Cell characteristics estimation device, cell characteristics estimation method, and program

Abstract

A cell characteristics estimation device according to an embodiment of the present invention has a data acquisition unit, a data extraction unit, a relevant information acquisition unit, and an estimation unit. The data acquisition unit acquires first data including at least one of a surface image which is an image of the surface of a perovskite layer in a perovskite solar cell, and an X-ray diffraction waveform obtained from the perovskite layer. The data extraction unit extracts, from the first data, second data indicating a parameter relating to the crystal state of the perovskite layer. The relevant information acquisition unit acquires relevant information in which the first data and / or the second data is associated with cell characteristics of the perovskite solar cell. The estimation unit estimates the cell characteristics from the first data and / or the second data on the basis of the relevant information.

Description

Battery Performance Estimation Device, Battery Performance Estimation Method, and Program

[0001] Embodiments of the present invention relate to a battery performance estimation device, a battery performance estimation method, and a program.

[0002] Conventionally, a technique for estimating the battery performance of a perovskite solar cell including a photoelectric conversion element using a perovskite material is known. For example, a battery performance estimation device has been proposed that acquires an image of a cut surface of a stacked structure of a perovskite solar cell in the stacking direction using an electron microscope and estimates battery performance such as current-voltage characteristics (I-V characteristics) based on the acquired cross-sectional image.

[0003] However, in conventional battery performance estimation devices, it is necessary to cut the stacked structure of the perovskite solar cell, so that it is not easy to estimate the battery performance, and since the cross-sectional image is a local image of the stacked structure, there is a possibility that the battery performance cannot be estimated with high accuracy.

[0004] Japanese Patent No. 7319733

[0005] The problem to be solved by the present invention is to provide a battery performance estimation device, a battery performance estimation method, and a program that can easily and accurately estimate the battery performance of a perovskite solar cell.

[0006] The battery performance estimation device according to the embodiment includes a data acquisition unit, a data extraction unit, a related information acquisition unit, and an estimation unit. The data acquisition unit acquires first data including at least one of a surface image that is an image of the surface of a perovskite layer in a perovskite solar cell and an X-ray diffraction waveform obtained from the perovskite layer. The data extraction unit extracts second data indicating parameters related to the crystal state of the perovskite layer from the first data. The related information acquisition unit acquires related information in which at least one of the first data and the second data is associated with the battery performance of the perovskite solar cell. The estimation unit estimates the battery performance from at least one of the first data and the second data based on the related information.

[0007] A diagram showing the structure of the perovskite solar cell 100. A block diagram showing the configuration of the battery performance estimation device 200. A diagram showing the first example of processing performed by the battery performance estimation device 200. A diagram explaining the related information 310 as a trained model. A diagram schematically showing a part of the neural network used in the related information 310 as a trained model. A diagram showing an example of analysis results obtained from a surface image of the perovskite layer 174. A diagram showing an example of analysis results obtained from data extracted from the X-ray diffraction waveform of the perovskite layer 174. A diagram showing the second example of processing performed by the battery performance estimation device 200. A diagram showing the third example of processing performed by the battery performance estimation device 200. A diagram showing another example of the learning phase processing performed by the battery performance estimation device 200.

[0008] The battery performance estimation device, battery performance estimation method, and program of the embodiment will be described below with reference to the drawings.

[0009] (1) Perovskite Solar Cell Figure 1 shows the structure of a perovskite solar cell 100. The perovskite solar cell 100 includes, for example, a cable 110, a junction box 120, a connector 130, a front sheet 140, a sealing material 150, an interconnector 160, a power generation element 170, a sealing material 180, and a back sheet 190.

[0010] Cable 110 is connected to connector 130 and interconnector 160 through junction box 120 respectively. Front sheet 140 is provided on the light-receiving surface side of perovskite solar cell 100 and is a sheet for protecting power generation element 170. Encapsulant 150 is provided on the light-receiving surface side of perovskite solar cell 100 and is a sheet for encapsulating power generation element 170. Interconnector 160 is a connector connected to power generation element 170. Power generation element 170 is an element that converts light energy into electrical energy. Encapsulant 180 is provided on the back surface side of perovskite solar cell 100 and is a sheet for encapsulating power generation element 170. Back sheet 190 is provided on the back surface side of perovskite solar cell 100 and is a sheet for protecting power generation element 170. In FIG. 1, as an example, junction box 120 is attached to the light-receiving surface side, but it may be attached to the back surface side or the side surface side.

