摘要
本文概述了人工智能在电池电解液领域的应用现状,并详细描述了电池电解液特征值的构建方法及人工智能模型的选择策略。从分子结构和分子描述符提取两方面重点阐述了当前可行的特征值构建方法,对卷积神经网络、迁移学习等模型在分子设计中的优缺点进行对照分析。同时指出,测试数据集的获取是制约人工智能辅助开发电解液的主要难点之一。未来,在描述符、多形态融合以及数据扩充等方面进行更深层次的研究,将有助于人工智能应用于电解液的快速研究发展。
This paper outlines the current applications of artificial intelligence in the field of battery electrolytes.It describes the methods for constructing feature representations of battery electrolytes and the selection of artificial intelligence models.Current feasible feature value construction methods are focused on both molecular structure and molecular descriptor extraction.The advantages and disadvantages of models such as convolutional networks and transfer learning in molecular design are investigated.It is also pointed out that obtaining testing datasets is a challenging issue that restricts the development of artificial intelligence-assisted electrolyte research.In the future,deeper research in areas such as descriptors,multi-modal fusion,and data expansion will facilitate the rapid development of artificial intelligence applications in electrolyte research.
作者
蒋锦天
徐娇
张咪
孙翔
张炜
张荣鑫
朱伟伟
JIANG Jintian;XU Jiao;ZHANG Mi;SUN Xiang;ZHANG Wei;ZHANG Rongxin;ZHU Weiwei(Zhejiang Research Institute of Chemical Industry Co.,Ltd.,Hangzhou,Zhejiang 310023,China)
出处
《浙江化工》
2026年第1期7-14,共8页
Zhejiang Chemical Industry
关键词
人工智能
电解液
分子描述符
分子设计
artificial intelligence
electrolyte
molecular descriptor
molecular design
