机器学习解析直接空气捕集用固体胺吸附剂的构效关系

周志斌 ,  张智渊 ,  邱雨晴 ,  董越 ,  赵国江 ,  曾垌皓 ,  吴晓宇 ,  郭本帅 ,  戴一阳 ,  周利 ,  刘冲 ,  代忠德 ,  吉旭

工程科学与技术 ›› 2025, Vol. 57 ›› Issue (05) : 79 -90.

PDF (2776KB)
工程科学与技术 ›› 2025, Vol. 57 ›› Issue (05) : 79 -90. DOI: 10.12454/j.jsuese.202400007
碳中和与清洁能源

机器学习解析直接空气捕集用固体胺吸附剂的构效关系

作者信息 +

Machine Learning Analysis of the Structure‒Property Relationship of Amine-based Solid Adsorbents for Direct Air Capture

Author information +
文章历史 +
PDF (2841K)

摘要

当前,直接空气捕集(DAC)技术蓬勃发展,已成为高效的负碳技术,也是实现二氧化碳净零排放的关键路径。高性能的吸附剂材料是保障DAC技术产业化部署的核心要素,其中固体胺吸附剂因性能优异备受关注。通过对全球固体胺吸附剂的CO2吸附实验数据进行系统性挖掘和整合,本文运用随机森林(RF)机器学习算法建立了吸附剂的24个特征描述符(涉及基底组分、孔隙性质、有机胺属性及实验条件等)与其CO2吸附容量之间的预测模型,获得了训练集决定系数R2=0.964、测试集R2=0.823的性能。单变量分析结果表明,固体胺吸附剂的CO2吸附容量受到多重因素影响,24个特征描述符与CO2吸附容量之间表现出明显的非线性关系。因此,本文进一步运用SHAP分析方法定量解析了机器学习模型的预测结果,揭示了决定固体胺吸附剂CO2吸附容量的关键参数。其中,有机胺负载量对吸附剂CO2吸附量的正向贡献度最高,可达0.46 mmol/g,其他重要的特征描述符还包括:测试所用CO2浓度、实验温度、吸附剂的BET表面积和孔隙率、基底的元素组成(如O、C、F)等。针对新型 DAC 固体胺吸附剂的开发,本文建议选择具有高孔隙率的杂化材料作为基底,并采用低聚合度的有机胺且实现其高负载量;同时,对吸附工艺也提出了一些建议,如采用更低的操作温度等。本研究聚焦新型DAC固体胺吸附剂的开发,有利于加快DAC负碳技术的开发与部署,助力“双碳”目标的顺利实现。