[0011] The power generation element 170 includes, for example, a film substrate 171, a transparent electrode 172, a hole transport layer 173, a perovskite layer 174, an electron transport layer 175, and a metal evaporation film 176. For example, the perovskite layer 174 includes at least one of the compounds represented by A １ BX １ ３ and the compounds represented by A ２ ２ A １ （ｍ－１） B ｍ X １ （３ｍ＋１） and includes at least one selected from the group consisting of Cs １ is a monovalent cation, and the R in the A １ in the A ＋ Rb ＋ K ＋ Na ＋ R １ NH ３ ＋ R １ ２ NH ２ ＋ and HC(NH ２ ) ２ ＋ and the A １ in the A１ This is at least one monovalent group selected from the group consisting of hydrogen, a linear alkyl group containing 1 to 18 carbon atoms, a branched alkyl group containing 1 to 18 carbon atoms, a cyclic alkyl group containing 1 to 18 carbon atoms, a substituted aryl group, an unsubstituted aryl group, a substituted heteroaryl group, and an unsubstituted heteroaryl group, and the above A ２ R １ HN ３ ＋ , R １ ２ NH ２ ＋ , C(NH ２ ) ３ ＋ , and R ２ C ２ H ４ NH ３ ＋ A monovalent cation comprising at least one selected from the group consisting of the above A ２ In the R ２ is at least one monovalent group selected from the group consisting of substituted aryl groups, unsubstituted aryl groups, substituted heteroaryl groups, and unsubstituted heteroaryl groups, and B is Pb ２＋ Sn ２＋ and Ge ２＋ A divalent cation comprising at least one selected from the group consisting of the above, and the X １ is F － , Cl － , Br － , I － SCN － , and CH ３ COO － It is at least one monovalent anion selected from the group, and m is an integer between 1 and 20. B is Pb ２＋ In this case, the ability to efficiently absorb light and generate charge carriers is high, which increases the likelihood of achieving high photoelectric conversion efficiency. Furthermore, compared to other divalent cation perovskite crystals with photoelectric conversion capabilities, it has the characteristic of being highly stable, being able to maintain its crystal structure relatively stably. The above B is Sn ２＋ In the case of Sn ２＋Although it has durability issues because it is easily oxidized, Pb ２＋ It has the characteristic of absorbing light in the longer wavelength region (near-infrared region), and can convert more sunlight into photoelectric energy, and as an environmental consideration, it does not contain harmful Pb ２＋ Because it has the advantage of being able to reduce [something], the features of the present invention can be fully utilized. The main element of B is Pb ２＋ or Sn ２＋ In this case, the crystal structure is stabilized and an improvement in power generation performance can be expected, therefore Ge ２＋ The perovskite layer 174 may have cubic (α) phase, tetragonal (β) phase, trigonal (δ) phase, and orthorhombic (γ) phase. Furthermore, the crystalline phase of the perovskite layer 174 changes with temperature. For example, if the perovskite material is methylammonium lead iodide (MAPbI) ３ In the case of ), the crystalline phase of the perovskite layer 174 transitions from α phase → β phase → γ phase in response to the change from high temperature to low temperature. Also, if the perovskite material is formamidinium triiodide lead (FAPbI), the crystalline phase of the perovskite layer 174 transitions from α phase → β phase → γ phase. ３ In this case, the crystalline phase of the perovskite layer 174 transitions from α phase → δ phase → β / γ phase in response to the change from high temperature to low temperature. The α phase is the most preferable crystalline phase for power generation because it has a high light absorption coefficient and good light absorption characteristics. The β / γ phase also generates electricity, but its power generation efficiency is not as good as that of the α phase. The δ phase does not generate electricity because it is photoelectrically inactive. For these reasons, the crystalline phase of the perovskite layer 174 in this embodiment may contain the δ phase, but it is preferable to contain at least one of the α, β, and γ phases, and it is more preferable to contain the most α phase. As a result, the perovskite solar cell 100 of this embodiment can improve its power generation efficiency.

[0012] When sunlight strikes the perovskite solar cell 100, electrons and holes are generated in the perovskite layer 174. The generated electrons are transported via the electron transport layer 175 to the metal vapor-deposited film 176, which functions as the cathode. The generated holes are transported via the hole transport layer 173 to the transparent electrode 172, which functions as the anode. This transfer of charge generates electricity in the perovskite solar cell 100. Figure 1 shows an example where the hole transport layer 173 is provided on the light-receiving side of the perovskite layer 174 and the electron transport layer 175 is provided on the back side, but the order can be reversed.

[0013] (2) Battery performance estimation device Figure 2 is a block diagram showing the configuration of the battery performance estimation device 200. The battery performance estimation device 200 is a computer that estimates the battery performance of the perovskite solar cell 100. The battery performance estimation device 200 includes, for example, an input unit 210, a display unit 220, a data acquisition unit 230, a data extraction unit 240, a related information acquisition unit 250, an estimation unit 260, a learning unit 270, a determination unit 280, an analysis unit 290, and a storage unit 300.

[0014] The input unit 210 may be, for example, an input device such as a keyboard, mouse, or touch panel. The display unit 220 may be a display device such as a CRT (Cathode Ray Tube) display, liquid crystal display, or organic EL (Electro-Luminescence) display.

[0015] The data acquisition unit 230, data extraction unit 240, related information acquisition unit 250, estimation unit 260, learning unit 270, determination unit 280, and analysis unit 290 are implemented, for example, by a hardware processor such as a CPU (Central Processing Unit) executing a program (software). Some or all of these components may be implemented by hardware (including circuitry) such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), and GPU (Graphics Processing Unit), or by the cooperation of software and hardware. The program may be stored in advance in a storage device such as an HDD (Hard Disk Drive) or flash memory (a storage device equipped with a non-transient storage medium), or it may be stored in a removable storage medium such as a DVD or CD-ROM (a non-transient storage medium) and installed in the storage device when the storage medium is mounted in a drive device.

[0016] The storage unit 300 can be an HDD, flash memory, RAM (Random Access Memory), etc. The storage unit 300 may also be a NAS (Network Attached Storage) device that can be accessed by the battery performance estimation device 200 via a network. The storage unit 300 stores information such as related information 310. Details of the related information 310 will be described later.

[0017] (3) Processing by the Battery Performance Estimation Device The battery performance estimation device 200 of this embodiment estimates the battery performance based on the relevant information 310, using at least one of the first data (for example, the surface image of the perovskite layer 174 and the X-ray diffraction waveform) and the second data (parameters relating to the crystal state of the perovskite layer 174). Three examples of the details of the processing performed by the battery performance estimation device 200 will be described below.

