm in at least one state of the hydrogen cell stack 1. Furthermore, the above example does not exclude a configuration in which the number nf of all types of main physical quantities Qm is different in at least one state of the hydrogen cell stack 1.

[0071] "M" is the total number of types of the main physical quantity Qm, and is expressed as an integer equal to or greater than 1. "m" is an ordinal number for identifying the type of the main physical quantity Qm, and is expressed as an integer equal to or greater than 1 and equal to or less than M.

[0072] For example, the feature quantity φ(0, n, m) is a feature quantity φ obtained by converting the nth physical quantity of the mth type of main physical quantity Qm obtained from the hydrogen cell stack 1 in a normal state. Furthermore, the feature quantity φ(1, n, m) is a feature quantity φ obtained by converting the nth physical quantity of the mth type of main physical quantity Qm obtained from the hydrogen cell stack 1 in a first abnormal state. Furthermore, the feature quantity φ(f, n, m) is a feature quantity φ obtained by converting the nth physical quantity of the mth type of main physical quantity Qm obtained from the hydrogen cell stack 1 in the fth abnormal state.

[0073] The state determination unit 54 determines the state of the operating hydrogen cell stack 1 from at least the distribution information Dt2 of the monitoring data Dt of the operating hydrogen cell stack 1 based on the correlation information Dc. For example, the state determination unit 54 reads corresponding programs, data, etc. from the storage unit 53 and performs state determination of the hydrogen cell stack 1. However, this example does not exclude a configuration in which the above-mentioned state determination of the hydrogen cell stack 1 is realized by hardware.

[0074] The state determination unit 54 includes an acquisition unit 541 , a conversion unit 542 , a correlation information generation unit 543 , and a determination unit 544 .

[0075] The acquisition unit 541 acquires the physical quantity Q and, for example, acquires at least distribution information Ds2, De2, and Dt2 indicating the distribution of the main physical quantity Qm in the hydrogen cell stack 1, which generates one of water and hydrogen through an electrochemical reaction using the other. In this embodiment, the acquisition unit 541 acquires operation information Ds1, De1, and Dt1, distribution information Ds2, De2, and Dt2, and detection information Ds3, De3, and Dt3 of the hydrogen cell stack 1. For example, the acquisition unit 541 acquires the physical quantity Q of the hydrogen cell stack unit 101 in various states, including a normal state and one or more different abnormal states. Note that this information may be data acquired by the hydrogen electrochemical system 100 or data input from an external source.

[0076] The converter 542 converts each of the primary physical quantities Qm into a feature quantity φ. FIG. 4 is a conceptual diagram showing an example of conversion from the physical quantity Q to the feature quantity φ. In this embodiment, the converter 542 generates distribution map data P(0,1,1) to P(L,N,M) from the primary physical quantity Qm acquired by the acquirer 541. Note that at this time, the secondary physical quantity Qs may also be used, as indicated by the dashed line in FIG. 4. The distribution map data P(0,1,1) to P(L,N,M) are image data, and may hereinafter be referred to as "distribution map data P" or "distribution map data P(f,n,m)."

[0077] For example, the converter 542 generates distribution map data P(f, n, m) from the primary physical quantity Qm (and the secondary physical quantity Qs) by a specific calculation. The specific calculation and the characteristic elements of the distribution map data P(f, n, m) are not particularly limited, and various methods can be used depending on the purpose. For example, the specific calculation can be performed using an existing method such as a regression method using at least the distribution information Ds2, De2, and Dt2.

[0078] The distribution map data P(f, n, m) is an element of the image set Im. The distribution map data P constituting the set P(m) is generated from the m-th type of main physical quantity Qm acquired from the hydrogen cell stack 1 in the f-th state. In other words, the distribution map data P(f, n, m) can be expressed by the following equation 1: P(f, n, m) ∈ Im (Equation 1)

[0079] Furthermore, the conversion unit 542 converts the distribution map data P into an α-dimensional feature quantity φ. Specifically, the conversion unit 542 can convert the distribution map data P into the feature quantity φ by performing operations such as extracting feature elements of the distribution map data P using existing technology. In this case, the extracted feature elements can include, for example, the type, intensity, lightness, and saturation of the color of the pixel, the color density distribution between adjacent pixels or within a predetermined range of pixel sets, and the rate of change of at least one of the color intensity, lightness, and saturation.

[0080] The conversion unit 542 extracts α types of feature elements from the distribution map data P, and converts them into numerically represented points on an α-dimensional coordinate axis using a mapping E_α consisting of these elements. In this embodiment, "α" is the number of feature elements to be extracted, and is an integer equal to or greater than 1. In detail, the number of feature elements α is expressed as the product of the number of feature element items β and the number of types γ for each item. For example, if the number of types of feature elements for each item is γ1, γ2, ..., γβ, the number of feature elements α can be expressed by the following formula 2: α = (γ1 + γ2 + ... + γβ) (Formula 2)

