Machine learning accelerates indoor stability analysis to predict long-term outdoor performance of perovskite solar cells

• Commercial development of perovskite solar cells (PSCs) has been significantly delayed by time-consuming degradation studies under real outdoor conditions. These are necessary steps to determine device lifetime, a problem PSCs have traditionally faced. Direct predictions of detailed time- and climatic-condition-dependent outdoor PSC performance, including daily power output and irreversible degradation, based on accelerated indoor testing are currently lacking.
• 

• Machine learning (ML) is a specialized field in artificial intelligence that focuses on developing algorithms to gain knowledge and make predictions or decisions by identifying patterns and insights in the data used for training. ML can be applied to the development of solar cells, including PSCs, for predicting material composition-property relationships, optimizing device structures and manufacturing processes, and reconstructing measurement data. There are few published works using ML tools to study factors affecting PSC stability.

  
  

  In this work, Ben-Gurion University Visoly-Fisher, Monica Lira-Cantu and Alessio Gagliardi et al. demonstrate that the outdoor degradation behavior of PSCs can be predicted by accelerated indoor stability analysis. Predictions are achieved using fast and accurate machine learning algorithms and mathematical decomposition processes. By training the algorithm using different indoor stability data sets, the most relevant stress factors are identified to reveal the degradation pathways of outdoor PV. This approach is not just specific to perovskite solar cells, but can equally be extended to other PV technologies where degradation and its mechanisms are key factors for widespread adoption.

  
  

• 
• 
• 
• 

• Ioannis Kouroudis, et al.Artificial Intelligence-Based, Wavelet-Aided Prediction of Long-Term Outdoor Performance of Perovskite Solar Cells. ACS Energy Letters 0, 9

  
  

  DOI: 10.1021/acsenergylett.4c00328

  
  

  https://doi.org/10.1021/acsenergylett.4c00328