Abstract

Objective Direct air capture (DAC) emerges as a promising negative emission technology to mitigate global warming. The performance of direct air capture adsorbents, particularly amine-based solid porous adsorbents, plays a critical role in the industrial deployment of DAC. Extensive experimental studies are conducted worldwide to investigate the CO2 capture capabilities of these adsorbents, generating a substantial volume of data. However, the CO2 capture performance is influenced by multiple parameters, requiring a systematic and comprehensive analysis to clarify the relationship between the structural and conditional parameters of amine-based adsorbents and their CO2 capture capabilities. Methods This study systematically compiled experimental data on amine-based solid DAC adsorbents from peer-reviewed scientific articles. Four different types of machine learning algorithms (random forest, artificial neural network, support vector machine, and ridge regression) were employed to construct a predictive model that correlated the features of amine-based solid adsorbents with their CO2 adsorption capacity values. In addition, the study utilized shapley additive explanations (SHAP) analysis to deconstruct the machine learning model’s predictive process, quantitatively revealing key parameters that determined the CO2 adsorption capacity of the adsorbents. Results and Discussions This study collected 629 valid data entries from 32 scientific publications, covering a wide range of CO2 capture capacities from 0 mmol/g to 5.0 mmol/g, to guide the design of new DAC adsorbents with enhanced CO2 capture performance. Each data entry was characterized by 24 descriptors, which encompassed information on the porous substrate components, textural properties, amine properties, and experimental conditions. An individual-variable analysis using the Pearson method revealed little linear correlation between the descriptors and CO2 capture capability, except for the amine loading in the adsorbents, with a Pearson correlation coefficient R = 0.543. Machine learning models were employed to uncover potential nonlinear and multivariable relationships. Four algorithms with good fitting capabilities and robustness against information noise were selected to build predictive models for the CO2 capacity of amine-based adsorbents, namely artificial neural network (ANN), support vector machine (SVM), ridge regression (Ridge), and random forest (RF). The dataset was split into training and test sets in a 4:1 ratio, and hyperparameters were optimized using grid search and validated with 5-fold cross-validation. After the optimization of hyperparameters, the RF model showed superior performance compared to the other selected models. The optimal RF model demonstrated the best performance in predicting CO2 adsorption capacity, with R2 = 0.823, MAE= 0.270, and RMSE= 0.372 in the test set. The performance of the RF model indicated that the 24 descriptors effectively covered the key factors that determined the CO2 capacity of amine-based adsorbents within the current experimental design space. A quantitative structure-property relationship (QSPR) analysis was conducted using the SHAP analysis method based on the mentioned reliable RF model. In this case, the SHAP method quantified the contribution of each descriptor of the DAC data entry to the output (predicted CO2 adsorption capacity) of the RF model, providing interpretability and insights into the model's decision-making process. The SHAP analysis results identified the most important descriptor influencing CO2 capacity, the amine loading, which exhibited a strong positive correlation. Other significant descriptors included the molecular weight of the incorporated amines (negative correlation), CO2 concentration (positive), porosity of the adsorbent (positive), and elemental contents in the substrate material, such as O (negative), C (positive), and F (positive). Accordingly, three experimental design strategies were proposed for further exploration of amine-based DAC adsorbents: 1) utilize substrates with high porosity, 2) avoid using amine-based polymers with excessively high molecular weight (> 10 000), and 3) select substrates containing C or F elements. These strategies contributed to the further enhancement of the CO2 adsorption performance of amine-based solid porous adsorbents and accelerated the development and deployment of DAC as a key negative emission technology. At the same time, the current limitations of chemical diversity and experimental space highlighted several areas and directions that required further research. For example, additional experimental studies were needed to investigate the CO2 capacity of amine-based adsorbents in humid and low-temperature environments, as well as the impacts of gas flow rate and gas mixture composition on the adsorbents’ CO2 capacities. In addition, an ideal DAC adsorbent should have met multiple criteria beyond CO2 capacity, such as efficient mass transfer and high stability in adsorption-desorption cycles. Currently available experimental data are still insufficient to address these aspects using machine learning strategies. Conclusions This research highlights the potential of machine learning in analyzing large-scale datasets to identify factors that influence the CO2 capture performance of DAC adsorbents. Critical factors were identified, including amine loading and adsorbent porosity. The findings provide insights that can guide the design of more effective DAC adsorbents and highlight areas requiring additional experimental research, particularly regarding the effects of environmental conditions and gas composition on adsorbent performance. This study contributes to the progress of DAC technologies as a feasible solution to achieving China's carbon dioxide peaking and carbon neutrality objectives by advancing the understanding of amine-based adsorbents.

Graphical abstract

关键词

二氧化碳 / 直接空气捕集 / 固体胺吸附剂 / 机器学习 / 碳捕集

Key words

CO2 / direct air capture / amine-based solid adsorbent / machine learning / carbon capture

引用本文

引用格式 ▾
周志斌,张智渊,邱雨晴,董越,赵国江,曾垌皓,吴晓宇,郭本帅,戴一阳,周利,刘冲,代忠德,吉旭. 机器学习解析直接空气捕集用固体胺吸附剂的构效关系[J]. 工程科学与技术, 2025, 57(05): 79-90 DOI:10.12454/j.jsuese.202400007

