加速钙钛矿太阳能电池开发：基于机器学习驱动框架的SHAP分析

打开文本图片集

关键词：钙钛矿太阳能电池；机器学习；夏普利加性解释（SHAP）分析；光电转换效率预测 中图分类号：TM914.4 文献标识码：A DOI：10.37188/CJL.20250149 CSTR：32170.14.CJL.20250149

Abstract：Perovskite solarcels （PSCs） have garnered significant attention in the realm of innovative photovoltaic technologiesdue to their impressiveperformance.Traditional trial-and-eror experimental methodsoftenresult in lengthy research cycles to enhance the power conversion eficiency （PCE）of PSCs.We propose a machine learning （ML）-based intelligent optimization strategy to accelerate research cycles in PSC fabrication.Byapplying various MLalgorithms to develop PCE prediction models，the gradient boosting（Gradientboosting，GB）model was chosen for Shapleyadditive explanations（SHAP）visualization analysis and experimental validation.Theexperimental resultsrevealed that the design guided by model predictions and SHAP analysisachievedaPCEof 21. 81% forwide bandgap（1.65eV）PSCs.This study effectivelyaddresses the limitations ofconventional trial-and-errorapproaches and overcomes the challengeoflow predictive accuracy in MLapplications withinthePSCdomain.It providesa new perspectiveandscientificbasis fortherapiddevelopmentofhigh-PCEPSCs，andalsoofersareferenceforthedevelopment of other new solar cell technologies.

Key Words：perovskite solar cells；machine learning；Shapley additive explanations（SHAP）analysis；power conversionefficiency（PCE）prediction

1引言

自2009年以来，钙钛矿太阳能电池（Perovskitesolarcells，PSCs）已成为新型光伏技术领域的重要研究方向。（剩余21169字）