基于材料基因工程的储氢材料数据库构建及其应用  

Materials genome engineering-based hydrogen storage materials database and its applications

作　　者：黄鹏儒 蔡丹 林怀周 刘佳溪 李子源 李彬 邹勇进[1] 褚海亮[1] 孙立贤[1] 徐芬[1] Pengru Huang;Dan Cai;Huaizhou Lin;Jiaxi Liu;Ziyuan Li;Bin Li;Yongjin Zou;Hailiang Chu;Lixian Sun;Fen Xu(Guangxi Key Laboratory of Information Materials,Guangxi Collaborative Innovation Centre of Structure and Property for New Energy and Materials,School of Material Science and Engineering,Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]广西电子信息材料构效关系重点实验室,广西新能源材料结构与性能协同创新中心,桂林电子科技大学材料科学与工程学院,桂林541004

出　　处：《中国科学：化学》2022年第10期1863-1870,共8页SCIENTIA SINICA Chimica

基　　金：国家重点基础研究发展计划(编号:2021YFB3802400,2018YFB1502105,2018YFB1502103);国家自然科学基金(编号:U20A20237,51863005,51462006,51102230,52101245,51871065,51971068,52161037);广西八桂学者基金;桂林市科学研究与技术开发计划(编号:20210102-4,20210216-1);中德国际合作项目(编号:GZ1528);广西信息材料重点实验室开放基金(编号:AA19182014,AD17195073,AA17202030-1);桂林电子科技大学研究生教育创新计划(编号:2022YCXS197,2022YCXS193)资助项目。

摘　　要：高密度储氢材料的加速研发对于我国能源经济转型、早日实现双碳目标至关重要.集成高通量计算、数据库及机器学习预测的数据驱动材料研发新范式有望缩短研发周期并降低研发成本.由于组分、结构、工艺及形貌等多重复杂性,目前储氢材料相关的数据驱动性能预测研究较少,尚缺乏一个较为系统的性质性能数据库.因此,本文中我们开发了智能化的数据挖掘引擎,通过已发表的学术论文中发掘储氢材料热力学、动力学储氢性能数据,以及现有的材料基因工程数据库数据中获取含氢材料物理化学性质,并结合高通量第一性原理计算数据,构建了储氢材料性质性能数据集.基于所构建的数据集进一步建立了储氢材料数据库,并应用晶体图形神经网络等机器学习方法对储氢材料的吸放氢质量、吸放氢温度进行预测.相关工作将数据驱动的材料研发新模式与储氢材料相结合,为发展实用高效的新型储氢材料提供有效的平台支持、数据支撑、方法指引.The accelerated research and development of high-density hydrogen-storage materials are critical to the transition of the energy economy and the realization of the carbon peaking and carbon neutrality goals.The data-driven research paradigm-integrated high-throughput calculations,database,and machine learning is appealing to accelerate new material development.Due to their complex compositions,structures and morphologies,the data-driven research on hydrogen-storage materials is still limited,and a database on the comprehensive properties is necessary to be established.Herein,we show our results on establishing a Materials Genome Initiative database and the property prediction based on machine learning for hydrogen-storage materials.The datasets are constructed from published papers,Materials Genome Initiative databases,and high-throughput first-principle calculations.A hydrogen-storage materials database was established and hydrogen property predictions are carried out by using machine-learning methods.We believe that the accelerated development of new hydrogen-storage materials would be benefited from our database and platforms.

关 键 词：储氢材料 热力学 动力学 数据库 材料设计 

分 类 号：TK91[动力工程及工程热物理] TP311.13[自动化与计算机技术—计算机软件与理论] TB34[自动化与计算机技术—计算机科学与技术]