[0081] The above-mentioned point quantified on the α-dimensional coordinate axis by the mapping E_α is the feature value φ expressed in coordinates of the virtual space (feature space) Sv defined in α dimensions. For example, the feature value φ(f,n,m) converted from the distribution map data P(f,n,m) is the mapped point E_α(P(f,n,m)). The mapped points correspond to the coordinates of each of the feature values ​​φ in the virtual space Sv in FIGS. 8 and 13 described below. In other words, the mapped point E_α(P(f,n,m)) is the feature value φ(f,n,m) expressed as coordinates in the α-dimensional virtual space Sv, and is an element of the set Jα of α types of feature elements. The feature value φ(f,n,m) and the mapped point E_α(P(f,n,m)) can be expressed by the following formula 3: φ(f,n,m)=E_α(P(f,n,m))∈Jα (Formula 3)

[0082] In this embodiment, the conversion unit 542 extracts and digitizes α=3 feature elements from the distribution map data P, and defines the points mapped onto three-dimensional coordinates with these as the coordinate axes as the feature quantities φ.

[0083] Furthermore, the conversion unit 542 can estimate another physical quantity Q from a specific physical quantity Q. For example, in this embodiment, the conversion unit 542 estimates the current density distribution of the hydrogen cell stack 1 by inverse problem analysis from the distribution of the magnetic field detected by each main sensor 31 (e.g., a Hall element). Furthermore, the conversion unit 542 can convert magnetic field distribution information Ds2, De2, Dt2 into current density distribution information Ds2, De2, Dt2.

[0084] At least one of the conversion from the primary physical quantity Qm (and the secondary physical quantity Qs) to the feature quantity φ and the estimation and conversion from a specific physical quantity Q to another physical quantity Q may be performed outside the hydrogen electrochemical system 100. In this case, the converted feature quantity φ and / or the converted physical quantity Q are acquired by the acquisition unit 541. Furthermore, the examples in this specification do not exclude a configuration in which the hydrogen electrochemical system 100 does not include the conversion unit 542.

[0085] The correlation information generator 543 constructs a correlation model between each of the feature quantities φ(0,1,1) to φ(L,N,M) and F (={1+L}) types of operating states of the hydrogen cell stack 1, and generates correlation information Dc based on the correlation model. The correlation information Dc is stored in the storage unit 53. In this embodiment, the feature quantity φ is represented using coordinates (see FIGS. 8 and 13, etc.). However, without being limited to this example, the feature quantity φ may be represented using techniques such as mathematical transformation, encoding, missing value processing, clustering, principal component analysis, interaction, feature selection, or an appropriate combination of these techniques.

[0086] The determination unit 544 determines the state of the hydrogen cell stack 1 from the monitoring data Dt of the operating hydrogen cell stack 1. For example, the determination unit 544 determines the state of the hydrogen cell stack 1 from at least the distribution information Dt2 of the monitoring data Dt of the operating hydrogen cell stack 1 based on the correlation information Dc.

[0087] <1-2-1. Determining the State of the Hydrogen Cell Stack 1> Next, an example of determining the state of the hydrogen cell stack 1 in the hydrogen electrochemical system 100 will be described with reference to Fig. 5. Fig. 5 is a flowchart illustrating a method for determining the state of the hydrogen cell stack 1 in the first embodiment.

[0088] 5 , the hydrogen electrochemical system 100 first creates correlation information Dc using the simulation data Ds and the experimental data De to determine a conversion model from the primary physical quantity Qm (and the secondary physical quantity Qs) to the feature quantity φ (step S1). Thereafter, for example, when the input unit 51 receives an operation input indicating the start of state determination of the operating hydrogen cell stack 1 (Yes in step S2), the hydrogen electrochemical system 100 determines the state of the hydrogen cell stack 1 from the monitoring data Dt based on the correlation information Dc (step S3).

[0089] The state determination in step S3 is then repeated, for example, until the input unit 51 receives an operational input indicating the end of state determination of the operating hydrogen cell stack 1. In other words, if the operational input indicating the end of the state determination is not received (No in step S4), the process returns to step S3. This allows the hydrogen electrochemical system 100 to determine the state of the operating hydrogen cell stack 1 in real time. On the other hand, if the operational input indicating the end of the state determination is received (Yes in step S4), the real-time state determination process of the hydrogen cell stack 1 ends.

[0090] <1-2-2. Example of Determining a Conversion Model> An example of determining a conversion model by creating correlation information Dc will be described with reference to FIGS. 6 to 8. FIG. 6 is a flowchart illustrating an example of a method for determining a conversion model by creating correlation information Dc in the first embodiment. FIG. 7 is a schematic diagram showing an example of converting feature quantities φ(0,1,1) to φ(L,N,M) using distribution information Ds2 and De2 in the first embodiment. FIG. 8 is a graph showing an example of classifying each feature quantity φ(0,1,1) to φ(L,N,M) into a normal state and L types of abnormal states using a virtual space Sv in the first embodiment. Note that FIG. 7 corresponds to step S1 in FIG. 5.

