| 郑玲春,何雨涵,李雯茜,任壕杰,向小凤,张书鸣.化学通报,2026,89(7):817-828. |
| 机器学习辅助氧还原反应设计催化剂应用研究进展 |
| Research Progress on the Application of Machine Learning in Oxygen Reduction Reaction Catalysts |
| 投稿时间：2026-02-25 修订日期：2026-03-30 |
| DOI： |
| 中文关键词: 机器学习 氧还原反应 电催化剂 描述符 材料筛选 |
| 英文关键词:Machine learning,Oxygen reduction reaction,electrocatalyst,descriptors,material screening |
| 基金项目:重庆市教育委员会科学技术研究计划项目(KJQN202401623)资助 |
|
| 摘要点击次数: 233 |
| 全文下载次数: 48 |
| 中文摘要: |
| 氧还原反应（Oxygen Reduction Reaction, ORR）是燃料电池等清洁能源技术的核心过程, 其效率受制于高性能电催化剂的开发。传统试错法与密度泛函理论计算存在效率低、成本高的瓶颈，机器学习（Machine Learning, ML）通过数据驱动方式, 能够高效挖掘催化剂结构、电子特征与ORR性能之间的构效关系, 实现快速预测与筛选。本文系统归纳了作为ML输入特征的几何、电子及活性描述符，以阐明它们在电催化剂设计中的一般规律。此外，文中详细阐述了ML在ORR催化剂设计中的具体应用，包括一般流程以及支持向量机（Support Vector Machine, SVM）、随机森林(Random Forest, RF)、梯度提升回归(Gradient Boosting Regression, GBR)和神经网络(Neural Network, NN)等常用ML算法的原理及实例, 体现了ML在降低计算成本、识别关键物理量、加速高通量筛选方面的优势。 目前该领域仍面临多尺度建模难、模型可解释性不足等挑战, 未来需进一步发展跨尺度模拟替代模型、并增强模型物理可解释性, 推动ORR催化剂设计向智能化、高效化发展。 |
| 英文摘要: |
| The oxygen reduction reaction (ORR) is a core process in clean energy technologies such as fuel cells, and its efficiency is constrained by the development of high-performance electrocatalysts. Traditional trial-and-error methods and density functional theory calculations are limited by low efficiency and high cost. Machine learning (ML), through a data-driven approach, can efficiently explore the structure-activity relationship between the catalyst"s geometry, electronic features, and ORR performance, enabling rapid prediction and screening. This review systematically summarizes the geometric, electronic, and activity descriptors used as input features for ML to clarify their general rules in the design of electrocatalysts. Additionally, it elaborates on the specific applications of ML in ORR catalyst design, including the general process and the principles and examples of commonly used ML algorithms such as support vector machine (SVM), random forest (RF), gradient boosting regression (GBR), and neural network (NN), highlighting the advantages of ML in reducing computational costs, identifying key physical quantities, and accelerating high-throughput screening. Currently, this field still faces challenges such as the difficulty of multi-scale modeling and insufficient model interpretability. In the future, it is necessary to further develop cross-scale simulation surrogate models and enhance the physical interpretability of models to promote the intelligent and efficient development of ORR catalyst design. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
京公网安备11010802047317号