Fuel cell lifespan prediction device and method

Abstract

A fuel cell lifespan prediction device according to one embodiment may comprise: an information collection unit for measuring the voltage of each of a plurality of unit cells provided in a fuel cell stack; a cell prediction model unit for generating a cell degradation prediction model so as to define a degradation pattern of each unit cell on the basis of the voltage measured in the unit cell; a lifespan prediction unit for predicting the lifespan of each unit cell on the basis of the cell degradation prediction model, and outputting the time required for the lifespan of each unit cell to reach a failure determination condition; and a stack prediction model unit for generating an integrated degradation prediction model so as to define a degradation pattern of the fuel cell stack on the basis of the lifespan of each unit cell.

Description

Fuel cell life prediction device and method

[0001] The present invention relates to a fuel cell life prediction device and method for predicting the life of a fuel cell.

[0002] Fuel cells, due to their complex component relationships and recent commercialization, lack the reliability of traditional products. In particular, compared to other energy sources, they are more susceptible to internal and external influences, resulting in poor durability. Understanding internal and external interactions and dependencies is crucial for ensuring precise reliability. Therefore, ensuring precise reliability of fuel cells is challenging. To enhance the reliability of lifespan predictions, various methods of fuel cell degradation pattern analysis are being conducted.

[0003] A fuel cell stack is a device that generates electricity through the chemical reaction of hydrogen and oxygen. It consists of multiple cells arranged in series in a stack assembly to generate the electrical energy required to operate a specific system. Due to this configuration, the overall performance of the fuel cell stack is highly dependent on the performance of each individual cell. If even one of the individual cells deteriorates, irreversibly reducing its performance, it will affect the entire stack's lifespan, making it difficult to achieve the expected stable performance. Therefore, reliable predictions of the lifespan of each individual cell, as well as the entire fuel cell stack, are essential.

[0004] Due to the series structure of fuel cell stacks, deterioration in one cell can affect adjacent cells. Therefore, when predicting fuel cell stack deterioration, it is necessary to consider not only the data characteristics of individual cells, but also the influence of adjacent cells and other external factors.

[0005] A fuel cell life prediction device and method are provided to improve the reliability of life prediction of each unit cell by modeling the deterioration pattern of each unit cell of a fuel cell stack.

[0006] A fuel cell life prediction device and method are provided that improve the reliability of life prediction of each unit cell by precisely correcting the deterioration pattern of each unit cell by considering the spatial correlation between unit cells and other unit cells.

[0007] A fuel cell life prediction device and method are provided to improve the reliability of life prediction of a fuel cell stack in which multiple unit cells are integrated through analysis of the deterioration pattern of the unit cells.

[0008] A fuel cell life prediction device and method are provided to prevent excessive replacement of the entire fuel cell stack by determining a failure of a unit cell and diagnosing replacement, thereby reducing costs and predicting an optimal repair cycle.

[0009] According to one embodiment, a fuel cell life prediction device includes: an information collection unit that measures voltages of a plurality of unit cells provided in a fuel cell stack; a cell prediction model unit that generates a cell deterioration prediction model to define a deterioration pattern of each unit cell based on the voltages measured in the unit cells; a life prediction unit that predicts the life of each unit cell based on the cell deterioration prediction model and outputs a time required for the life of each unit cell to reach a failure determination condition; and a stack prediction model unit that generates an integrated deterioration prediction model to define a deterioration pattern of the fuel cell stack based on the life of each unit cell.

[0010] The above cell degradation prediction model defines a first degradation pattern based on a voltage measured from the unit cell, generates a Mahalanobis distance between the plurality of unit cells based on the first degradation pattern, and defines a second degradation pattern of each unit cell through an exponential covariance function based on the Mahalanobis distance.

[0011] The above integrated degradation prediction model defines the degradation pattern of the fuel cell stack based on the lifespan of the unit cell with the shortest lifespan among the lifespans of each unit cell.