登录浏览全文

4963

注册一个新账户 忘记密码

本刊网刊
针对2015年《巴黎协定》关于全球升温控制在1.5 ℃以内的明确而严格的目标,联合国倡导在全球范围内实现净零排放。2023年11月20日联合国环境规划署发布的《2023年排放差距报告》[1]显示,全球的温室气体排放量在2021~2022年间增加了1.2%,达到创纪录的5.74×1010 t CO2当量。因此,各国必须采取比《巴黎协定》中更为强硬的减排措施,才可能避免导致2030年全球变暖2.5~2.9 ℃的极端后果出现。该报告同时指出,全球的碳排放量到2030年必须下降28%~42%,才有可能实现《巴黎协定》中提出的全球升温可控目标[1]。因此,除碳减排之外,净零排放的实现还必须依赖高效和大规模的负碳技术[2],这使得相关的科学研究和技术开发非常活跃,其中的代表性方案之一是直接空气捕集(DAC)技术[3]。基于《巴黎协定》1.5 ℃目标而建立的多种气候变化综合评估模型在充分考查和比较各种负碳技术的基础上总结了DAC的优势[34],主要包括:1)其对土地和水资源的需求很小,尤其是相对于生物质能‒碳捕集与封存技术而言[5];2)大规模和高效的DAC部署将使得基于传统化石燃料的工业部门得以继续存在,增加现有经济体系的缓冲能力[6];3)DAC对于电力、交通乃至油气等领域的深度脱碳是不可或缺的负碳技术,在多种模型中都是实现净零排放的关键[78]。因此,DAC技术的开发对于中国实现“双碳”目标具有至关重要的意义[911]
从基础研究层面来看,开发新的固体吸附剂是DAC技术研发的重点之一。当前用于DAC的主流吸附剂包括沸石、高分子材料和金属有机框架材料(MOFs)等。在众多备选的固体吸附剂中,一类负载有机胺的多孔固体吸附剂(简称固体胺吸附剂)表现出优良的性能,在科学界与工业界都受到了广泛关注。例如:Kwon等[12]用二氧化硅负载的聚乙烯亚胺(PEI)在30 ℃、400 ppm和19%相对湿度的条件下实现了3.36 mmol/g的优异CO2吸附量;Darunte等[13]用MOFs材料结合有机小分子三氨乙基胺(TREN)对25 ℃、400 ppm的干燥CO2实现了2.76 mmol/g的吸附量。
近期发表的一些综述对固体胺DAC吸附剂的研究现状、面临的挑战与下一步的研究策略进行了总结与论述[2,14]。总的来看,当前已用于或潜在可用于DAC的固体胺吸附材料种类繁多,它们的可调参数也有很多,因此尚未建立起具备高度系统性的研究策略。为使材料开发具有更高的效率和目的性,亟需系统地建立智能化筛选和评价机制。虽然运用人工智能工具推进此类DAC吸附剂开发的成果目前少见报道,但研究人员已经基于大数据和机器学习等方法对CO2及其他气体吸附剂开展了相关的筛选和评价工作。例如,Boyd等[15]完成了数据驱动的面向烟道气CO2捕集的MOFs材料设计与开发。Avci等[16]也基于机器学习和分子模拟,开展了面向CO2捕集和H2纯化的高通量MOFs吸附剂(及膜)筛选研究。Leperi等[17]构建了神经网络模型来模拟和优化工业尺度上的CO2变压吸附过程。国内的人工智能辅助吸附剂设计与开发也取得了一系列重要的成果。例如,Yuan等[18]运用机器学习工具预测了生物质衍生的多孔碳材料用于CO2吸附的性能。Situ等[19]面向MOFs材料用于烟道气CO2捕集开展了大规模的机器学习筛选。本团队也已开展了一系列基于数据科学的大规模多孔材料筛选与人工智能辅助设计工作,并用于多种分离和纯化的应用场景[2025]。这些研究在人工智能与化学、化工结合方面具有显著的创新性,也为开展DAC吸附剂的系统筛选和评估奠定了基础。
固体胺吸附剂的传统试错法研究高度依赖研究者的个人经验,难以对材料的参数空间进行全面、高效的探索和分析。同时,传统研究方案往往缺乏对材料构效关系的定量评估,从而降低了研究迭代的效率,提高了实验试错的成本。因此,本文试图建立机器学习辅助DAC吸附剂材料的评估与筛选策略,实现对数据的系统化分析并设计合理的特征描述符组合,确保模型的可靠性和泛化能力。本文对近年来固体胺吸附剂相关的实验研究进行了系统总结,运用机器学习算法基于吸附剂相关特征描述符构建了其CO2吸附容量的预测模型,并揭示了吸附剂特征描述符与其CO2吸附容量之间的构效关系。