[0018] (3-1) First Example Figure 3 is a diagram showing a first example of the processing performed by the battery performance estimation device 200. As a first example, an example of estimating battery performance from first data and second data based on related information 310 will be explained for each of the three phases (learning phase, estimation phase, and analysis phase).

[0019] (3-1-1) Learning Phase The learning phase is a phase in which the learning unit 270 generates related information 310 by learning training data. First, the data acquisition unit 230 acquires first data. The first data includes at least one of the following: a surface image, which is an image of the surface of the perovskite layer 174 in the perovskite solar cell 100, and an X-ray diffraction waveform obtained from the perovskite layer 174. The surface image may be an image of the surface of the perovskite layer 174 taken by a scanning electron microscope (SEM) (surface SEM image). The surface image is an image observed by irradiating the sample with an electron beam, and includes not only secondary electron images, but also backscattered electron images, electron beam backscatter images, and compositional images from characteristic X-rays. The X-ray diffraction waveform is a waveform that shows the X-ray diffraction pattern (relationship between diffraction angle and diffraction intensity) when the perovskite layer 174 is irradiated with X-rays. When the angle of incidence of X-rays on the surface of the perovskite layer 174 is θ, it is preferable that the X-ray diffraction waveform includes waveforms in the range of 5° ≤ 2θ ≤ 45°. This is because characteristic peaks of the perovskite crystal structure and unreacted raw materials / impurities often appear in the range of 5° ≤ 2θ ≤ 45° in the X-ray diffraction waveform. The data acquisition unit 230 outputs the acquired first data to the data extraction unit 240 and the learning unit 270.

[0020] The data extraction unit 240 extracts second data from the first data acquired by the data acquisition unit 230. The second data is data indicating parameters related to the crystal state of the perovskite layer 174. For example, if the first data includes a surface image of the perovskite layer 174, the data extraction unit 240 may extract at least one of the following as second data: the number of crystals in the perovskite layer 174, the size of the crystals, the shape of the crystals, and the amount of foreign matter. Also, if the first data includes an X-ray diffraction waveform, the data extraction unit 240 may extract at least one of the following as second data: the position, intensity, width, distortion due to fitting, and degree of crystallinity of the peaks in the X-ray diffraction waveform. For example, the data extraction unit 240 may perform fitting on the peaks in the X-ray diffraction waveform using a Voigt function, which is a combination of a Gaussian function and a Lorentz function. Furthermore, the data extraction unit 240 may extract the degree to which the peak waveform obtained by fitting is asymmetrical with respect to the peak position (the degree to which it is biased to the left or right) as distortion due to fitting. Also, the degree of crystallinity is a value that indicates the proportion of crystalline material to the whole sample. The data extraction unit 240 may extract the degree of crystallinity from the X-ray diffraction waveform by any method. For example, the data extraction unit 240 may extract the degree of crystallinity from the X-ray diffraction waveform using a method for calculating the degree of crystallinity by quantitative analysis of crystalline or amorphous components, a method for calculating the degree of crystallinity by peak separation, the Ruland method, or the Vonk method. The data extraction unit 240 outputs the extracted second data to the learning unit 270.

[0021] Next, the learning unit 270 acquires first data from the data acquisition unit 230 and second data from the data extraction unit 240. The learning unit 270 also acquires the battery performance of the perovskite solar cell 100, which has a perovskite layer 174 from which the first and second data have been acquired. The battery performance acquired here is not an estimated value, but the known battery performance of the perovskite solar cell 100. The battery performance may be, for example, the current-voltage characteristics (I-V characteristics) of the perovskite solar cell 100, or the durability of the perovskite solar cell 100. Durability may be, for example, a characteristic that shows the relationship between durability time and conversion efficiency. For example, durability may be measured by a Damp Heat (high temperature and high humidity) test or a Light-soaking (light irradiation) test as specified by IEC 61646, one of the international standards for thin-film solar cell modules. Hereafter, "durability" will have the same meaning. The battery performance may be a value entered by the user using the input unit 210, a value stored in the storage unit 300 beforehand, or a value obtained from an external device via a network.

[0022] Next, the learning unit 270 generates related information 310 based on the first data, the second data, and the battery performance. The related information 310 may be, for example, a trained model in which the first data, the second data, and the battery performance are associated using a neural network. Specifically, the learning unit 270 generates the related information 310, which is a trained model, by learning training data in which the first data, the second data, and the battery performance are associated. The learning unit 270 stores the generated related information 310 in the storage unit 300. The related information 310 generated here will be used in the battery performance estimation process during the estimation phase.

[0023] Figure 4 illustrates the related information 310 as a trained model. The related information 310 includes an input layer, an intermediate layer, and an output layer. The related information 310 may include multiple intermediate layers.

[0024] When the first data (surface image, X-ray diffraction waveform) and the second data (number of crystals, crystal size, crystal shape, amount of foreign matter, peak intensity, peak width, etc.) are input to the input layer of the related information 310, the battery performance (I-V characteristics, durability, etc.) is output from the output layer. If the first data is a surface image, the foreign matter is unreacted raw material (e.g., PbI ２ Ya SnI ２ This may include dust, fine metal particles from the film deposition apparatus / tray, or residual solvent. Furthermore, if the first data is an X-ray diffraction waveform, the foreign matter may be unreacted raw materials (e.g., PbI). ２ Ya SnI ２ ), or fine metals derived from the film deposition apparatus / tray, etc.