[0012] The life prediction unit predicts the life of the fuel cell stack based on the integrated deterioration prediction model and outputs the time required until a failure determination condition is reached.

[0013] The above cell prediction model section is characterized by defining a cell degradation prediction model as in [Mathematical Formula 1] below.

[0014] It further includes a storage unit that stores information about the voltage measured in the information collection unit, the cell degradation prediction model generated in the cell prediction model unit, and the integrated degradation prediction model generated in the stack prediction model unit.

[0015] A fuel cell life prediction method according to another embodiment includes an information collection step of measuring voltages of a plurality of unit cells provided in a fuel cell stack, respectively; a cell prediction step of generating a cell deterioration prediction model to define a deterioration pattern of each unit cell based on the voltages measured in the unit cells; a cell life prediction step of predicting the life of each unit cell based on the cell deterioration prediction model and outputting the time required for the life of each unit cell to reach a failure determination condition; and a stack prediction step of generating an integrated deterioration prediction model to define a deterioration pattern of the fuel cell stack based on the life of each unit cell.

[0016] The method further includes a stack life prediction step of predicting the life of the fuel cell stack based on the integrated prediction model and outputting the time required for each unit cell to reach a failure judgment condition.

[0017] The above cell degradation prediction model defines a first degradation pattern based on a voltage measured from the unit cell, generates a Mahalanobis distance between the plurality of unit cells based on the first degradation pattern, and defines a second degradation pattern of each unit cell through an exponential covariance function based on the Mahalanobis distance.

[0018] According to a fuel cell life prediction device and method according to one aspect, the reliability of life prediction of each unit cell is improved through modeling of the deterioration pattern of each unit cell of a fuel cell stack.

[0019] According to a fuel cell life prediction device and method according to one aspect, the reliability of life prediction of a unit cell is improved by precisely predicting the deterioration pattern of each unit cell by considering the spatial correlation between a unit cell and other unit cells.

[0020] According to a fuel cell life prediction device and method according to one aspect, the reliability of life prediction of a fuel cell stack in which a plurality of unit cells are integrated is improved through analysis of the deterioration pattern of the unit cell.

[0021] According to a fuel cell life prediction device and method according to one aspect, by determining a failure of a unit cell and diagnosing replacement, excessive replacement of the entire fuel cell stack is prevented, thereby reducing costs and predicting an optimal repair cycle.

[0022] FIG. 1 is a drawing illustrating a fuel cell stack and a fuel cell life prediction device according to one embodiment of the present invention.

[0023] FIG. 2 is a diagram illustrating the configuration of a fuel cell life prediction device according to one embodiment of the present invention.

[0024] Figure 3 is a graph of voltage over time that predicts the deterioration pattern of unit cell 1 without considering the correlation between unit cells in the fuel cell life prediction device of the present invention.

[0025] Figure 4 is a graph of voltage over time that predicts a deterioration pattern by reflecting the spatial correlation between unit cells 1 and 4 in a fuel cell life prediction device of the present invention.

[0026] Figure 5 is a graph of voltage over time that predicts a deterioration pattern by reflecting the spatial correlation between unit cells 1 and 10 in a fuel cell life prediction device of the present invention.

[0027] Figure 6 is a graph of voltage over time that predicts a deterioration pattern by reflecting the spatial correlation between unit cells 1 and 17 in a fuel cell life prediction device of the present invention.

[0028] FIG. 7 is a diagram illustrating an operation flow for a fuel cell life prediction method according to one embodiment of the present invention.

[0029] FIG. 8 is a diagram illustrating an operation flow for a cell degradation prediction model of a cell prediction model unit according to one embodiment of the present invention.

[0030] Like reference numerals refer to like elements throughout the specification. This specification does not describe all elements of the embodiments, and any general information within the technical field to which the present invention pertains or any information that overlaps between the embodiments is omitted.

[0031] Throughout the specification, when a part is said to be "connected" to another part, this includes not only direct connection but also indirect connection, and indirect connection includes connection via a wireless communication network.

