Combined Machine Learning Techniques for Analyzing the Back Contact Influence on the Stability of Perovskite-Based Solar Cells

In the last ten years, perovskite-based solar cells (PSCs) have attracted a significant attention due to the fast rise in their power conversion efficiency (PCE). Nevertheless, the problem of the device stability remains a dilemma that is responsible for the solar cell degradation that needs to be investigated intensively. Therefore, in this work, based on a dataset containing 140 data points extracted from previous experimental works on PSCs, the influence of the back contact on the device stability by is analyzed applying the machine learning (ML) distribution of eXtreme Gradient Boosting (XGBoost). Besides, Multi-layer Perceptron Regressor (MLPRegressor) technique is used to analyze the effect of back contact alternative materials. The obtained results show a 15% influence on the total solar cell stability owing to the back contact. In terms of the most stable material, we found that Carbon is the best alternative according to our dataset. It was observed that these results align with many experimental works in the literature. Hence, we assume that the present work could be useful in the design of long-term stable PSCs.