[0091] In Figures 6 to 8, the current density distribution estimated from the magnetic field is converted into feature quantities φ(0,1,1) to φ(L,N,1) having three-dimensional coordinates, and the correlation information Dc is created based on these. That is, M = 1 and α = 3. In Figures 6 to 8, for ease of explanation, the numbers n0 to nL of experimental and simulation results for the normal state and the first to Lth abnormal states are each set to six. However, this example is not limited to this, and the numbers n0 to nL of experimental and simulation results for the normal state and the first to Lth abnormal states may be a number other than six, or may not be the same number. Furthermore, the first to Lth abnormal states each represent a state in which one of the following conditions occurs: flooding, dryout, deterioration of the electrolyte membrane 11, catalyst poisoning, catalyst deterioration, or a deficiency (or an unknown abnormality) of an electrochemical reactant. Since the case where α = 3 is illustrated here, the virtual space Sv in Figure 8 is expressed three-dimensionally, consisting of the X, Y, and Z directions. The X-axis, Y-axis, and Z-axis each represent one of three types of regression-based characteristic elements converted from the distribution of current density estimated by inverse problem analysis from the magnetic field detection results of the monitoring data Dt.

[0092] First, as shown in FIG. 6 , the converter 542 converts the N magnetic field distributions included in the distribution information Ds2 and De2 into N current density distributions through inverse problem analysis (step S101). Then, as shown in FIG. 7 , the converter 542 creates N current density distribution map data P(0,1,1) to P(L,N,1) based on the N current density distribution information (step S102). Furthermore, the converter 542 performs image analysis on the n current density distribution map data P(0,1,1) to P(L,N,1) and converts each piece of current density distribution information into N feature quantities φ each having an α (=3)-dimensional coordinate (step S103). For example, the multiple pixels constituting each distribution map data can be divided into pixel ranges of a predetermined width using a finite element method or the like. Therefore, for the α (=3) coordinate axes of the feature φ, for example, any of the mean value, variance, skewness, moment, gradient, etc. of brightness, lightness, hue, or saturation in each pixel range is adopted.

[0093] For example, in Figure 7, the features in the normal state are φ(0,1,1), ..., φ(0,n0,1). Note that the α1 coordinate values ​​of φ(0,1,1), φ(0,2,1), ..., φ(0,n0,1) are the first features (e.g., the above-mentioned mean values) converted from the distribution map data P(0,1,1), P(0,2,1), ..., P(0,n0,1), respectively. The α2 coordinate values ​​of φ(0,1,1), φ(0,2,1), ..., φ(0,n0,1) are the second features (e.g., the above-mentioned variances) converted from the distribution map data P(0,1,1), P(0,2,1), ..., P(0,n0,1), respectively. The α3 coordinate values ​​of φ(0,1,1), φ(0,2,1), ..., φ(0,n0,1) are the third feature values ​​(for example, the skewness mentioned above) converted from the distribution map data P(0,1,1), P(0,2,1), ..., P(0,n0,1), respectively.

[0094] 7, the feature quantities in the first abnormal state are φ(1,n0+1,1), ..., φ(1,n0+n1,1). Note that the α1 coordinate values ​​of φ(1,n0+1,1), ..., φ(1,n0+n1,1) are the first feature quantities (e.g., the above-mentioned average value) converted from the distribution map data P(1,n0+1,1), ..., P(1,n0+n1,1), respectively. The α2 coordinate values ​​of φ(1,n0+1,1), ..., φ(0,n0+n1,1) are the second feature quantities (e.g., the above-mentioned variance) converted from the distribution map data P(1,n0+1,1), ..., P(1,n0+n1,1), respectively. The α3 coordinate values ​​of φ(0,n0+1,1), ..., φ(0,n0+n1,1) are the third feature values ​​(for example, the skewness mentioned above) converted from the distribution map data P(1,n0+1,1), ..., P(1,n0+n1,1), respectively.

[0095] 7, the feature quantities in the Lth abnormal state are φ(L,N-nL+1,1), ..., φ(L,N,1). The α1 coordinate values ​​of φ(L,N-nL+1,1), ..., φ(L,n,1) are first feature quantities (e.g., the above-mentioned average value) converted from the distribution map data P(L,N-nL+1,1), ..., P(L,N,1), respectively. The α2 coordinate values ​​of φ(L,6N-nL+1,1), ..., φ(L,N,1) are second feature quantities (e.g., the above-mentioned variance) converted from the distribution map data P(L,N-nL+1,1), ..., P(L,N,1). The α3 coordinate value of φ(L, N-nL+1, 1), ..., φ(L, N, 1) is the third feature (for example, the above-mentioned skewness) converted from the distribution map data P(L, N-nL+1, 1), ..., P(L, N, 1).

[0096] Next, the correlation information generator 543 plots each of the feature quantities φ(0,1,1) to φ(L,N,1) in an α (=3)-dimensional virtual space Sv (step S104), as shown in FIG. 8 . The correlation information generator 543 then determines coordinate ranges C0 to CL corresponding to each state of the hydrogen cell stack 1 based on the feature quantities φ(0,1,1) to φ(0,N,1) for each state (step S105). For example, the correlation information generator 543 determines the coordinate range C0 corresponding to the normal state based on the feature quantities φ(0,1,1) to φ(0,n0,1) for the normal state. Furthermore, the correlation information generator 543 determines coordinate ranges C1 to CL corresponding to the first to Lth abnormal states, respectively, based on the feature amounts φ(1,n0+1,1) to φ(1,n0+n1,1) in the first abnormal state, ..., and the feature amounts φ(L,N-nL+1,1) to φ(L,N,1) in the Lth abnormal state. Note that, in order to make the graph easier to see, the coordinate ranges C2 to C(L-1) corresponding to the second to (L-1)th abnormal states are not shown in FIG. 8.