[0032] Additionally, when a part is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise specifically stated.

[0033] Singular expressions include plural expressions unless the context clearly indicates otherwise.

[0034] Additionally, terms such as "~part", "~device", "~block", "~absence", and "~module" may refer to a unit that processes at least one function or operation. For example, the terms may refer to at least one piece of hardware such as an FPGA (field-programmable gate array) / ASIC (application specific integrated circuit), at least one piece of software stored in memory, or at least one process processed by a processor.

[0035] The symbols attached to each step are used to identify each step and do not indicate the order of the steps, and the steps may be performed in a different order than stated unless the context clearly indicates a specific order.

[0036] Hereinafter, with reference to the attached drawings, an embodiment of a fuel cell life prediction device and a control method thereof according to one aspect will be described in detail.

[0037] FIG. 1 is a drawing illustrating a fuel cell stack and a fuel cell life prediction device according to one embodiment of the present invention, and FIG. 2 is a drawing illustrating a configuration of a fuel cell life prediction device according to one embodiment of the present invention.

[0038] As illustrated in FIG. 1, a fuel cell stack (10) according to one embodiment of the present invention includes a plurality of unit cells (11), and a fuel cell life prediction device (100) is connected to each unit cell (11) to predict the life of the fuel cell.

[0039] A fuel cell stack (10) can be used to provide power to drive an electric vehicle. The scope of application of the fuel cell stack (10) is not limited to vehicle systems, and can also be used to provide electrical energy to electronic equipment and household electrical systems.

[0040] A fuel cell stack (10) may be composed of a plurality of unit cells (11). Specifically, the fuel cell stack (10) may refer to a stack composed by repeatedly stacking and connecting a plurality of unit cells (11).

[0041] As charge and discharge are repeated, the fuel cell stack (10) deteriorates and deteriorates, causing a decrease in available capacity. At this time, since the fuel cell stack (10) has multiple unit cells (11) connected in series, the deterioration of a unit cell (11) can electrochemically and physically affect the deterioration of adjacent unit cells (11) within the stack. Therefore, in order to predict the lifespan of the fuel cell stack (10), it is necessary to consider not only the deterioration pattern of each unit cell (11), but also the correlation between the unit cells (11) within the fuel cell stack (10).

[0042] The fuel cell life prediction device (100) can collect and store information on the unit cell (11), for example, information on the current (I) and voltage (V) of the unit cell (11) in real time or at a predetermined time cycle.

[0043] A fuel cell life prediction device (100) can generate a cell deterioration prediction model (M1) that predicts the deterioration pattern of a unit cell (11) by analyzing variability such as voltage drop based on the voltage (V) of a stored unit cell (11).

[0044] The fuel cell life prediction device (100) can quantitatively predict the deterioration of a unit cell (11) after a predetermined period of time based on a cell deterioration prediction model (M1) to be described later, and quantitatively derive the time (lifespan) required until a point in time when a failure determination condition is met. At this time, the failure determination condition (failure critical point) is set to a predetermined value, and for example, can be set to a point in time when the measured voltage of the unit cell (11) is determined to have deteriorated by 10% compared to the initial voltage.

[0045] The fuel cell life prediction device (100) can generate an integrated deterioration prediction model (M2) that predicts the deterioration pattern of the fuel cell stack (10) based on the life of the unit cell (11). Specifically, the integrated deterioration prediction model (M2) can evaluate the life of the fuel cell stack based on the unit cell with the shortest life or the unit cell that reaches the critical point first among the lifespans of each unit cell predicted by the life prediction unit (190).

[0046] The fuel cell life prediction device (100) according to the present invention can be implemented inside a vehicle. In this case, the fuel cell life prediction device (100) can be formed integrally with the internal control units of the vehicle, or can be implemented as a separate device and connected to the control units of the vehicle by a separate connection means. Here, the device (100) can operate in conjunction with the engine and motor of the vehicle, and can also operate in conjunction with a control unit that controls the operation of the engine or motor.