[0025] Figure 5 schematically shows a part of the neural network used for the related information 310 as a trained model. x represents the value of the unit (node) in each layer. The value of each unit is calculated using a predetermined function by multiplying the output of the unit in the previous layer by the weight w and adding the results. These can be expressed by the following equations.

[0026] y t m =w t-1 1m ・x t-1 1 + w t-1 2m ・x t-1 2 + ... + w t-1 nm ・x t-1 n +b t-1 x t m = f(y t m ) ... (1)

[0027] Here, b is the bias term. Functions such as ReLU and sigmoid functions are used for function f. By repeating the calculation in equation (1) above from the input layer to the output layer, the battery performance based on the related information 310 can be determined. "Learning" means adjusting the weights w and bias term b to appropriate values. Learning is performed using methods such as stochastic gradient descent. That is, random values ​​are given to the weights w and bias term b of each layer, and the training data dataset is provided to the input layer. In the initial stages of learning, the output value output from the output layer will be an incorrect value, but in the case of training data, the value that should be output (target value) is known, so the weights are updated in reverse order from the output layer to the input layer so that the difference (error) between the output value and the target value becomes small. This is called backpropagation, and the error between the output value and the target value is expressed as squared error or cross-entropy. At this time, if the error is expressed as a differentiable function, the amount to adjust (gradient) to reduce the error can be calculated. Note that the trained model shown in Figure 5 is just one example, and the training method is not limited to the explanation above.

[0028] (3-1-2) Estimation Phase Next, the estimation phase will be explained. The estimation phase is the phase in which the estimation unit 260 estimates the battery performance of the perovskite solar cell 100. First, the related information acquisition unit 250 acquires related information 310 from the storage unit 300. As mentioned above, the related information 310 may be a trained model used to estimate the battery performance. The related information acquisition unit 250 outputs the acquired related information 310 to the estimation unit 260.

[0029] The data acquisition unit 230 acquires first data of the perovskite solar cell 100 whose battery performance is to be estimated. As described above, the first data includes a surface image of the perovskite layer 174 and an X-ray diffraction waveform. The data acquisition unit 230 outputs the acquired first data to the data extraction unit 240 and the estimation unit 260.

[0030] The data extraction unit 240 extracts second data for the perovskite solar cell 100, whose battery performance is to be estimated, from the first data acquired by the data acquisition unit 230. As described above, the second data is data indicating parameters related to the crystal state of the perovskite layer 174. The data extraction unit 240 outputs the extracted second data to the estimation unit 260.

[0031] Next, the estimation unit 260 acquires related information 310 from the related information acquisition unit 250, acquires first data from the data acquisition unit 230, and acquires second data from the data extraction unit 240. The estimation unit 260 also estimates the battery performance from the first data and second data based on the related information 310. Specifically, the estimation unit 260 inputs the first data and second data into the related information 310, which is a trained model, to obtain the battery performance output from the related information 310. As mentioned above, the battery performance may be, for example, the current-voltage characteristics (I-V characteristics) of the perovskite solar cell 100, or the durability of the perovskite solar cell 100. After that, the estimation unit 260 outputs the battery performance to the display unit 220.

[0032] Next, the display unit 220 displays the battery performance output from the estimation unit 260. This allows the battery performance estimation device 200 to inform the user of the battery performance of the perovskite solar cell 100 before the perovskite solar cell 100 is manufactured.

[0033] Furthermore, the estimation unit 260 outputs the first data, the second data, battery performance, and related information 310 to the analysis unit 290. This information will be used in the analysis phase.

[0034] (3-1-3) Analysis Phase Next, the analysis phase will be described. The analysis phase is the phase in which the analysis unit 290 analyzes the parts of the first data and second data that contribute to battery performance. First, the analysis unit 290 obtains the first data, the second data, battery performance, and related information 310 from the estimation unit 260. Next, the analysis unit 290 analyzes the parts of at least one of the first data and second data that contribute to battery performance based on the first data, the second data, battery performance, and related information 310. Specifically, the analysis unit 290 may perform the analysis processing using XAI (Explainable Artificial Intelligence) methods such as GradCAM. XAI is called explainable AI and is an AI that can explain the process that led to the output result and the basis for the judgment.

[0035] The analysis unit 290 outputs the analysis results to the display unit 220. This results in the analysis results being displayed on the display unit 220. The details of the analysis results from the analysis unit 290 are described below.

[0036] Figure 6 shows an example of analysis results obtained from a surface image of the perovskite layer 174. As shown in Figure 6, the surface image IM1 and the analysis result image IM2 are displayed on the display unit 220 as analysis results. Surface image IM1 is an image of the surface of the perovskite layer 174 included in the first data. Analysis result image IM2 is an image in which the surface image is color-coded according to the degree of contribution to battery performance. Analysis result image IM2 may be a heat map, for example, where the higher the contribution, the closer it is to red, and the lower the contribution, the closer it is to blue. By checking this analysis result, the user can easily understand which parts of the surface image of the perovskite layer 174 contribute to battery performance.