[0097] The means for determining the coordinate ranges C0 to CL is not particularly limited. For example, the coordinate ranges C0 to CL of each state (class) can be set using a decision boundary obtained by solving a multi-class classification problem in machine learning such as a support vector machine (SVM).

[0098] Alternatively, the coordinate range C0 to CL can be set using a statistical method or the like. For example, the correlation information generation unit 543 may calculate the peak-top value Va, minimum value Vb, and maximum value Vc of the distribution for each coordinate value of the feature amount φ, and set the lower and upper limits of each coordinate value in the coordinate range C0 to CL based on these values. For example, the lower limit of each coordinate value may be (Va-A x |Va-Vb|) obtained by subtracting A (≧1) times the absolute value |Va-Vb| of the difference between the peak-top value Va and the minimum value Vb. Note that if (Va-A x |Va-Vb|) < 0, the lower limit is set to 0. Furthermore, the upper limit of each coordinate value may be (Va + B x |Va-Vc|) obtained by adding B (≧1) times the absolute value |Va-Vc| of the difference between the peak-top value Va and the maximum value Vc.

[0099] The correlation information generation unit 543 generates correlation information Dc, which is a correspondence between each of the determined coordinate ranges C0 to CL and the F types of states of the hydrogen cell stack 1 (i.e., normal state and the first to Lth abnormal states), and stores the correlation information Dc in the memory unit 53 (step S106).

[0100] 1-2-3. Example of determining the state of the hydrogen cell stack 1 during operation Next, an example of determining the state of the hydrogen cell stack 1 during operation will be described with reference to Figs. 8 to 10. Fig. 9 is a flowchart illustrating an example of a method for determining the state of the hydrogen cell stack 1 during operation based on the correlation information Dc in the first embodiment. Fig. 10 is a schematic diagram showing an example of determining the state of the hydrogen cell stack 1 during operation based on the correlation information Dc in the first embodiment. Note that Fig. 9 corresponds to the processing in step S3 in Fig. 5.

[0101] As shown in Fig. 9 , the conversion unit 542 first converts magnetic field distribution information included in distribution information Dt2 of the monitoring data Dt detected from the operating hydrogen cell stack 1 into current density distribution information by inverse problem analysis (step S301). The conversion unit 542 then creates current density distribution map data Pr based on the current density distribution information (step S302), as shown in Fig. 10 . Furthermore, as shown in Fig. 10 , the conversion unit 542 performs image analysis of the current density distribution map data Pr to convert the current density distribution information into a feature value φr having α (= 3)-dimensional coordinates (step S303).

[0102] Next, the determination unit 544 references the correlation information Dc and determines which of the coordinate ranges C0 to CL in the virtual space Sv shown in FIG. 8 contains the feature value φr (step S304). For example, if the feature value φr is contained within the coordinate range C0 corresponding to the normal state, such as when φr = φr1 in FIG. 8 , the determination unit 544 determines that the operating hydrogen cell stack 1 is operating normally. Alternatively, if the feature value φr is contained within the coordinate range C1 corresponding to the first abnormal state, such as when φr = φr2 in FIG. 8 , the determination unit 544 determines that the operating hydrogen cell stack 1 is in the first abnormal state. Alternatively, if the feature value φr is contained within the coordinate range CL corresponding to the Lth abnormal state, such as when φr = φr3 in FIG. 8 , the determination unit 544 determines that the operating hydrogen cell stack 1 is in the Lth abnormal state. Note that if the feature value φr is not contained within all of the coordinate ranges C0 to CL, the determination unit 544 determines that the operating hydrogen cell stack 1 is in an unknown abnormal state, for example. The determination process of FIG. 9 then ends.

[0103] <1-2-4. Notes> In the above example, the distribution information Ds2, De2 of one type of main physical quantity Qm (current density distribution) is used to create the correlation information Dc and to determine the state of the operating hydrogen cell stack 1. However, this is not limiting, and these may be performed using distribution information Ds2, De2 of a main physical quantity Q other than the current density distribution, or may be performed using distribution information Ds2, De2 of multiple types of main physical quantities Qm.

[0104] <1-3. First Modification of First Embodiment> In the example of the first embodiment described above, α feature quantities φ (such as mean value, variance, and skewness) are used for each of the coordinate axes α1, α2, and α3 of the α-dimensional virtual space Sv. However, the example is not limited to this example, and M primary physical quantities Qm may be used for each coordinate value. Below, configurations of the first modification that differ from the first embodiment described above will be described. Furthermore, the same components as those in the first embodiment described above will be assigned the same reference numerals, and descriptions of similar configurations may be omitted.