[0047] Referring to FIG. 2, the fuel cell life prediction device (100) may include a control unit (110), an interface unit (120), a communication unit (130), a storage unit (140), an information collection unit (150), a cell prediction model unit (160), a stack prediction model unit (180), and a life prediction unit (190). Here, the control unit (110) may process signals transmitted between each component of the device (100).

[0048] The interface unit (120) may include an input means for receiving a control command from a user and an output means for outputting the operating status and results of the device (100).

[0049] Here, the input means may include a key button, a mouse, a joystick, a jog shuttle, a stylus pen, etc. In addition, the input means may include a soft key implemented on the display.

[0050] The output means may include a display, and may also include an audio output means such as a speaker. In this case, if a touch sensor such as a touch film, a touch sheet, or a touch pad is provided in the display, the display operates as a touch screen, and the input means and the output means may be implemented in an integrated form. In this case, the display may include at least one of a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), a flexible display, a field emission display (FED), and a 3D display.

[0051] The communication unit (130) may include a communication module that supports a communication interface with electrical components and / or control units installed in the vehicle. As an example, the communication module may be connected to a fuel cell stack (10) of the vehicle (or a system that manages information of the fuel cell stack (10)) to receive information on the current and voltage of the unit cell (11) and the fuel cell stack (10).

[0052] Here, the communication module may include a module that supports vehicle network communication such as CAN (Controller Area Network) communication, LIN (Local Interconnect Network) communication, and Flex-Ray communication.

[0053] Additionally, the communication unit (130) may include a communication module that supports a communication interface with an external device. As an example, the communication module may transmit the life prediction results of the fuel cell stack (10) of the vehicle to a vehicle management system that manages the status of the vehicle.

[0054] At this time, the communication module may include a module for wireless Internet access or a module for short-range communication. Here, wireless Internet technologies may include Wireless LAN (WLAN), Wireless Broadband (Wibro), Wi-Fi, and WiMAX (World Interoperability for Microwave Access, Wimax), and short-range communication technologies may include Bluetooth, ZigBee, UWB (Ultra Wideband), RFID (Radio Frequency Identification), and Infrared Data Association (IrDA).

[0055] The storage unit (140) can store data and / or algorithms necessary for the operation of the fuel cell life prediction device (100).

[0056] For example, the storage unit (140) may store information on the current and voltage of the unit cell (11) and the fuel cell stack (10) received through the communication unit (130). In addition, the storage unit (140) may store commands and / or algorithms for generating a life prediction model, and a cell deterioration prediction model (M1) and an integrated deterioration prediction model (M2) may be stored.

[0057] In addition, the storage unit (140) may store commands and / or algorithms for predicting the life of the fuel cell stack (10), and the life prediction results of the fuel cell stack (10) using a life prediction model may also be stored.

[0058] Here, the storage unit (140) may include a storage medium such as a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), a programmable read-only memory (PROM), or an electrically erasable programmable read-only memory (EEPROM).

[0059] The information collection unit (150) collects the voltage (V) of the unit cell (11) and the fuel cell stack (10) and stores it in the storage unit (140). The information collection unit (150) can collect current and voltage from the unit cell (11) and the fuel cell stack (10) in real time, at a predetermined time cycle, or at an irregularly predetermined time.

[0060] The cell prediction model unit (160) extracts voltage information stored in the storage unit (140) and generates a cell degradation prediction model (M1) that defines the degradation pattern of a unit cell (11) according to time changes.

[0061] The cell degradation prediction model (M1) can model degradation patterns as a nonlinear stochastic process based on the relationship between accumulated voltage and time. The cell degradation prediction model (M1) can be constructed as a nonlinear Wiener process that determines the relationship between voltage and time, and can be defined as shown in [Mathematical Equation 1] below.