[0037] Figure 7 shows an example of analysis results obtained from data extracted from the X-ray diffraction waveform of the perovskite layer 174. Figure 7 is a conceptual diagram to explain the peak position and intensity of the X-ray diffraction waveform and does not accurately reflect the actual peak position and intensity. As shown in Figure 7, the X-ray diffraction waveform IM3 and the analysis result image IM4 are displayed on the display unit 220 as analysis results. Similar to Figure 7, the X-ray diffraction waveform IM3 is the X-ray diffraction waveform included in the first data. For example, if the angle of incidence of X-rays on the surface of the perovskite layer 174 is θ, the horizontal axis may be 2θ and the vertical axis may be the intensity of the X-rays. The analysis result image IM4 may, for example, have the degree of contribution to battery performance (contribution) on the horizontal axis and the items of data extracted from the X-ray diffraction waveform on the vertical axis. The data extracted from the X-ray diffraction waveform on the vertical axis is data included in the second data extracted by the data extraction unit 240, and may be, for example, the position, intensity, and width of the X-rays. The user can confirm these analysis results. It is easy to determine which peaks in the X-ray diffraction waveform of the perovskite layer 174 contribute to battery performance, based on their position, intensity, and width.

[0038] Furthermore, using a surface image of the perovskite layer 174 is more effective than using a cross-sectional image. For example, when using a cross-sectional image, it is necessary to cut the laminated structure (power generation element 170) of the perovskite solar cell 100, making it difficult to easily estimate the battery performance. Also, since the cross-sectional image is a localized image of the laminated structure, it is not possible to estimate the battery performance with high accuracy. In contrast, when using a surface image of the perovskite layer 174, it is not necessary to cut the laminated structure, the crystal grain size in the perovskite layer 174 can be easily measured, and foreign matter attached to the surface can be easily detected. For these reasons, using a surface image of the perovskite layer 174 is superior to using a cross-sectional image.

[0039] Furthermore, using the X-ray diffraction waveform of the perovskite layer 174 is more effective than using cross-sectional images. As mentioned above, cross-sectional images are localized images of the stacked structure and therefore cannot accurately estimate battery performance. In contrast, using the X-ray diffraction waveform of the perovskite layer 174 allows for the measurement of data regarding the crystalline state of the perovskite layer 174 over a wider area than using cross-sectional images. For this reason, using the X-ray diffraction waveform of the perovskite layer 174 is superior to using cross-sectional images.

[0040] As described above, in the first example of this embodiment, the data acquisition unit 230 acquires first data, the data extraction unit 240 extracts second data from the first data, the related information acquisition unit 250 acquires related information 310 from the storage unit 300, and the estimation unit 260 estimates the battery performance from the first data and second data based on the related information 310. As a result, the battery performance estimation device 200 of this embodiment can easily and accurately estimate the battery performance of the perovskite solar cell 100.

[0041] Furthermore, in the first example of this embodiment, the analysis unit 290 analyzes the parts of the first data and the second data that contribute to the battery performance, based on the first data, the second data, the battery performance, and the related information 310. This allows the user to easily understand which parts of the first data and the second data contribute to the battery performance.

[0042] (3-2) Second Example Figure 8 shows a second example of the processing performed by the battery performance estimation device 200. As a second example, an example of estimating battery performance from the first data based on related information 310 will be explained for each of the three phases (learning phase, estimation phase, and analysis phase). In this second example, the second data will not be used.

[0043] (3-2-1) Learning Phase The learning phase is a phase in which the learning unit 270 generates related information 310 by learning training data. First, the data acquisition unit 230 acquires first data. As described above, the first data includes at least one of the surface image of the perovskite layer 174 and the X-ray diffraction waveform. The data acquisition unit 230 outputs the acquired first data to the learning unit 270.

[0044] Next, the learning unit 270 acquires first data from the data acquisition unit 230. The learning unit 270 also acquires the battery performance of the perovskite solar cell 100, which has a perovskite layer 174 from which the first data has been acquired. The battery performance acquired here is not an estimated value, but the known battery performance of the perovskite solar cell 100. As mentioned above, the battery performance may be, for example, the current-voltage characteristics (I-V characteristics) of the perovskite solar cell 100, or the durability of the perovskite solar cell 100.

[0045] Next, the learning unit 270 generates related information 310 based on the first data and the battery performance. The related information 310 may be, for example, a trained model in which the first data and the battery performance are associated using a neural network. Specifically, the learning unit 270 generates the related information 310, which is a trained model, by learning training data in which the first data and the battery performance are associated. The learning unit 270 stores the generated related information 310 in the storage unit 300. The related information 310 generated here will be used in the battery performance estimation process during the estimation phase.

[0046] (3-2-2) Estimation Phase Next, the estimation phase will be explained. The estimation phase is the phase in which the estimation unit 260 estimates the battery performance of the perovskite solar cell 100. First, the related information acquisition unit 250 acquires related information 310 from the storage unit 300. As mentioned above, the related information 310 may be a trained model used to estimate the battery performance. The related information acquisition unit 250 outputs the acquired related information 310 to the estimation unit 260.

[0047] The data acquisition unit 230 acquires first data of the perovskite solar cell 100 whose battery performance is to be estimated. As described above, the first data includes a surface image of the perovskite layer 174 and an X-ray diffraction waveform. The data acquisition unit 230 outputs the acquired first data to the estimation unit 260.

[0048] Next, the estimation unit 260 acquires related information 310 from the related information acquisition unit 250 and acquires first data from the data acquisition unit 230. The estimation unit 260 also estimates the battery performance from the first data based on the related information 310. Specifically, the estimation unit 260 inputs the first data into the related information 310, which is a trained model, and acquires the battery performance output from the related information 310. As mentioned above, the battery performance may be, for example, the current-voltage characteristics (I-V characteristics) of the perovskite solar cell 100, or the durability of the perovskite solar cell 100. After that, the estimation unit 260 outputs the battery performance to the display unit 220.

