Summary
Hybrid perovskite photovoltaics (PV) are considered one of the most promising emerging PV technologies that have the potential to becoming a vital part of future’s renewable energy production needed for combating climate change. However, the community still struggles with the scaling and reproducibility when printing perovskite thin-film absorber layers from solution due to the complex perovskite formation process that is susceptible to a variety of environment and process parameters.
This project proposes a novel interface between perovskite solution printing and algorithmic optimization and control theory by equipping a roll-to-roll perovskite printing line with in situ characterization, computational data processing and automatically adjustable process parameters. The characterization is given by point-probe reflection/absorption as well as luminescence imaging measurements. Feedback for control is calculated in real-time using simple control algorithms, at first, and progressing in complexity toward the employment of deep reinforcement learning (DRL), later on. The images are analyzed offline using convolutional neural networks for optimizing absorption/reflection set points for the above-described feedback control.
The proposal has two main goals: 1) Boosting the optimization of perovskite solution printing by effectively balancing exploration and exploitation of the large parameter space 2) demonstrating enhanced process control of certain system states to achieve higher process reproducibility and resilience in perovskite printing. These goals are complementary in a sense that control is exceeded on rapidly accessible parameters while optimization is performed with slowly accessible parameters.
In conclusion, this proposal presents a fundamental and concise new methodology of addressing reproducibility and scalability in perovskite solution printing that could be used for printing of functional thin-film, as well, highlighting its generality.
This project proposes a novel interface between perovskite solution printing and algorithmic optimization and control theory by equipping a roll-to-roll perovskite printing line with in situ characterization, computational data processing and automatically adjustable process parameters. The characterization is given by point-probe reflection/absorption as well as luminescence imaging measurements. Feedback for control is calculated in real-time using simple control algorithms, at first, and progressing in complexity toward the employment of deep reinforcement learning (DRL), later on. The images are analyzed offline using convolutional neural networks for optimizing absorption/reflection set points for the above-described feedback control.
The proposal has two main goals: 1) Boosting the optimization of perovskite solution printing by effectively balancing exploration and exploitation of the large parameter space 2) demonstrating enhanced process control of certain system states to achieve higher process reproducibility and resilience in perovskite printing. These goals are complementary in a sense that control is exceeded on rapidly accessible parameters while optimization is performed with slowly accessible parameters.
In conclusion, this proposal presents a fundamental and concise new methodology of addressing reproducibility and scalability in perovskite solution printing that could be used for printing of functional thin-film, as well, highlighting its generality.
Unfold all
/
Fold all
More information & hyperlinks
| Web resources: | https://cordis.europa.eu/project/id/101107885 |
| Start date: | 01-09-2023 |
| End date: | 31-08-2025 |
| Total budget - Public funding: | - 188 590,00 Euro |
Cordis data
Original description
Hybrid perovskite photovoltaics (PV) are considered one of the most promising emerging PV technologies that have the potential to becoming a vital part of future’s renewable energy production needed for combating climate change. However, the community still struggles with the scaling and reproducibility when printing perovskite thin-film absorber layers from solution due to the complex perovskite formation process that is susceptible to a variety of environment and process parameters.This project proposes a novel interface between perovskite solution printing and algorithmic optimization and control theory by equipping a roll-to-roll perovskite printing line with in situ characterization, computational data processing and automatically adjustable process parameters. The characterization is given by point-probe reflection/absorption as well as luminescence imaging measurements. Feedback for control is calculated in real-time using simple control algorithms, at first, and progressing in complexity toward the employment of deep reinforcement learning (DRL), later on. The images are analyzed offline using convolutional neural networks for optimizing absorption/reflection set points for the above-described feedback control.
The proposal has two main goals: 1) Boosting the optimization of perovskite solution printing by effectively balancing exploration and exploitation of the large parameter space 2) demonstrating enhanced process control of certain system states to achieve higher process reproducibility and resilience in perovskite printing. These goals are complementary in a sense that control is exceeded on rapidly accessible parameters while optimization is performed with slowly accessible parameters.
In conclusion, this proposal presents a fundamental and concise new methodology of addressing reproducibility and scalability in perovskite solution printing that could be used for printing of functional thin-film, as well, highlighting its generality.
Status
SIGNEDCall topic
HORIZON-MSCA-2022-PF-01-01Update Date
31-07-2023
Images
No images available.
Geographical location(s)
