作者：Xuefeng Zhao（赵雪峰）,Peng Deng, Guibin Shen*,You-hua Chen*（陈有华）

期刊：《Applied Energy》

出版时间：NOV 15 2026， Vol.423

校内级别：T2类

DOI: 10.1016/j.ijpe.2025.109888

Abstract

AI families, including machine learning, deep learning, and large language models, significantly drive solar cell (SC) industry upgrades. However, current research exhibits disciplinary fragmentation, typically focusing on single AI models within specific technologies like Silicon (SSC), Dye-sensitized (DSC), and Perovskite (PSC) cells. This creates a critical gap: the lack of comprehensive macro level framework explaining how the AI family propels innovation across distinct technological pathways. Therefore, our research constructs a universal innovation framework designed for analyzing the driving logic of the AI family on the development of SSC, DSC, and PSC. The framework leverages a large-scale dataset of patent documents and applies advanced Natural Language Processing techniques, including Named Entity Recognition and BERT-based models, to quantify AI's impact. The analysis is further synthesized through visualizations such as spherical networks and canopy graphs. Our research reveals that AI integration in the solar industry is not an arbitrary process but follows a structured trajectory governed by three core principles. First, a technology's maturity dictates the complexity of AI integration, with more mature technologies like SSC exhibiting more intricate AI applications. Second, AI tools are selected with high specificity to address the unique technical challenges of each cell type. Third, AI adoption follows a predictable sequence, progressing from foundational Machine Learning, to advanced Deep Learning, and finally toward cutting-edge Large Language Models. Significantly, the proposed framework successfully decodes predictable patterns in AI driven solar innovation, establishing a theoretical foundation for accelerating technological breakthroughs across the entire SC industrial chain. The three distinct AI and SC phenomena enable industry practitioners to optimize resource allocation, thereby enhancing the overall efficiency of future energy transitions.

Keywords: AI; Solar cell; Perovskite; Natural language processing

编辑：董晓玲

复审：周伟

终审：黄松