[0049] Next, the display unit 220 displays the battery performance output from the estimation unit 260. This allows the battery performance estimation device 200 to inform the user of the battery performance of the perovskite solar cell 100 before the perovskite solar cell 100 is manufactured.

[0050] Furthermore, the estimation unit 260 outputs the first data, battery performance, and related information 310 to the analysis unit 290. This information will be used in the analysis phase.

[0051] (3-2-3) Analysis Phase Next, the analysis phase will be described. The analysis phase is the phase in which the analysis unit 290 analyzes the parts of the first data that contribute to the battery performance. First, the analysis unit 290 obtains the first data, battery performance, and related information 310 from the estimation unit 260. Next, the analysis unit 290 analyzes the parts of the first data that contribute to the battery performance based on the first data, battery performance, and related information 310. Specifically, the analysis unit 290 may perform the analysis processing using XAI methods such as GradCAM.

[0052] The analysis unit 290 outputs the analysis results to the display unit 220. As a result, the analysis results are displayed on the display unit 220. For example, the surface image IM1 and analysis result image IM2 shown in Figure 6 may be displayed on the display unit 220 as analysis results. Alternatively, the X-ray diffraction waveform IM3 and analysis result image IM4 shown in Figure 7 may be displayed on the display unit 220 as analysis results.

[0053] As described above, in the second example of this embodiment, the data acquisition unit 230 acquires first data, the related information acquisition unit 250 acquires related information 310 from the storage unit 300, and the estimation unit 260 estimates the battery performance from the first data based on the related information 310. As a result, the battery performance estimation device 200 of this embodiment can easily and accurately estimate the battery performance of the perovskite solar cell 100.

[0054] Furthermore, in the second example of this embodiment, the analysis unit 290 analyzes the parts of the first data that contribute to the battery performance based on the first data, the battery performance, and the related information 310. This allows the user to easily understand which parts of the first data contribute to the battery performance.

[0055] (3-3) Third Example Figure 9 shows a third example of the processing performed by the battery performance estimation device 200. As the third example, an example of estimating battery performance from the second data based on related information 310 will be explained for each of the three phases (learning phase, estimation phase, and analysis phase). In this third example, the first data will not be used.

[0056] (3-3-1) Learning Phase The learning phase is a phase in which the learning unit 270 generates related information 310 by learning from training data. First, the data acquisition unit 230 acquires first data. The first data includes at least one of the surface image of the perovskite layer 174 and the X-ray diffraction waveform. The data acquisition unit 230 outputs the acquired first data to the data extraction unit 240.

[0057] The data extraction unit 240 extracts second data from the first data acquired by the data acquisition unit 230. The second data is data indicating parameters related to the crystal state of the perovskite layer 174. The data extraction unit 240 outputs the extracted second data to the learning unit 270.

[0058] Next, the learning unit 270 acquires second data from the data extraction unit 240. The learning unit 270 also acquires the battery performance of the perovskite solar cell 100, which has a perovskite layer 174 from which the first and second data have been acquired. The battery performance acquired here is not an estimated value, but the known battery performance of the perovskite solar cell 100. As mentioned above, the battery performance may be, for example, the current-voltage characteristics (I-V characteristics) of the perovskite solar cell 100, or the durability of the perovskite solar cell 100.

[0059] Next, the learning unit 270 generates related information 310 based on the second data and the battery performance. The related information 310 may be, for example, a trained model in which the second data and the battery performance are associated using a neural network. Specifically, the learning unit 270 generates the related information 310, which is a trained model, by learning training data in which the second data and the battery performance are associated. The learning unit 270 stores the generated related information 310 in the storage unit 300. The related information 310 generated here will be used in the battery performance estimation process during the estimation phase.

[0060] (3-3-2) Estimation Phase Next, the estimation phase will be explained. The estimation phase is the phase in which the estimation unit 260 estimates the battery performance of the perovskite solar cell 100. First, the related information acquisition unit 250 acquires related information 310 from the storage unit 300. As mentioned above, the related information 310 may be a trained model used to estimate the battery performance. The related information acquisition unit 250 outputs the acquired related information 310 to the estimation unit 260.

[0061] The data acquisition unit 230 acquires first data of the perovskite solar cell 100 whose battery performance is to be estimated. As described above, the first data includes a surface image of the perovskite layer 174 and an X-ray diffraction waveform. The data acquisition unit 230 outputs the acquired first data to the data extraction unit 240.

[0062] The data extraction unit 240 extracts second data for the perovskite solar cell 100, whose battery performance is to be estimated, from the first data acquired by the data acquisition unit 230. As described above, the second data is data indicating parameters related to the crystal state of the perovskite layer 174. The data extraction unit 240 outputs the extracted second data to the estimation unit 260.

[0063] Next, the estimation unit 260 acquires related information 310 from the related information acquisition unit 250 and second data from the data extraction unit 240. The estimation unit 260 also estimates the battery performance from the second data based on the related information 310. Specifically, the estimation unit 260 inputs the second data into the related information 310, which is a trained model, to obtain the battery performance output from the related information 310. As mentioned above, the battery performance may be, for example, the current-voltage characteristics (I-V characteristics) of the perovskite solar cell 100, or the durability of the perovskite solar cell 100. After that, the estimation unit 260 outputs the battery performance to the display unit 220.

