In this work, we use surrogate models to accelerate the optimization of an adsorption process for H₂ purification and CO₂ capture within the context of fossil-based low-carbon H₂ production. A one dimensional column model was used to generate a training set for different feed compositions with up to four impurities and for varying process conditions. Subsequently, an artificial neural network (ANN) surrogate model was trained for six key performance indicators, achieving adjusted $R^2$ values over 0.999 for all indicators. Finally, the ANN was used for the constrained optimization of the H₂ separation performance and of the process performance (energy consumption and productivity), and the results were compared to full model optimization results. The agreement is very good for the H₂ separation performance, and in the case of low H₂ purity targets (99%) also for the process performance. For higher H₂ purities, however, the energy-productivity Pareto front features a very high sensitivity toward the H₂ purity constraint, and even small deviations in H₂ purity between full model and ANN surrogate model can translate to big deviations in the energy-productivity Pareto front. This is particularly pronounced for poorly sampled input regions at the edge of the sampling domain, for example for a binary H₂-CO₂ feed. If the ANN surrogate model is used at such edges of the sampling domain, additional sampling is required in this region to increase its accuracy. For all other cases, the deviations between the full model and the ANN surrogate model regarding the minimum energy consumption (or the maximum productivity) were well below 5%, showing that the ANN can be used with good accuracy instead of the full model.

在这项工作中，我们使用替代模型来加速优化基于化石的低碳 H ₂生产中 H ₂净化和 CO _{2捕获的吸附过程。}使用一维柱模型为具有多达四种杂质的不同进料组成和不同的工艺条件生成训练集。随后，针对六个关键性能指标训练了人工神经网络（ANN）代理模型，实现了调整后的$R^2$所有指标的值都超过 0.999。最后，将人工神经网络用于H ₂分离性能和工艺性能（能耗和生产率）的约束优化，并将结果与​​全模型优化结果进行比较。该协议对于 H ₂分离性能非常好，在低 H ₂纯度目标 (99%) 的情况下也适用于工艺性能。然而，对于更高的 H _{2纯度，能量生产率 Pareto 前沿对 H 2}纯度约束具有非常高的敏感性，甚至 H _{2的偏差很小}完整模型和人工神经网络代理模型之间的纯度可以转化为能量生产率帕累托前沿的大偏差。这对于在采样域边缘采样不佳的输入区域尤其明显，例如对于二元 H ₂ -CO ₂进料。如果在采样域的此类边缘使用 ANN 代理模型，则需要在该区域进行额外采样以提高其准确性。对于所有其他情况，完整模型和人工神经网络替代模型之间关于最小能耗（或最大生产力）的偏差远低于 5%，这表明人工神经网络可以以良好的准确度代替完整模型使用。