论文标题
部分可观测时空混沌系统的无模型预测
Data-driven design of new catalytic materials in methane oxidation based on a site isolation concept
论文作者
论文摘要
储层计算是预测湍流的有力工具,其简单的架构具有处理大型系统的计算效率。然而,其实现通常需要完整的状态向量测量和系统非线性知识。我们使用非线性投影函数将系统测量扩展到高维空间,然后将其输入到储层中以获得预测。我们展示了这种储层计算网络在时空混沌系统上的应用,该系统模拟了湍流的若干特征。我们表明,使用径向基函数作为非线性投影器,即使只有部分观测并且不知道控制方程,也能稳健地捕捉复杂的系统非线性。最后,我们表明,当测量稀疏、不完整且带有噪声,甚至控制方程变得不准确时,我们的网络仍然可以产生相当准确的预测,从而为实际湍流系统的无模型预测铺平了道路。
The conversion of natural gas (methane) to ethane and ethylene (OCM: oxidative coupling of methane) facilitates its transportation and provides a way to synthesize higher value chemicals. The search for high-performance catalysts to achieve this conversion is the main scope of most corresponding studies in the field of OCM. Here, we present a general data-driven strategy for the search of novel catalytic materials, focusing particularly on materials useful for the OCM reaction. Our strategy is based on consistent experimental measurements and includes ab initio thermodynamics calculations and active screening. Based on our experiments, which showed unique volcano-type dependence of the performance on the stability of formed carbonates attributed to the site isolation concept, we developed a method for efficient and inexpensive DFT calculations of the formation energies of carbonates with prediction accuracy 0.2 eV. This method was implemented into a high-throughput screening scheme, which includes both general requirements for catalyst candidates and an actively done artificial intelligence part. Experimental validation of some of the candidates obtained during the screening showed successful reproduction of the initial volcano dependence. Moreover, several new materials were found to outperform standard OCM catalysts, specifically at lower temperatures.