Scholarlink

<p>In perovskite solar cells (PSCs), key quantities crucial for understanding physical processes, like electronic/ionic parameters, are difficult or impossible to measure directly. This study uses Autoencoders (AEs) to provide parameter values for different physical quantities of PSCs. The data set used for Machine Learning (ML) is simulated with a 1D drift-diffusion (DD) model, mimicking real-world devices. AEs are trained with the simulated data and, after learning, the encoder part of the AE is used for parameter estimation of measured PSCs. These estimates are retaken for new DD simulations that are compared to the measurements for validation, since the true device parameters are unknown. Furthermore, the decoder can be regarded as a device simulator restricted to the training data regime. The results show that AEs obtain device parameter estimates within seconds for the studied PSC devices. This procedure is applicable to gain a deeper understanding of device behaviour, for example, to study effects of degradation or different manufacturing processes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Autoencoder for parameter estimation and current-voltage curve simulation of perovskite solar cells

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月4日

查看全文
  • Oliver Zbinden,
  • Ennio Luigi Comi,
  • Evelyne Knapp,
  • Wolfgang Tress

摘要

In perovskite solar cells (PSCs), key quantities crucial for understanding physical processes, like electronic/ionic parameters, are difficult or impossible to measure directly. This study uses Autoencoders (AEs) to provide parameter values for different physical quantities of PSCs. The data set used for Machine Learning (ML) is simulated with a 1D drift-diffusion (DD) model, mimicking real-world devices. AEs are trained with the simulated data and, after learning, the encoder part of the AE is used for parameter estimation of measured PSCs. These estimates are retaken for new DD simulations that are compared to the measurements for validation, since the true device parameters are unknown. Furthermore, the decoder can be regarded as a device simulator restricted to the training data regime. The results show that AEs obtain device parameter estimates within seconds for the studied PSC devices. This procedure is applicable to gain a deeper understanding of device behaviour, for example, to study effects of degradation or different manufacturing processes.