[0105] <1-3-1. Example of Determining a Conversion Model> An example of determining a conversion model by creating correlation information Dc will be described with reference to FIGS. 11 to 13. FIG. 11 is a flowchart illustrating an example of a method for determining a conversion model by creating correlation information Dc in the first modified example of the first embodiment. FIG. 12 is a schematic diagram illustrating an example of converting feature quantities φ(0,1,1) to φ(L,N,M) using distribution information Ds2 and De2 in the first modified example of the first embodiment. FIG. 13 is a graph illustrating an example of classifying each feature quantity φ(0,1,1) to φ(L,N,M) into a normal state and L types of abnormal states using a virtual space Sv in the first modified example of the first embodiment. Note that FIG. 11 corresponds to the processing in step S1 of FIG. 5.

[0106] To facilitate understanding, in FIGS. 12 and 13 , distribution information on M types of primary physical quantities Qm is converted into feature quantities φ(0,1,1) to φ(L,N,M) having three-dimensional coordinates φ1, φ2, and φ3, and correlation information Dc is created based on these. That is, M=3 and α=3. In addition, in FIGS. 11 to 13 , for ease of explanation, the number of experimental / simulation results in the normal state and the first to Lth abnormal states is set to six. However, this is not limited to this example, and the number of experimental / simulation results in the normal state and the first to Lth abnormal states may be a number other than six, or may not be the same number. Furthermore, the first to Lth abnormal states each indicate a state in which one of flooding, dryout, deterioration of the electrolyte membrane 11, catalyst poisoning, catalyst deterioration, or deficiency of an electrochemical reactant (and an unknown abnormality) has occurred. Furthermore, since the example illustrates the case where α = 3, the virtual space Sv in Fig. 13 is expressed three-dimensionally consisting of three directions, X, Y, and Z. The X-axis, Y-axis, and Z-axis are feature quantities φ obtained by converting the main physical quantity Qm included in the distribution information Ds2, De2 and the detection information Ds3, De3, respectively. For example, the X-axis is a feature quantity φ obtained by converting the current density estimated by inverse problem analysis from the magnetic field detection results of the monitoring data Dt. The Y-axis is a feature quantity φ obtained by converting the temperature detection results of the monitoring data Dt. The Z-axis is a feature quantity φ obtained by converting the humidity detection results of the monitoring data Dt.

[0107] 11 , the converter 542 generates (N×3) pieces of distribution map data P(0,1,1) to P(L,N,3) based on N pieces of distribution information Ds2, De2 for M (=3) types of primary physical quantities Qm (step Sa101). For example, for each of the N experimental / simulation results, the converter 542 generates N pieces of distribution map data P(0,1,1) to P(L,N,1) for current density, N pieces of distribution map data P(0,1,2) to P(L,N,2) for temperature, and n pieces of distribution map data P(0,1,3) to P(L,N,3) for humidity, all of which are converted from the distribution information of the magnetic field by inverse problem analysis.

[0108] Next, the converter 542 performs image analysis on the N pieces of distribution map data P(0, 1, 1) to P(L, N, 3) for each of the main physical quantities Qm, and converts the distribution information Ds2, De2 of the M (= 3) types of main physical quantities Qm into (N × 3) feature quantities φ(0, 1, 1) to φ(L, N, 3) having an α (= 3)-dimensional coordinate (step Sa102).

[0109] Next, the correlation information generator 543 plots each of the feature quantities φ(0,1,1) to φ(L,N,M) in an α (=3)-dimensional virtual space Sv (step Sa103), as shown in FIG. 13. The correlation information generator 543 then determines coordinate ranges C0 to CL corresponding to each state of the hydrogen cell stack 1 based on the feature quantities φ(0,1,1) to φ(0,N,M) for each state (step S104). For example, the correlation information generator 543 determines the coordinate range C0 corresponding to the normal state based on the feature quantities φ(0,1,1) to φ(0,n0,3) for the normal state. Furthermore, the correlation information generator 543 determines coordinate ranges C1 to CL corresponding to the first to Lth abnormal states, respectively, based on the feature amounts φ(1,n0+1,1) to φ(1,N0+n1,3) in the first abnormal state, ..., and the feature amounts φ(L,N-nL+1,1) to φ(L,N,3) in the Lth abnormal state. Note that, to make the graph easier to see, FIG. 13 omits the illustration of coordinate ranges C2 to C(L-1) corresponding to the second to (L-1)th abnormal states. Note that the coordinate ranges C1 to CL are determined, for example, in the same manner as in the first embodiment.

[0110] The correlation information generation unit 543 generates correlation information Dc, which represents the correspondence between each of the determined coordinate ranges C0 to CL and the state of the hydrogen cell stack 1 (i.e., the normal state and the first to Lth abnormal states), and stores the correlation information Dc in the memory unit 53 (step Sa105).

[0111] 1-2-2. Example of determining the state of the hydrogen cell stack 1 during operation Next, an example of determining the state of the hydrogen cell stack 1 during operation will be described with reference to Figs. 13 to 15. Fig. 14 is a flowchart illustrating an example of a method for determining the state of the hydrogen cell stack 1 during operation based on the correlation information Dc. Fig. 15 is a schematic diagram showing an example of determining the state of the hydrogen cell stack 1 during operation based on the correlation information Dc in a first modified example of the first embodiment. Note that Fig. 14 corresponds to the processing in step S3 of Fig. 5.