[0062]

[0063] In [Mathematical Formula 1], i is an index for a unit cell, which is an integer greater than or equal to 1 and less than or equal to n; j is an arbitrary point in time, which is an integer greater than or equal to 1 and less than or equal to n; is the time Voltage changes in ; is the parameter Transformation function for time; σ is the diffusion parameter; means Brownian motion.

[0064] For example, the cell prediction model unit (160) can define a cell degradation prediction model (M1) using the voltage of the accumulated unit cell (11) at each time point (1st to jth time point) that meets a predetermined condition such as a voltage drop, and store the model formula in the storage unit (140). Accordingly, the storage unit (140) stores the cell degradation prediction model (M1) for each of the i unit cells (11), and the cell degradation prediction model (M1) of each unit cell (11) can store a model generated at j time points.

[0065] The cell degradation prediction model (M1) generates a Mahalanobis distance to consider the spatial correlation of multiple unit cells (11), and can accurately predict the degradation pattern of each unit cell (11) through an exponential covariance function based on the Mahalanobis distance.

[0066] Specifically, the cell degradation prediction model (M1) may be configured to define a first degradation pattern that models a voltage drop pattern as a nonlinear stochastic process based on the voltage measured in each unit cell (11), generate a Mahalanobis distance that considers the correlation of a plurality of unit cells (11) based on the first degradation pattern of each unit cell (11), and redefine a second degradation pattern that models the voltage drop pattern as a stochastic process based on an exponential covariance function based on the Mahalanobis distance.

[0067] Mahalanobis distance is used to measure the distance between points and distributions, or the similarity between distributions, and is a concept that considers covariance. Mahalanobis distance can be defined as shown in [Mathematical Equation 2] below.

[0068]

[0069] In [Equation 2], d(s i , s j ) is Mahalanobis Street; s i , s j are the position distribution values ​​in space i,j respectively; w ij is the spatial weight; C is the covariance.

[0070] If the Mahalanobis distance is generated from the position distribution value of each unit cell based on the first degradation pattern, the second degradation pattern that considers the correlation with other unit cells (11) can be redefined by reflecting the exponential covariance function based on the generated Mahalanobis distance to the degradation pattern. At this time, the first degradation pattern and the second degradation pattern are defined by the above-described [Mathematical Formula 1], and the second degradation pattern can be defined by reflecting the Mahalanobis distance in the parameter.

[0071] Cell numberCELL 1CELL 2CELL 3CELL 4CELL 5CELL 6d(s i , s j )52.3062839.9289238.1648939.4088528.6907524.99809Cell numberCELL 7CELL 8CELL 9CELL 10CELL 11CELL 12d(s i , s j )16.0263427.9808318.47535010.8862515.48087Cell numberCELL 13CELL 14CELL 15CELL 16CELL 17CELL 18d(s i , s j )18.7174020.3042833.0284437.9172939.4492955.04064

[0072] Referring to [Table 1], [Table 1] is an example of data calculated using [Mathematical Formula 2] for the Mahalanobis distance index of a fuel cell stack consisting of 20 unit cells. Here, the Mahalanobis distance is calculated based on unit cell 10, and it can be seen that the Mahalanobis distance increases as the location gets farther away from unit cell 10.

[0073] When reflecting the exponential covariance function based on the Mahalanobis distance to the degradation pattern, the Mahalanobis distance can be calculated and applied in units of time.

[0074] When the exponential covariance function is reflected in the degradation pattern based on the Mahalanobis distance, the influence is greater when the distance from the failure judgment reference cell is close, and the influence can be set to attenuate as the distance between cells increases. This is because, when reflecting the spatial correlation based on the Mahalanobis distance of a unit cell for life prediction, the more the spatial correlation of adjacent cells is reflected, the closer the predicted life is estimated to the actual life.

[0075] Specifically, the exponential covariance function based on the Mahalanobis distance can be expressed by the following mathematical expression 3, where the cell prediction model unit (160) After modifying the part with the Mahalanobis distance index, it is reflected in the deterioration pattern.