1 方 法

1.1 皮尔森分析

吸附剂特征描述符与CO2吸附容量之间的皮尔森相关性系数R计算公式如下[26]

R=nxy-xy(nx2-x2)(ny2-y2)

式中:x为24个特征描述符[28]中的任意一个;y为实验测得的CO2吸附容量;n为全部的实验样本数,即629。

1.2 特征数据归一化

依照式(2)将每个特征描述符的数值等比缩放到[0,1]的区间内:

x˜ij=xij-minxjmaxxj-minxj

式中,ij分别为数据样本编号和特征编号,xij 为样本i的第j特征值,x˜ij为归一化后的xij,min(xj )和max(xj )分别为全部样本中第j特征的最小值和最大值。

1.3 机器学习与超参数优化

机器学习的过程(包括数据集的抽样和拆分、五重交叉验证、模型的选择和训练、模型性能的评估)使用scikit-learn Python库[27]完成。在文中被提及的模型都进行了超参数优化,具体模型的超参数名称和超参数可能的选项(即超参数空间)列于表1。超参数的搜索方法为网格搜索。特定超参数下的模型性能使用五重交叉验证来评估:训练集被随机5等分,每次验证轮换地用其中4份数据集训练一个平行验证模型,剩余1份作为验证集验证该模型的泛化能力。超参数的优化目标为最大化模型在五重交叉验证时的平均决定系数。

1.4 模型性能指标

文中出现的各个性能指标的计算方法如下。

决定系数R2

R2=1-n2p-y2ny-y2

平均绝对误差EMA

EMA=1np-y

均方根误差ERMS

ERMS=1np-y2

式(3)~(5)中,p为机器学习模型的预测值,n=629。

1.5 SHAP分析

对于样本i的第j个特征描述符的SHAP值ϕij,计算公式[28]如下:

  ϕij=SS!M-S-1!M!μpy|xiSxij-μpy|xiS

式中,S为任意一个不包含j的特征描述符子集,S为子集S中的特征描述符数量,M为总的特征描述符数量(本文中M=24),μp为被解析模型的CO2吸附容量的预测期望,xiS为样本i对于特征描述符子集S的特征向量值。

2 结果与讨论

2.1 数据集与特征描述符

本文对近年来发表在学术期刊中的固体胺DAC吸附剂相关的学术论文进行了系统性整理,收集了802条包括吸附剂CO2吸附容量的实验记录和相应的信息(基底组分、孔隙性质、有机胺属性、吸附实验条件),并对其进行了特征化处理,表示为24个特征描述符,列于表2。802条实验记录中,有173条由于缺失表2所列特征描述符中的一项或数项而从最终数据集中删除;剩余629条特征描述符信息完整的实验记录来自32篇研究论文[2960],涵盖16种基底多孔材料(如多孔二氧化硅、沸石、MOFs、多孔有机硅烷、多孔碳等)及18种有机胺(如PEI、聚丙烯亚胺PPI、四乙撑五胺TEPA、聚烯丙胺PAA、3-氨丙基甲基二乙氧基硅烷APDES等),还包括多个变化的实验参数,如温度、CO2浓度和相对湿度。

