Machine Learning in Perovskite Solar Cells: Recent Developments and Future Perspectives

Abstract

Within a short period of time, perovskite solar cells (PSC) have attracted paramount research interests among the photovoltaic (PV) community. Usage of machine learning (ML) into PSC research is significantly accelerated their holistic understanding of device requisite properties. ML techniques are increasingly employed to discover stable perovskite materials, optimize device architecture and processing, and analyze PSC characterization data. This review provides an in‐depth exploration of ML applications in PSC advancement through an analysis of existing literature. The review commences with an introduction to the ML workflow, detailing each step, followed by concise overviews of perovskite materials and PSC operation. Later sections explore the diverse ways ML contributes to PSC development, which ranges from the optoelectronic property prediction of perovskites, discovery of novel perovskites, PSC device structure optimization, and comprehensive PSC analyses. The challenges impeding PSC commercialization are discussed, along with ML's potential to mitigate them. The review concludes by highlighting current limitations in employing ML for PSC research and suggests potential solutions. It also outlines prospective research directions for ML applications in PSC research, aiming to develop highly efficient and stable PSCs.

Machine Learning Guided Strategies to Develop High Efficiency Indoor Perovskite Solar Cells

“…ML is boosting PSC research by discovering new and stable perovskites, optimizing the device designs and the fabrication processes, property predictions, and in analyzing the device performance. [ 8 ] Li et al built two ML models utilizing ML algorithms like Linear Regression (LR), Support Vector Regression (SVR), K‐nearest Neighbors (KNN), Random Forest (RF), and Artificial Neural Network (ANN); one for bandgap prediction and another model for predicting the parameters relating to the performance of solar cells, including open circuit voltage ( V OC ), short circuit current ( J SC ), fill factor (FF), and power conversion efficiency (PCE). [ 9 ] The bandgap prediction model uses perovskite compositions as input features whereas bandgap, energy offset between the hole transport layer (HTL) and the highest occupied molecular orbital (HOMO) of perovskites, and energy offset between the lowest unoccupied molecular orbital (LUMO) perovskites and electron transport layer (ETL) were used as input features for predicting performance metrics.…”

Screening of Functional Small Molecules via Modified Machine Learning Strategy toward Efficient All-Inorganic Perovskite Solar Cells

“…Recently, many researchers began to use machine learning to process experimental data, which greatly saved the cost of experimental research time and was of great significance to the development of the perovskite field. 23 For instance, Liu et al used machine learning to screen interface materials for perovskite solar cells, which successfully improved the PCE and V OC of devices and accelerated the exploration of new materials for perovskite solar cells. 24 Zhi et al used the machine learning method to study the relationship between molecular characteristics of ammonium salt and PCE improvement of perovskite solar cells and selected interface modification materials according to the prediction of the model, successfully improved the device PCE, and proved the feasibility of machine learning in screening of functional molecules toward efficient perovskite solar cells.…”

“…In this regard, the machine learning strategy has strong data processing ability and can establish accurate prediction models, which plays an important role in assisting and guiding scientific research. − For perovskite modification, there are a large number of small molecules that can be used for interface modification, so the introduction of machine learning methods helps to screen suitable small molecules more quickly and accurately. Recently, many researchers began to use machine learning to process experimental data, which greatly saved the cost of experimental research time and was of great significance to the development of the perovskite field . For instance, Liu et al used machine learning to screen interface materials for perovskite solar cells, which successfully improved the PCE and V OC of devices and accelerated the exploration of new materials for perovskite solar cells .…”

Predicting photovoltaic parameters of perovskite solar cells using machine learning

“…ML techniques are increasingly used to optimize device structures and analyze the properties of PSCs. The optimization of device structure includes optimizing the perovskite layer, ETL, HTL, etc for better performance [21]. By determining the structure of the device, ML can quickly make fast and accurate predictions on the properties of PSCs.…”