[0076]

[0077] Here k 2 The variance of the variance-covariance function, is a parameter that adjusts the length scale, are two points x i Wow y i It represents the Euclidean distance between the cell prediction model unit (160). Part of Mahalanobis Street This is replaced with . By utilizing this, the influence according to distance in the exponential covariance function formula reflects the influence of the Mahalanobis distance form, which can additionally consider the covariance between data for two variables that produce distance values ​​compared to the existing Euclidean distance form.

[0078] The stack prediction model unit (180) creates an integrated degradation prediction model (M2) to define the degradation pattern of the fuel cell stack (10) based on the lifespan of each unit cell and stores the model formula in the storage unit (140). Specifically, the integrated degradation prediction model (M2) can evaluate the lifespan of the fuel cell stack based on the unit cell (11) with the shortest lifespan or the unit cell (11) that reaches the critical point first among the lifespans of each unit cell (11) predicted by the lifespan prediction unit (190).

[0079] At the same time, the life prediction unit (190) can predict the life of the fuel cell stack (10) based on the integrated deterioration prediction model (M2) and output the time required for the predicted life to reach the failure judgment condition.

[0080] When a life prediction event occurs when a predetermined condition is satisfied or a user input signal, etc. occurs, the life prediction unit (190) calls the cell deterioration prediction model (M1) and integrated deterioration prediction model (M2) stored in the storage unit (140).

[0081] The life prediction unit (190) can predict the life of each unit cell (11) through the called cell deterioration prediction model (M1) and output the time required for the life of the unit cell (11) to reach a predetermined deterioration judgment condition. In addition, the life prediction unit (190) can predict the life of the fuel cell stack (10) through the called integrated deterioration prediction model (M3) and output the time required for the life of the fuel cell stack (10) to reach a predetermined deterioration judgment condition.

[0082] The life prediction unit (190) can store the life prediction results of the unit cell (11) and the fuel cell stack (10) in the storage unit (140). In addition, the life prediction unit (190) can transmit the life prediction results of the unit cell (11) and the fuel cell stack (10) to a vehicle management system inside or outside the vehicle through the communication unit (130).

[0083] Accordingly, the fuel cell life prediction device (100) of the present invention can predict not only the life of the fuel cell stack (10), but also the life of each individual unit cell (11) within the stack, and thereby determine a failure of the unit cell and predict the replacement cycle.

[0084] FIG. 3 is a graph showing a deterioration pattern of unit cell 1 predicted without considering the correlation between unit cells in the cell prediction model section of the fuel cell life prediction device of the present invention, and FIGS. 4 to 6 are graphs showing a deterioration pattern predicted by a cell deterioration prediction model (M1) so as to reflect the correlation between unit cell 1 and unit cells 4, 10, and 17 in the cell prediction model section.

[0085] Referring to Fig. 3, when predicting the degradation pattern of unit cell 1 without considering the correlation between unit cells, the change in the life of the unit cell according to time is shown. Referring to Figs. 4 to 6, the cell degradation prediction model (M1) is applied to reflect the correlations of unit cell 1, unit cells 4, 10, and 17 among 20 unit cells in the cell prediction model section, and the change in the life of the unit cell according to time is shown. At this time, the reliability of the life prediction of each experiment was evaluated by the mean square error of the performance of the formed degradation paths, and the average and median values ​​among them are selected as representative values ​​to determine the reliability of each model.

[0086] Referring to [Table 2], when the correlation between unit cells is not considered, the mean square error (MSE) and mean absolute error (MSE_mean) are 5.624 and 7.946, respectively, whereas when the correlation between unit cells is considered, the mean square error (MSE) is evaluated to be 4.5 or less and the mean absolute error (MSE_mean) is evaluated to be 5.2 or less, indicating that the reliability of deterioration pattern prediction is high when the correlation is considered.

