TECHNICAL FIELD OF THE INVENTION

The invention relates to a computer-implemented method for predicting a performance value of a perovskite-based solar cell based on electrical impedance spectroscopy carried out on the solar cell or parts thereof. Furthermore, the inven tion relates to a computer-implemented method for providing a trained function which involves training a function which is able to predict a performance value of a perovskite-based so lar cell based on electrical impedance spectroscopy carried out on the solar cell or parts thereof. Further, the inven tion relates to a corresponding providing system and a corre sponding computer-readable medium.

BACKGROUND OF THE INVENTION

Perovskite-based solar cells are a new, promising photovolta ic technology. During a relatively short time of development, cell efficiencies above 25% have been achieved on a laborato ry scale. This is competitive with regard to record efficien cies of conventional multicrystalline or monocrystalline sil icon solar cells, which are in a range of 23% and 26% on lab scale, respectively. For tandem setups comprising a perov skite-based solar cell and a silicon solar cell, a record photoconversion efficiency of 28% has already been demon strated.

Nonetheless, upscaling the manufacturing of perovskite-based solar cells and corresponding photovoltaic represent new challenges. In contrast to the widely available and well- established processes for the manufacturing of silicon-based solar cells, the manufacturing process of perovskite-based solar cells still needs to be optimized. For the manufactur ing of tandem solar cells even more manufacturing process steps have to be established, some of which are even not de veloped yet for large industrial scale.

Precise monitoring of perovskite-based solar cells during the sequential manufacturing process steps and analysis of devia tions in performance traced back to specific process steps would enable cost efficient process development and produc tion of perovskite-based solar cells.

Today, in order to judge the performance of perovskite-based solar cells quantitatively, the fully processed cell is char acterized in a controlled environment under exactly known il lumination. Concretely, an I(V)-curve (current-voltage-curve) of the cell is recorded at a solar simulator, which is a de vice that provides illumination approximating natural sun light. Subsequently, the main performance values of the test ed solar cell are deduced thereof.

However, this procedure requires significant invest, is time consuming and needs to be carried out at the fully processed solar cell. In other words, there is no established fast and reliable way to check the interim steps of the perovskite- based solar cell production chain. To evaluate the quality of the process, all preparation steps must be followed through and the characterization of the resulting cell performance can only take place once the solar cell has been finished. Typically, the 1-V measurements at full illumination are car ried out according to established standards such as IEC 60904-9. These measurements are used to evaluate the main performance parameters of the perovskite-based solar cell.

Thus, there exists the desire to develop a concept how to predict the performance of a perovskite-based solar cell in an alternative manner compared to the state of the art. Ide ally, the results may be used to monitor and optimize the processing of the solar cell faster and more directly.

DESCRIPTION OF THE INVENTION This objective is achieved by the subject-matters of the in dependent claims. Advantageous embodiments and variations are disclosed in the dependent claims and the description.

According to the invention, there is provided a computer- implemented method for providing output data, wherein the output data comprises a performance value of a solar cell and the solar cell comprises perovskite material. The method com prising the following steps:

- receiving input data, wherein the input data comprises data from electrical impedance spectroscopy carried out on the so lar cell or parts thereof,

- applying a trained function to the input data, wherein the output data is generated and

- providing the output data.

A key aspect of the present invention is the use of electri cal impedance spectroscopy to predict one or more performance values of the perovskite-based solar cell. Using electrical impedance spectroscopy is attractive, because it does not need to be carried out on a fully manufactured (in other words: fully processed) solar cell but can in principle also be carried out on intermediate cell material, i.e. before the completion of the processing of the cell. In other words, the electrical impedance spectroscopy measurements may be carried out on the fully processed or partly processed solar cell.

Electrical impedance spectroscopy (EIS) is a technique for evaluating the current response to the application of an al ternating voltage as a function of frequency. In particular, EIS measures the resistance and capacitance properties of a material via application of a sinusoidal alternating excita tion signal in the range of e.g. 2-10 mV. An impedance spec trum is obtained by varying the frequency over a predefined range. For instance, EIS is known to investigate the kinetics of electrical processes, including the clarification of rele- vant ionic and electronic processes that occur at different interfaces in various electrical devices.

However, EIS has never been used to predict a performance value of a solar cell so far, in particular of a perovskite- based solar cell.

EIS does not require a solar simulator, compared to conven tional I-V measurements. Instead, EIS measurements may be carried out at low light or even at no illumination at all.

The electrical impedance spectroscopy may for example be car ried out at frequencies in the range between 100 milli Hertz (mHz), i.e. 0.1 Hz, and 1 Mega Hertz (MZh), i.e. 1000 000 Hz.

In an advantageous embodiment of the invention, the electri cal impedance spectroscopy is carried out at a limited number of pre-defined frequencies. This has the advantage that the measurement time may be significantly reduced. Measurement times in the range between a few seconds down to smaller than 1 second may thus be achieved. This has the advantage that the EIS measurements may potentially be carried out inline during the manufacturing process of the perovskite-based so lar cell. Advantageously, the frequencies are chosen such that they are well suited to characterize the tested material and thus are able to form the basis of performance value pre dictions. Choosing the suitable frequencies is typically fa cilitated in practice, because, as normally the same cell type is manufactured along one production process, the best suited frequencies are known.

