基于数据挖掘与机器学习的重要能源转化过程电催化剂的设计与创新

燃料电池和金属空气电池是大规模能源转化与储存领域中极具前景的技术。电催化还原CO2、N2等生成燃料或高附加值化学品也是被寄予厚望的研究方向。上述研究的关键都在于高效电催化剂的研发。本项目中，我们首先利用第一性原理计算设计电催化剂并研究其催化性能，建立催化反应数据库。进而基于所建数据库进一步深入分析催化剂的化学组成、原子结构、电子结构等重要参数，并利用机器学习中神经网络和回归分析等算法，训练机器学习催化剂性质与催化活性之间的关系，获得可以准确描述催化活性的结构描述符，利用结构描述符快速评测催化活性，摆脱对高耗时量子化学计算的依赖，为快速大规模搜寻高效催化剂提供理论依据。

Fuel cells and metal-air batteries are promising technologies for large-scale energy conversion and storage. Electrocatalytic reduction of CO2 and N2 to fuels and chemicals with high added value is also a hopeful direction in this field. The key in the above research lies in the development of high-efficiency electrocatalysts. In this proposal we will first design novel electrocatalysts and investigate the corresponding catalytic performance by means of first-principles computations. Then we will establish a reaction database based on the results. By means of machine learning algorithms such as neural networks and regression analysis, with the chemical composition, atomic and electronic structure as the input, and the catalytic activity as output, the relationship between them can be obtained from the reaction database. The objective of this proposal is to obtain a reactivity descriptor and establish a quantitative relationship between the catalytic performance and the descriptor. Finally, the catalytic performance can be evaluated accurately and quickly, without conventional dependence on highly time-consuming quantum chemical computations. It is expected that this project could build the theoretical foundation for accelerating the development of outstanding electrocatalysts.