[0087] Mihalanobis distance MSEMSE mean degrees 3-5.6247.964 degrees 423.14.4825.194 degrees 5104.624.3384.908 degrees 6152.064.0064.764

[0088] [Table 3] below shows the estimated values ​​reflecting spatial correlation (unit cell 18 degradation model, Mahalla, TH=92%), and [Table 4] shows the estimated values ​​not reflecting spatial correlation (unit cell 18 degradation model TH=92%).

[0089] Referring to [Table 3], when considering the correlation of the cell located closest to the unit, the MTTF (Mean Time To Failure) is estimated to be 2,749 CYCLE, which is higher than the MTTF of 2,685 CYCLE that does not reflect spatial correlation. This confirms that the MTTF that reflects spatial correlation is estimated to be closer to the actual MTTF than the MTTF that does not, compared to the MTTF of 2,975 CYCLE at which unit cell 18 actually reaches the failure threshold.

[0090] Spatial location (mahalla) αβk 2 lσ 2 B 10MTTF cell 17-2.683e-056.545e-051.7691.3297.666e-062.3182.749(-2.721e-05, -2.646e-05)(6.532e-05, 6.559e-05(1.769, 1.770)(1.329, 1.330)cell 15-2.798e-056.413e-051.9611.9717760e-062,2482,690(-2.818e-05,-2.77e-05)(6.403e-05,6.423e-05)(1.961,1.8623)(1.971,1.972) cell 13-2.751e-056.110e-051.8271.6537.335e-062,2812,673(-2.756e-05,-2.746e-05)(6.109e-05,6.112e-05)(1.827,1.828)(1.653,1.654) cell 11-2.544e-056.509e-051.6121.3867.623e-062,2712,674(-2.549e- 05,-2.539e-05)(6.508e-05,6.511e-05)(1.612,1.613)(1.386,1.387

[0091] Here, the formula corresponding to [Table 3] (reflecting spatial correlation) is as shown in Mathematical Formula 4 below, and α and β are parameters of the 'exponent + power' formula, k 2 and are parameters that adjust the variance and length scale in the exponential covariance function, f(distance) is a distance index (Euclidean, Mahalanobis, etc.) to be included in the exponential covariance function, and σ corresponds to the diffusion coefficient of the Wiener process model. In the disclosed embodiment, spatial correlation information is reflected in f(distance) as Mahalanobis distance, not Euclidean distance. B 10 And MTTF is not a parameter but an indicator related to the fuel cell life.

[0092]

[0093] Parameter Estimate 95% Confidence Interval α-2.761e-05-2.845e-05-2.676e-05β1.897e-021.897e-021.898e-02c1.965e-031.964e-031.966e-03σ2 8.771e-06B 10 2,215MTTF2,685

[0094] The equation corresponding to Table 4 (excluding spatial correlation) is as shown in Mathematical Equation 5 below, where α, β, and c are all parameters of the 'exponent + power' equation. Here, α, β, and c are parameters that adjust the overall size or scale of the function, adjust the degree of data degradation, and adjust the growth rate of the function over time, respectively.

[0095]

[0096] Below, a method for predicting the life of a fuel cell according to an embodiment of the present invention is described.

[0097] FIG. 7 is a diagram illustrating an operation flow for a fuel cell life prediction method according to one embodiment of the present invention, and FIG. 8 is a diagram illustrating an operation flow for a cell deterioration prediction model of a cell prediction model unit.

[0098] Referring to FIG. 7, a fuel cell life prediction method (S100) may include an information collection step (S110) of measuring the voltage of each of a plurality of unit cells provided in a fuel cell stack, a cell prediction step (S120) of generating a cell degradation prediction model (M1) to define a degradation pattern of each unit cell based on the voltage measured in the unit cell, a cell life prediction step (S130) of predicting the life of each unit cell based on the cell degradation prediction model (M1) and outputting the time required for the life of each unit cell to reach a failure determination condition, and a stack prediction step (S140) of generating an integrated degradation prediction model to define a degradation pattern of the fuel cell stack based on the life of each unit cell. In addition, the fuel cell life prediction method may further include a stack life prediction step (S150) of predicting the life of the fuel cell stack based on the integrated prediction model and outputting the time required for the life of each unit cell to reach a failure determination condition.