629个DAC实验数据样本的CO2吸附容量分布直方图如图1所示,CO2吸附容量分布在0~5.0 mmol/g的区间内。

2.2 单变量分析

本文首先对实验数据样本进行了单变量分析,以考察提取的特征描述符与固体胺吸附剂的CO2吸附容量之间的相关性,用皮尔森相关性系数(R)来表征。基于计算结果,特征描述符与CO2吸附容量之间的相关性强弱与方向可以被定量判断,如表3所示。

24个特征描述符与CO2吸附容量之间的皮尔森回归结果以及2维散点分布如图2所示。由图2表3可知,有机胺负载量与CO2吸附容量之间呈现出较强相关性(R=0.543),其余的特征描述符与CO2吸附容量的相关性不强,其中中度相关2个,弱相关21个。这一结果可能是因为固体胺吸附剂的CO2吸附容量并不由单一因素所决定。想要做出准确的判断并量化不同参数与吸附量之间的相关性,需要进一步运用机器学习、SHAP分析等方法。

2.3 机器学习

为揭示固体胺吸附剂24个特征描述符与CO2吸附容量之间的多元相关性,本文选用一组多样化的机器学习算法提升建模的成功率和可靠性。被选用的机器学习模型应当满足以下条件:1)具有不同的拟合方法,以尽可能提升建模过程的成功率;2)具有较强的拟合能力,能够处理复杂高维的数据关系;3)面对实验测量和记录过程中难免存在的偏差等问题,算法应当具有良好的鲁棒性,对噪声和异常值具有一定的容忍度,能够在复杂的数据环境中获得稳定的预测结果;4)尽可能具有良好的可解释性,以方便后续的构效关系解析。因此,本文最终选择了RF[61]、ANN[62]、SVM[63]、Ridge[64]等4种模型来尝试建模。

在模型训练时,本文将629条数据以8∶2的比例拆分为训练集和测试集。其中,训练集用于模型训练,测试集用于模型泛化能力的评估。24个输入特征描述符具有不同的数值分布区间,可能在模型训练过程中带来不利影响,如扭曲ANN的梯度表面、误导SVM对于距离的感知、使Ridge对不同特征的正则化效果不平衡等,为消除特征数值分布差异带来的干扰,在模型训练前对所有特征值都进行了归一化处理,方法见第1.2节。

按照第1.3节描述的方法,对所有模型的超参数都进行了调节与优化。基于最优的超参数(表1中用下划线标注),使用训练集中的全部数据样本针对每个机器学习算法训练了一个最优模型,4个最优模型的性能测试结果如图3所示。

图3可知,RF模型性能表现优于其他3个模型,无论是在验证集(五重交叉验证R2值分别为0.842、0.775、0.839、0.855、0.816,平均值为0.825,方差为0.040),还是测试集上都具有不错的预测精度和稳定性(性能指标汇总结果见表4)。其他3个机器学习模型可能由于各自的算法特性不适于解决本问题,因而性能表现较差。例如,Ridge主要用于处理线性问题;SVM尽管可以处理非线性问题,但是总体而言更加倾向于处理线性问题。而本文中固体胺吸附剂的特征描述符与CO2吸附容量之间的关系表现出较为明显的非线性特征。ANN具有较强的非线性拟合与特征提取能力,但对大数据样本依赖性较强,而目前通过文献挖掘能够获得的数据集仍较小。

RF模型的优异表现也证明本文初始选择的特征描述符集合能够基本覆盖决定固体胺吸附剂CO2吸附容量的关键参数。尤其需要指出的是,如图1所示,高性能吸附剂(吸附容量>4.0 mmol/g)在原始训练样本中的数量很少(只占约1.43%),因此对高CO2吸附容量的固体胺吸附剂进行可靠预测具有挑战性。但RF模型仍能对测试集中吸附剂的CO2吸附容量进行准确的预测,这表明基于现有特征描述符集合的RF模型可以实现一定程度的外推。