The EIS measurements are carried out on the solar cell or parts thereof. The solar cell does not need to be fully manu factured yet. For example, anti-reflection coating does not need to be applied yet. Furthermore, the front contact (e.g. transparent conducting oxide, TCO) does not need to be ap plied yet either. As a voltage needs to be applied to the ma- terial, some type of contact needs to be present, though. If the solar cell's contact does not exist yet, there can, for instance, be provided a small dot-like contact e.g. at the border, in particular at a corner, of the solar cell materi al. In principle, even the back contact (e.g. realized by en tirely covering the back side of the solar cell with alumi num) can be substituted by a test structure.

The purpose of the method according to the invention is to predict a performance value of a perovskite-based solar cell. A perovskite-based solar cell may also be denoted as a perov- skite solar cell, or PSC in short. In particular, the solar cell comprises a photosensitive layer and the photosensitive layer contains the perovskite material.

A perovskite-based solar cell is a type of solar cell which includes a perovskite structured compound, most commonly a hybrid organic-inorganic lead or tin halide-based material, as the light-harvesting active layer, i.e. the photosensitive layer. The general chemical structure of the perovskite is ABX ₃ and a common specification is (CH ₃ NH ₃₎ PbX ₃ .

As input data, the inventive method proposes to use data from electrical impedance spectroscopy carried out on the solar cell or parts thereof.

In one embodiment of the invention, the input data, i.e. the EIS data, comprises impedance, admittance, modulus function, complex dielectric constant and/or dielectric permittivity of the solar cell or parts thereof. These data may in principle be deduced from the Nyquist or Bode plots of the EIS measure ments.

In another embodiment of the invention, the input data are derived from modelling one or more electrical equivalent cir cuits based on the electrical impedance spectroscopy carried out on the solar cell or parts thereof. From the modelled electrical equivalent circuit(s), parameters such as the se- rial resistance, the capacity or the Warburg impedance of the solar cell or parts thereof can be deduced. These parameters may then be used as input data for the trained function which is used to predict the performance value of the solar cell.

In yet another embodiment of the invention, the input data further comprises at least one process parameter regarding the processing of the solar cell. In particular, the process parameter may comprise physical or chemical properties of at least one of the substances used for processing of the solar cell, such as the composition of the material or its viscosi ty. In other words, in addition to the input data gained from EIS measurements, process parameters from the processing of the solar cell which is tested may be used to predict one or several performance values of the solar cell.

The process parameters may also comprise deposition parame ters during a deposition process of elements of the solar cell. Here, the temperature, pressure, printing or spin coat ing parameters appear to be particularly valuable as input data.

Finally, also physical and chemical properties of the ele ments of the solar cell, such as the layer thickness, can be considered and added to the input data.

As output data, one or several performance values of the so lar cell are provided. In particular, the performance value are chosen from the group of the short circuit current, the open circuit voltage, the voltage of the solar cell at the maximum power point (which is also referred to by the maximum power point voltage), the current of the solar cell at the maximum power point (which is referred to by the maximum pow er point current), the maximum power (which is the product of the maximum power point voltage and the maximum power point current), the fill factor (which is the product of the short circuit current and the open circuit voltage, divided by the maximum power) and the cell efficiency, which is also re- ferred to as the photoconversion efficiency of the solar cell. These performance values are generally also referred to as "the main cell parameters" of a solar cell. A quantitative prediction of one or several of these parameters (or: values) give a concrete and quantitative estimation of the quality and performance of the solar cell.

According to the invention, an analytic function between the input data and the output data is not sought. Instead, a re lation, which is called a function in this patent applica tion, between the input data and the output data is set up. This function is trained such that it represents as best as possible the relation between input data and output data.

One approach to train the function is the application of mul tivariate statistics. Another approach is the use of neural networks, such as convolutional neural networks.

The function is trained according to the following method: Firstly, input training data is received, wherein the input data comprises exemplary data from electrical impedance spec troscopy applied to a perovskite-based solar cell or parts thereof. Secondly, output training data is received, wherein the output data comprises a performance value of the solar cell. The output training data is related to the input train ing data, which means that each output training data is asso ciated (in other words: related) to input training data. In a third step, a function is trained based on the input training data and the output training data. This results in the fourth and last step: the provision of a so-called trained function.

Descriptively speaking, by providing sets of input/output training data, a function is trained. The function is able, once that the training is completed, to predict output data given unknown input data with the help of the training.

Further, the invention is also directed to a providing sys tem, comprising a first interface, configured for receiving input data, wherein the input data comprises data from elec trical impedance spectroscopy carried out on a perovskite- based solar cell or parts thereof, a second interface, con figured for providing output data, wherein the output data comprises a performance value of the solar cell and a compu tation unit, configured for applying a trained function to the input data, wherein the output data is generated.