[0064] Next, the display unit 220 displays the battery performance output from the estimation unit 260. This allows the battery performance estimation device 200 to inform the user of the battery performance of the perovskite solar cell 100 before the perovskite solar cell 100 is manufactured.

[0065] Furthermore, the estimation unit 260 outputs the second data, battery performance, and related information 310 to the analysis unit 290. This information will be used in the analysis phase.

[0066] (3-3-3) Analysis Phase Next, the analysis phase will be described. The analysis phase is the phase in which the analysis unit 290 analyzes the parts of the second data that contribute to battery performance. First, the analysis unit 290 obtains the second data, battery performance, and related information 310 from the estimation unit 260. Next, the analysis unit 290 analyzes the parts of the second data that contribute to battery performance based on the second data, battery performance, and related information 310. Specifically, the analysis unit 290 may perform the analysis processing using XAI methods such as GradCAM.

[0067] The analysis unit 290 outputs the analysis results to the display unit 220. As a result, the analysis results are displayed on the display unit 220. For example, the X-ray diffraction waveform IM3 and the analysis result image IM4 shown in Figure 7 may be displayed on the display unit 220 as analysis results.

[0068] As described above, in the third example of this embodiment, the data acquisition unit 230 acquires first data, the data extraction unit 240 extracts second data from the first data, the related information acquisition unit 250 acquires related information 310 from the storage unit 300, and the estimation unit 260 estimates the battery performance from the second data based on the related information 310. As a result, the battery performance estimation device 200 of this embodiment can easily and accurately estimate the battery performance of the perovskite solar cell 100.

[0069] Furthermore, in the third example of this embodiment, the analysis unit 290 analyzes the parts of the second data that contribute to the battery performance based on the second data, the battery performance, and the related information 310. This allows the user to easily understand which parts of the second data contribute to the battery performance.

[0070] (4) Figure 10 of another embodiment shows another example of the learning phase processing performed by the battery performance estimation device 200. In the learning phases of the first and second examples described above, the learning unit 270 acquires first data from the data acquisition unit 230, and in the learning phases of the first and third examples described above, the learning unit 270 acquires second data from the data extraction unit 240, but is not limited to this. For example, the learning unit 270 may acquire at least one of the first data and the second data from the determination unit 280.

[0071] To estimate the battery performance of the perovskite solar cell 100 with higher accuracy, it is effective to use only data with the same prototype conditions for the perovskite layer 174 as training data during the learning phase. The prototype conditions may include, for example, at least one of the following: the materials used in the perovskite layer 174 and their molar ratios, the solvent used, and the temperature conditions (not limited to the time of coating, but may also include the temperature of the coating film firing).

[0072] For example, the determination unit 280 may determine whether or not to use data associated with battery performance, at least one of the first data and second data, as training data, based on the prototype conditions of the perovskite layer 174. For example, the determination unit 280 may acquire the prototype conditions of the perovskite layer 174 and output at least one of the first data and second data to the learning unit 270 only if it matches the pre-set prototype conditions. Alternatively, the learning unit 270 may generate a trained model as related information 310 by learning the training data associated with battery performance, at least one of the first data and second data acquired from the determination unit 280. This allows the battery performance estimation device 200 to estimate the battery performance of the perovskite solar cell 100 with higher accuracy.

[0073] The first data is defined as at least one of the surface image and X-ray diffraction waveform of the perovskite layer 174, but is not limited to this. For example, the first data may further include a cross-sectional image, which is an image of the cross-section obtained by cutting the stacked structure of the perovskite solar cell 100 in the stacking direction. If the first data includes a cross-sectional image, the data extraction unit 240 may extract at least one of the thickness, flatness, and presence or absence of grain boundaries of each layer in the stacked structure as second data. As a result, the learning unit 270 can use the cross-sectional image of the stacked structure as training data to generate more accurate relevant information. Therefore, the battery performance estimation device 200 can estimate the battery performance of the perovskite solar cell 100 with greater accuracy.

[0074] As described above, the battery performance estimation device 200 of the embodiment includes a data acquisition unit 230, a data extraction unit 240, a related information acquisition unit 250, and an estimation unit 260. The data acquisition unit 230 acquires first data including at least one of a surface image, which is an image of the surface of the perovskite layer 174 in the perovskite solar cell 100, and an X-ray diffraction waveform obtained from the perovskite layer 174. The data extraction unit 240 extracts second data from the first data, which indicates parameters related to the crystalline state of the perovskite layer 174. The related information acquisition unit 250 acquires related information 310, which associates at least one of the first data and the second data with the battery performance of the perovskite solar cell 100. The estimation unit 260 estimates the battery performance from at least one of the first data and the second data based on the related information 310. As a result, the battery performance estimation device 200 of the embodiment can easily and accurately estimate the battery performance of the perovskite solar cell 100.

[0075] Furthermore, the related information 310 may be a trained model in which at least one of the first data and the second data is associated with battery performance using a neural network. The learning unit 270 may also generate a trained model by learning training data in which at least one of the first data and the second data is associated with battery performance. This makes it easy to learn even when complex data such as the surface image of the perovskite layer 174 or X-ray diffraction waveform is included in the first data.

[0076] Furthermore, the analysis unit 290 analyzes the portion of at least one of the first data and the second data that contributes to the battery performance, based on the battery performance and related information 310. This allows the user to easily understand which portion of at least one of the first data and the second data contributes to the battery performance.