[0099] Referring to FIG. 8, the cell degradation prediction model (M1) may be configured to define a first degradation pattern based on a voltage measured in a unit cell (S121), generate a Mahalanobis distance between the plurality of unit cells based on the first degradation pattern (S122), and define a second degradation pattern of each unit cell using an exponential covariance function based on the Mahalanobis distance (S123).

[0100] The disclosed embodiments have been described with reference to the attached drawings as described above. Those skilled in the art will understand that the present invention can be implemented in forms other than the disclosed embodiments without altering the technical spirit or essential features of the present invention. The disclosed embodiments are illustrative and should not be construed as limiting.

Claims

1. An information collection unit that measures the voltage of each of a plurality of unit cells provided in a fuel cell stack; A cell prediction model unit that generates a cell degradation prediction model to define a degradation pattern of each unit cell based on the voltage measured from the unit cell; A life prediction unit that predicts the life of each unit cell based on the cell deterioration prediction model and outputs the time required for the life of each unit cell to reach a failure judgment condition; and A fuel cell life prediction device including a stack prediction model section that generates an integrated deterioration prediction model to define a deterioration pattern of the fuel cell stack based on the life of each unit cell.

2. In paragraph 1, The above cell deterioration prediction model is A fuel cell life prediction device which defines a first degradation pattern based on a voltage measured from the unit cell, generates a Mahalanobis distance between the plurality of unit cells based on the first degradation pattern, and defines a second degradation pattern of each unit cell by considering an exponential covariance function based on the Mahalanobis distance.

3. In paragraph 2, The above integrated deterioration prediction model is A fuel cell life prediction device that defines a deterioration pattern of the fuel cell stack based on the life of the unit cell with the shortest life among the respective unit cells.

4. In paragraph 3, The above life prediction section A fuel cell life prediction device that predicts the life of the fuel cell stack based on the above integrated deterioration prediction model and outputs the time required until a failure judgment condition is reached.

5. In paragraph 1, The above cell prediction model part A fuel cell life prediction device characterized by defining a cell deterioration prediction model as in the following [Mathematical Formula 1]. [Mathematical formula 1] (Here, i is an index for a unit cell, an integer greater than or equal to 1 and less than or equal to n; j is an arbitrary point in time, an integer greater than or equal to 1 and less than or equal to n; is the time Voltage changes in ; is the parameter Transformation function for time; σ is the diffusion parameter; stands for Brownian motion) 6. In paragraph 1, A fuel cell life prediction device further comprising a storage unit that stores information on the voltage measured by the information collection unit, the cell deterioration prediction model generated by the cell prediction model unit, and the integrated deterioration prediction model generated by the stack prediction model unit.

7. Information collection step of measuring the voltage of each of the multiple unit cells provided in the fuel cell stack; A cell prediction step for generating a cell degradation prediction model to define a degradation pattern of each unit cell based on the voltage measured in the unit cell; A cell life prediction step for predicting the life of each unit cell based on the cell deterioration prediction model and outputting the time required for the life of each unit cell to reach a failure judgment condition; and A fuel cell life prediction method comprising a stack prediction step of generating an integrated degradation prediction model to define a degradation pattern of a fuel cell stack based on the lifetime of each of the unit cells.

8. In paragraph 7, A fuel cell life prediction method further comprising a stack life prediction step of predicting the life of the fuel cell stack based on the integrated prediction model and outputting the time required for each unit cell to reach a failure judgment condition.

9. In paragraph 8, The above cell deterioration prediction model is A fuel cell life prediction method comprising: defining a first degradation pattern based on a voltage measured from the unit cell; generating a Mahalanobis distance between the plurality of unit cells based on the first degradation pattern; and defining a second degradation pattern of each unit cell by considering an exponential covariance function based on the Mahalanobis distance.

Images