[0112] 15 , the conversion unit 542 generates distribution map data Pra(1) to Pra(3) of each of the main physical quantities Qm (current density, temperature, and humidity estimated from the magnetic field) detected from the operating hydrogen cell stack 1 based on the distribution information Dt2 (step Sa301). Furthermore, as shown in FIG. 15 , the conversion unit 542 performs image analysis on each of the distribution map data Pra(1) to Pra(3) and converts the M (= 3) types of main physical quantities Qm included in the monitoring data Dt into coordinate quantities φra(1) to φra(3) of the feature quantity φra (step Sa302).

[0113] Next, the determination unit 544 references the correlation information Dc and determines which of the coordinate ranges C0 to CL in the virtual space Sv shown in FIG. 13 contains the feature quantity φr (step Sa303). For example, if the feature quantity φr is contained within the coordinate range C0 corresponding to the normal state, as in φra1 in FIG. 13, the determination unit 544 determines that the operating hydrogen cell stack 1 is operating normally. Alternatively, if the feature quantity φr is contained within the coordinate range C1 corresponding to the first abnormal state, as in φra2 in FIG. 13, the determination unit 544 determines that the operating hydrogen cell stack 1 is in the first abnormal state. Alternatively, if the feature quantity φr is contained within the coordinate range CL corresponding to the Lth abnormal state, as in φra3 in FIG. 13, the determination unit 544 determines that the operating hydrogen cell stack 1 is in the Lth abnormal state. Note that if the feature quantity φr is not contained within all of the coordinate ranges C0 to CL, the determination unit 544 determines that the operating hydrogen cell stack 1 is in an unknown abnormal state, for example. The determination process of FIG. 14 then ends.

[0114] <1-3-3. Remarks> In the example of the first embodiment, the feature elements α1, α2, and α3 of the distribution map data P generated from the primary physical quantity Qm are used for the coordinate axes of the α-dimensional virtual space Sv, and in the example of the first modified example of the first embodiment, the feature quantities φ1, φ2, and φ3 converted from the primary physical quantity Qm are used. However, the coordinate values ​​of the α-dimensional virtual space Sv are not limited to these examples. For example, the feature quantity φ of the M types of primary physical quantities Qm may be used for the coordinate values ​​of the α-dimensional virtual space Sv.

[0115] <1-4. Second Modification of First Embodiment> In the examples of the first embodiment and its first modification described above, the state of the hydrogen cell stack 1 during operation is determined using an α-dimensional virtual space Sv. However, this example is not limiting, and the state of the hydrogen cell stack 1 during operation may also be determined using a machine-learned classifier Ic. In other words, the correlation information Dc may be learning data obtained by machine learning of the classifier Ic. Below, configurations of the second modification that differ from those of the first embodiment and its first modification described above will be described. Furthermore, the same reference numerals will be used to designate components that are the same as those of the first embodiment and its first modification described above, and descriptions of similar configurations may be omitted.

[0116] In a second modified example of the first embodiment, the state determination unit 54 includes a learning unit 545 (see FIG. 1 ). The learning unit 545 performs machine learning of a classifier Ic using the distribution information Ds2, De2 and / or the detection information Ds3, De3. The classifier Ic is a learning algorithm stored in the storage unit 53, and estimates the state of the operating hydrogen cell stack 1 based on the detection results of the sensor 3. The classifier Ic may be, for example, a linear classifier such as a support vector machine (SVM), a classifier using a neural network, or a classifier generated by deep learning using a multilayer neural network. The trained neural network may be, for example, a convolutional neural network (CNN), a Bayesian neural network (BNN), or the like.

[0117] FIG. 16 is a flowchart illustrating an example of a state determination method for the hydrogen cell stack 1 in a second modified example of the first embodiment. In the second modified example of the first embodiment, as shown in FIG. 16 , the hydrogen electrochemical system 100 first performs machine learning on the classifier Ic using the simulation data Ds and the experimental data De (step Sb1). In other words, the learning unit 545 performs machine learning on the classifier Ic based on the primary physical quantity Qm (and the secondary physical quantity Qs) acquired by the acquisition unit 541 and the state of the hydrogen cell stack 1 corresponding to the primary physical quantity Qm (and the secondary physical quantity Qs). For example, as shown in FIG. 17 , the learning unit 545 inputs the simulation data Ds and the experimental data De as training data into the input layer of the neural network. Furthermore, the learning unit 545 performs calculations using, for example, the backpropagation algorithm, using the output values ​​of multiple neurons in the output layer of the neural network and the above-mentioned data Ds and De. At this time, the learning unit 545 updates the weighting coefficients of the neural network to maximize the output value of the neuron corresponding to the above-mentioned data Ds and De among the multiple neurons in the output layer. The learning unit 545 performs calculations using the backpropagation method on a larger amount of simulation data Ds and experimental data De, thereby optimizing the weighting coefficients of the neural network.