[0077] In this example, the related information 310 is assumed to be a trained model that associates the first data, the second data, and battery performance using a neural network, but it is not limited to this. For example, as in the third example, if the learning unit 270 performs the learning process using the second data (parameters related to the crystal state of the perovskite layer 174) without using the first data, the related information 310 may be a regression line or regression curve calculated by regression analysis. Since the second data is not as complex as the first data, the processing speed in the learning process can be further improved by using regression analysis.

[0078] Furthermore, while the battery performance estimation device 200 is designed to perform both the learning phase and the estimation phase, it is not limited to this configuration. For example, the learning phase and the estimation phase may be performed by separate devices. Specifically, the estimation phase may be performed by the battery performance estimation device 200, while the learning phase may be performed by a device different from the battery performance estimation device 200.

[0079] Furthermore, the battery performance estimation device 200 described above can also be realized by using a general-purpose computer device as the basic hardware. That is, the data acquisition unit 230, data extraction unit 240, related information acquisition unit 250, estimation unit 260, learning unit 270, determination unit 280, and analysis unit 290 can be realized by having a processor mounted on the above-mentioned computer device execute a program. In this case, the battery performance estimation device 200 may be realized by pre-installing the above-mentioned program on the computer device, or by storing the above-mentioned program on a storage medium such as a CD-ROM, or by distributing the above-mentioned program via a network and appropriately installing this program on the computer device. In addition, the storage unit 300 can be realized by appropriately utilizing memory, hard disk, or storage medium such as CD-R, CD-RW, DVD-RAM, DVD-R, etc., which are built into or attached to the above-mentioned computer device.

[0080] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents.

[0081] 100... Perovskite solar cell 174... Perovskite layer 200... Battery performance estimation device 210... Input unit 220... Display unit 230... Data acquisition unit 240... Data extraction unit 250... Related information acquisition unit 260... Estimation unit 270... Learning unit 280... Judgment unit 290... Analysis unit 300... Memory unit 310... Related information

Claims

1. A battery performance estimation device comprising: a data acquisition unit that acquires first data including at least one of a surface image, which is an image of the surface of the perovskite layer in a perovskite solar cell, and an X-ray diffraction waveform obtained from the perovskite layer; a data extraction unit that extracts second data indicating parameters relating to the crystal state of the perovskite layer from the first data; a related information acquisition unit that acquires related information relating at least one of the first data and the second data to the battery performance of the perovskite solar cell; and an estimation unit that estimates the battery performance from at least one of the first data and the second data based on the related information.

2. The battery performance estimation device according to claim 1, wherein, if the first data includes the surface image, the data extraction unit extracts at least one of the number of crystals in the perovskite layer, the size of the crystals, and the amount of foreign matter as the second data.

3. The battery performance estimation device according to claim 1, wherein the first data further includes a cross-sectional image which is an image of a cross-section obtained by cutting the stacked structure of the perovskite solar cell in the stacking direction.

4. The battery performance estimation device according to claim 3, wherein, if the first data includes the cross-sectional image, the data extraction unit extracts at least one of the thickness, flatness, and presence or absence of grain boundaries of each layer in the laminated structure as the second data.

5. The battery performance estimation device according to claim 1, wherein, if the first data includes the X-ray diffraction waveform, the data extraction unit extracts at least one of the peak position, intensity, width, distortion due to fitting, and degree of crystallinity of the X-ray diffraction waveform as the second data.

6. The battery performance estimation device according to claim 1, wherein the related information is a trained model in which at least one of the first data and the second data is associated with the battery performance using a neural network.

7. The battery performance estimation device according to claim 6, further comprising a learning unit that generates the trained model by learning training data in which at least one of the first data and the second data is associated with the battery performance.

8. The battery performance estimation device according to claim 7, further comprising a determination unit that determines whether or not to use data relating at least one of the first data and the second data to the battery performance as training data, based on the prototype conditions of the perovskite layer.

9. The battery performance estimation device according to claim 1, further comprising an analysis unit that analyzes a portion of at least one of the first data and the second data that contributes to the battery performance, based on at least one of the first data and the second data, the battery performance, and the related information.

10. The battery performance estimation device according to claim 1, wherein the perovskite layer comprises tin (Sn) or lead (Pb).

11. The battery performance estimation device according to claim 1, wherein the crystalline phase of the perovskite layer contains the most abundant cubic (α) phase.

12. The battery performance estimation device according to claim 1, wherein, when the angle of incidence of X-rays on the surface of the perovskite layer is θ, the X-ray diffraction waveform includes waveforms in the range of 5° ≤ 2θ ≤ 45°.

13. A battery performance estimation method comprising: a battery performance estimation device acquiring first data including at least one of a surface image, which is an image of the surface of the perovskite layer in a perovskite solar cell, and an X-ray diffraction waveform obtained from the perovskite layer; extracting second data indicating parameters relating to the crystal state of the perovskite layer from the first data; acquiring related information relating at least one of the first data and the second data to the battery performance of the perovskite solar cell; and estimating the battery performance from at least one of the first data and the second data based on the related information.

14. A program that causes a battery performance estimation device to acquire first data including at least one of a surface image, which is an image of the surface of the perovskite layer in a perovskite solar cell, and an X-ray diffraction waveform obtained from the perovskite layer; extract second data indicating parameters related to the crystal state of the perovskite layer from the first data; acquire related information relating at least one of the first data and the second data to the battery performance of the perovskite solar cell; and estimate the battery performance from at least one of the first data and the second data based on the related information.

Images