All embodiments and variants which have been discussed in context with the computer-implemented method also apply to the mentioned providing system. For the sake of conciseness, they are not repeated here expressis verbis.

Finally, the invention is also directed to a computer- readable medium comprising instructions which, when the pro gram is executed by a providing system, cause the providing system to carry out a method according to one of the embodi ments described above.

BRIEF DESCRIPTION OF THE DRAWINGS

In the following, the invention is illustrated by the help of the accompanying drawing, wherein

Fig. 1 shows Nyquist plots of three perovskite-based solar cells.

DETAILED DESCRIPTION OF THE DRAWINGS

Fig. 1 shows frequency-dependent real and imaginary part of the electrical impedance in a form known as Nyquist plot. The real part of the impedance in Ohm is plotted at the abscissa 10, the imaginary part of the impedance in Ohm is plotted at the ordinate 20. The first Nyquist plot 31 and the second Nyquist plot 32 result from electrical impedance spectroscopy measurements carried on a first type of perovskite-based so lar cells. They are two different cells, the first Nyquist plot 31 belonging to a first cell and the second Nyquist plot 32 belonging to a second cell, but both cells have been pro cessed by the same process. The third Nyquist plot 33 results from electrical impedance spectroscopy measurements carried on a third perovskite-based solar cell. The third solar cell has been processed by a distinctly different process. Thus, the third solar cell may be attributed to a second type of perovskite-based solar cells.

The first solar cell features a high performance, thus lead ing to a good selected performance value. The selected per formance value may e.g. be the photoconversion efficiency of the solar cell. The second solar cell features a low perfor mance, leading to a poor performance value, in this example a low photoconversion efficiency.

By comparing the first Nyquist plot 31 and the second Nyquist plot, it can be seen that significantly different Nyquist plots are obtained for two solar cells which belong to the same type of solar cell but show a different performance. In other words, the difference in the performance value (in this example, the cell efficiency) is already visible in the Nyquist plots obtained from EIS measurements. A key aspect of the invention is that a trained function would be able to recognize (and hence, predict) the poor performance of the second solar cell from the Nyquist plot only.

Fig. 1 shows, however, that care needs to be taken if EIS measurements from different cell types are compared. The third Nyquist plot 33 belongs to a different cell type, com pared to the first and second plots 31, 32. Thus, a different function applies when predicting the performance value(s) of the third solar cell, which needs to trained separately.

In essence, the present invention allows to predict the per formance of a perovskite-based solar cell from EIS measure ments, which otherwise would only be possible by sophisticat ed and costly I-V measurements at a solar simulator. The pre dicted performance can be valuable in several ways: The obtained result can be used to select and categorize the cells during manufacturing. The categorization is useful to build photovoltaic modules with matching cells in order to optimize the performances of the individual modules.

The obtained result may also be used to optimize the produc tion process itself. The measurements can be performed di rectly after an interim process step, i.e. before the pro cessing of the cell is completed. In contrast to many inves tigative tools, EIS is non-destructive (just contacting with two electrodes is required). Thus, on the one hand, the ob tained result may be used as a feedback to adjust a pro cessing machine in order to optimize the process for the next batch. On the other hand, the result may be used to as a feedforward to the next processing machine in order to com pensate deviations from the previous process step and thus "correct" the most recent batch. Particularly for tandem cells which require even more process steps (sometimes done at different manufacturers and shipped in between), EIS is a valuable tool for inspection of incoming pre-products before the processes at the next site start.

Yet another application of the present invention relates to the monitoring of a perovskite-based solar cell during its operation. As it is known, one challenge with perovskite- based solar cells relates to their potential degradation over time. The present method is also applicable to evaluate the degradation of the cell's performance over time independently of the illumination state. Although the aging mechanisms are generally known, the achievable stability during operation is not fully understood yet. For these reasons, perovskite-based solar cells may exhibit different degradation over time lead ing to a mismatch of cell performance in a module or in a power plant. This mismatch can be quantified using the de scribed method in order to estimate the overall loss in yield. The expected loss can be partially compensated by ad justing the operation of the solar cell considering the EIS measurement results, e.g. to avoid complete degradation or destruction of the cells.

In already operating photovoltaic sites the performance of cells or modules over time can be measured. When using EIS, those measurements do not require a solar simulator providing well defined illumination but could be performed whenever suitable (e.g. at nighttime) without interrupting the normal day-to-day operation. The acquired data can be analyzed and presented to stakeholders online and in real-time.

In summary, an attractive concept to predict one or more per formance values of a perovskite-based solar cell is present ed. This concept cannot only be used monitoring and optimiz- ing a production process of the solar cell, but also, for in stance, to monitor the performance of a perovskite-based so lar cell during its operation in the field.

LIST OF REFERENCE SIGNS

10 abscissa 20 ordinate 31 first Nyquist plot

32 second Nyquist plot

33 third Nyquist plot