[0118] Thereafter, for example, when the input unit 51 receives an operational input indicating the start of determining the state of the operating hydrogen cell stack 1 (Yes in step Sb2), the determination unit 544 uses the classifier Ic as shown in FIG. 18 to determine the state of the operating hydrogen cell stack 1 from the monitoring data Dt (particularly at least the distribution information Dt2) of the operating hydrogen cell stack 1 (step Sb3).

[0119] The state determination in step Sb3 is then repeated, for example, until the input unit 51 receives an operational input indicating the end of state determination of the operating hydrogen cell stack 1. In other words, if the operational input indicating the end of the state determination is not received (No in step Sb4), the process returns to step Sb3. This allows the hydrogen electrochemical system 100 to determine the state of the operating hydrogen cell stack 1 in real time, even in the second modified example of the first embodiment. On the other hand, if the operational input indicating the end of the state determination is received (Yes in step Sb4), the real-time state determination process of the hydrogen cell stack 1 shown in FIG. 16 ends.

[0120] 2. Second Embodiment Next, a second embodiment will be described with reference to FIGS. 19 to 21 . FIG. 19 is a schematic diagram showing a configuration example of a hydrogen electrochemical system 100 according to a modification of the second embodiment. FIG. 20 is a perspective view showing a configuration example of a hydrogen cell 10a according to the second embodiment. FIG. 21 is an exploded perspective view of the hydrogen cell 10a according to the second embodiment. Below, configurations of the hydrogen cell stack unit 101 according to the second embodiment that differ from those of the first embodiment, its first modification, and its second modification will be described. Furthermore, components similar to those of the first embodiment, its first modification, and its second modification will be assigned the same reference numerals, and descriptions of similar configurations may be omitted.

[0121] The hydrogen cell stack unit 101 according to the second embodiment is a hydrogen generator 1012 that generates hydrogen through a water electrolysis reaction. The hydrogen generator 1012 according to this embodiment is a polymer electrolyte membrane electrolyzer cell (PEMEC) that generates hydrogen gas by performing an electrochemical reaction (i.e., water electrolysis) using liquid water using externally supplied power. However, the present invention is not limited to this example, and the hydrogen generator 1012 may be a fuel cell other than a polymer electrolyte membrane electrolyzer. For example, the hydrogen generator 1012 may be a solid oxide electrolysis cell (SOEC) or a device that generates hydrogen gas through high-temperature steam electrolysis. Alternatively, the hydrogen generator 1012 may be a device that generates hydrogen gas through the electrolysis of alkaline water, such as an alkaline water electrolysis (AWE) or an anion exchange membrane water electrolysis (AEMWE).

[0122] The control unit 102 controls each component of the hydrogen electrochemical system 100 (for example, the hydrogen generation device 1012, the material supply unit 103, the supply fluid control unit 104, the discharge fluid control unit 105, and the discharge unit 106), and also functions as a status determination device that determines the status of the hydrogen generation device 1012 in real time.

[0123] The material supply unit 103 supplies a fluid serving as a material to the hydrogen generation device 1012 via the supply fluid control unit 104. The supplied fluid may be, for example, liquid water, water vapor, or alkaline water. The supply fluid control unit 104 controls the pressure, flow rate, etc. of the fluid supplied from the material supply unit 103 to the hydrogen generation device 1012. The discharge fluid control unit 105 controls the pressure, flow rate, etc. of the fluid discharged from the hydrogen generation device 1012 to the discharge unit 106. The discharge fluid may be, for example, hydrogen gas, oxygen gas, or water vapor. The discharge unit 106 is connected to the hydrogen cell stack unit 101 via the discharge fluid control unit 105 and processes the fluid discharged from the hydrogen cell stack unit 101. The discharge unit 106 may be a mechanism that discharges the fluid to the outside by, for example, being open to the atmosphere, or may be a mechanism that traps the fluid and stores it in a storage tank or the like.

[0124] In the hydrogen cell stack unit 101 according to the second embodiment, the anode of the fuel cell 1011 in the first embodiment described above functions as the cathode of the hydrogen generation device 1012. Furthermore, the cathode of the fuel cell 1011 in the first embodiment described above functions as the anode of the hydrogen generation device 1012. Other configurations are the same as those of the hydrogen cell stack unit 101 (fuel cell 1011) in the first embodiment described above.

[0125] The cathode electrode 12a of the hydrogen generation device 1012 has a cathode catalyst layer 121a and a cathode water transport layer 122a. The cathode water transport layer 122a is a porous layer that is water permeable and conductive. At the cathode electrode 12a, oxygen gas is produced and hydrogen ions are generated by electrolysis of supplied water. The oxygen gas, together with excess water, is discharged from the cathode electrode 12a through the cathode-side flow path 141a to the discharge channel 241. The hydrogen ions move to the anode electrode 13a through the electrolyte membrane 11.

[0126] The anode electrode 13a of the hydrogen generation device 1012 has an anode catalyst layer 131a and an anode gas diffusion layer 132a. At the anode electrode 13a, hydrogen gas is generated from hydrogen ions. The hydrogen gas, together with water that seeps out of the electrolyte membrane 11 as the hydrogen gas is generated, is discharged from the anode electrode 13a to the discharge channel 221 via the anode-side flow channel 142a.

[0127] In the hydrogen electrochemical system 100 according to the second embodiment, the state of the hydrogen cell stack 1 can be determined in the same manner as in the first embodiment and its first and second modifications described above.

[0128] 2. Remarks The above describes the embodiments of the present invention. Note that the above embodiments are merely examples, and it will be understood by those skilled in the art that various modifications are possible in the combination of each component and each process, and that such modifications are within the scope of the present invention.

[0129] The present invention is useful for devices, systems, programs, etc. that evaluate units that generate either water or hydrogen through an electrochemical reaction using the other. The evaluation includes determining the state. The present invention can also be applied to devices, systems, programs, etc. that evaluate hydrogen cell stacks or hydrogen electrochemical systems.

Claims

1. An acquisition unit that acquires at least distribution information indicating the distribution of physical quantities in a hydrogen cell stack that generates one of water and hydrogen by an electrochemical reaction using the other; a storage unit that stores correlation information indicating the correlation between at least a feature quantity converted from the distribution information in the hydrogen cell stack in a state including a normal state and one or more abnormal states and the state of the hydrogen cell stack; and a determination unit that determines the state of the hydrogen cell stack during operation based on the distribution information of at least the hydrogen cell stack based on the correlation information. A hydrogen electrochemical system comprising:

2. The hydrogen electrochemical system according to claim 1, wherein the physical quantity is at least one of temperature, humidity, electric field, and magnetic field.

3. The abnormal state includes at least one of the following: floating in which a substance generated by an electrochemical reaction in the hydrogen cell stack stays in the hydrogen cell stack; drying out in which an electrolyte membrane, an anode electrode, or a cathode electrode of the hydrogen cell stack dries; deterioration of the electrolyte membrane of the hydrogen cell stack; poisoning of a catalyst included in the hydrogen cell stack; deterioration of a catalyst included in the hydrogen cell stack; and lack of a substance that undergoes an electrochemical reaction in the hydrogen cell stack. The hydrogen electrochemical system according to claim 1 or claim 2.

4. Further comprising the hydrogen cell stack in which a plurality of hydrogen cells are stacked and a sensor that detects the distribution of the physical quantity in the hydrogen cell stack, wherein the hydrogen cell is formed by laminating an anode separator, an anode electrode, an electrolyte membrane, a cathode electrode, and a cathode separator in this order, and the sensor is arranged in plurality on at least one of the anode electrode and the cathode electrode. The hydrogen electrochemical system according to any one of claims 1 to 3.

5. Further comprising a conversion unit, wherein the distribution information includes the distribution of the magnetic field in the hydrogen cell stack, the conversion unit estimates and converts the distribution of the current density from the distribution of the magnetic field by inverse problem analysis, and the determination unit determines the state of the hydrogen cell stack during operation based on the distribution information of at least the current density of the hydrogen cell stack based on the correlation information. The hydrogen electrochemical system according to any one of claims 1 to 4.

6. The hydrogen electrochemical system according to claim 5, further comprising a learning unit, wherein the acquisition unit acquires a physical quantity of the hydrogen cell stack in a state including a normal state and one or more abnormal states, the learning unit generates a conversion model of a predetermined feature quantity based on the physical quantity acquired by the acquisition unit, and the conversion unit performs conversion into the feature quantity using the conversion model.

7. The hydrogen electrochemical system according to any one of claims 1 to 5, further comprising a learning unit having a classifier, wherein the acquisition unit acquires a physical quantity of the hydrogen cell stack in a state including a normal state and one or more abnormal states, the learning unit causes the classifier to perform machine learning based on the physical quantity acquired by the acquisition unit and the state of the hydrogen cell stack corresponding to the physical quantity, and the determination unit determines the state of the hydrogen cell stack during operation from at least the distribution information of the hydrogen cell stack during operation using the classifier.

8. A method for determining the state of a hydrogen cell stack, comprising: an acquisition step of acquiring at least distribution information indicating a distribution of a physical quantity in a hydrogen cell stack that generates one of water and hydrogen by an electrochemical reaction using the other; and a determination step of determining the state of the hydrogen cell stack during operation from at least the distribution information of the hydrogen cell stack during operation based on correlation information indicating a correlation between a feature quantity converted from at least the distribution information in the hydrogen cell stack in a state including a normal state and one or more abnormal states and the state of the hydrogen cell stack.

9. A program for causing a computer to execute a method for determining the state of a hydrogen cell stack, the program causing the computer to function as means for executing: an acquisition step of acquiring at least distribution information indicating a distribution of a physical quantity in a hydrogen cell stack that generates one of water and hydrogen by an electrochemical reaction using the other; and a determination step of determining the state of the hydrogen cell stack during operation from at least the distribution information of the hydrogen cell stack during operation based on correlation information indicating a correlation between a feature quantity converted from at least the distribution information in the hydrogen cell stack in a state including a normal state and one or more abnormal states and the state of the hydrogen cell stack.

Images