2.4 构效关系解析

基于图3(a)的最优机器学习模型,本文使用SHAP分析方法来揭示不同特征描述符对于固体胺吸附剂CO2吸附容量的贡献值。SHAP值的计算过程见第1.5节。根据SHAP分析的结果,将影响固体胺吸附剂CO2吸附容量最重要的9个特征描述符以及它们的平均ϕij展示于图4(a)中。同时,在图4(b)中以蜂群图的形式展示了最重要的9个特征描述符在所有样本中的ϕij分布,按式(2)归一化的特征值x˜ij的大小用颜色变化(从大到小对应从红到蓝)进行可视化。SHAP值ϕij代表在第i个吸附样本中,特征描述符j的值xij使得模型预测的CO2吸附容量平均改变了ϕij mmol/g。ϕij的绝对值越大表示其对于模型决策的影响力越大,重要度越高。ϕij>0代表特征取值xij在样本i中起到正向作用;ϕij<0代表特征取值xij在样本i中起到负向作用。

因此,由图4(a)可见,有机胺的负载量对于固体胺吸附剂的CO2吸附容量影响最大,平均ϕij达到0.46 mmol/g,这与之前单变量分析的结果相一致。其他几个有较大影响的特征描述符分别为:吸附实验中CO2的浓度(ϕij=0.20 mmol/g)、负载BET表面积(ϕij=0.14 mmol/g)和负载有机胺的平均分子量(ϕij=0.12 mmol/g)。由图4(b)可以判定各个特征描述符对于CO2吸附容量的贡献方向。有机胺负载量与实验中CO2浓度都有明显的正向贡献,ϕij高值都来自于高的有机胺负载量与高的CO2实验浓度。NMw对于CO2吸附容量的贡献方向较为模糊,但大概率为负向贡献,这是因为高ϕij的样本全部归属于低的NMw值,而高NMw值样本的ϕij几乎全为负。排名前9位的其他特征描述符中,负载BET表面积和负载孔隙率PVP都为负向贡献;实验温度T为明显的负向贡献;基底材料的组成也对固体胺吸附剂CO2吸附容量有一定的影响,其中CO为负向贡献,CCCF为正向贡献。

2.5 固体胺吸附剂设计策略

构效关系解析是对过往大量实验数据的分析与总结。在开发高CO2吸附容量的固体胺吸附剂时,研究人员可以将其作为实验设计的参考,强化有利因素并尽量避开不利因素,同时探索更多未知参数区域。基于第2.4节分析得到的构效关系,本文提出如下的实验策略:

1)使用高孔隙率的多孔材料作为基底。无论单变量分析还是SHAP分析的结果都表明,高有机胺负载量是固体胺吸附剂获得高CO2吸附容量的关键。为了使多孔吸附剂能够负载更多有机胺并保持其孔道的可及性,应该使用具有更高孔隙率的基底材料。另外,还可以考虑制造同时具有微孔和介孔/大孔的多级孔基底材料,用介孔或大孔负载更多的有机胺,用微孔保证CO2分子通过。例如,Cogswell等[65]采用具有多级孔结构和高孔隙率的3维硅基材料作为基底,成功实现了有机胺的更高负载并保持了孔道的畅通。

2)避免使用过高分子量的有机胺。高分子量的有机胺对于CO2吸附能力具有不利影响,这在实验研究中难以被直观发现,本文所做的SHAP分析从大量实验数据中揭示了这一规律。高分子量有机胺相对而言不易在多孔基底材料中流动和分散,从而降低了孔道堵塞的风险,避免CO2的传质阻力增大带来的吸附容量下降。根据SHAP分析得到的结果,对CO2吸附容量具有正向贡献的有机胺分子量NMw往往小于10 000 Da。同时,过低分子量的有机胺容易挥发,导致其在吸附剂再生过程中损失。因此,有机胺的分子量应该控制在一个比较适当的范围内,如根据Heydari‒Gorji等[66]的研究,为避免PEI的蒸发损失,其NMw至少应该大于400 Da。

3)在满足上述条件的前提下,考虑使用含有C与F的杂化材料作为基底,避免使用单纯的多孔氧化物。从SHAP分析所得到的构效关系可知,基底中包含C与F等元素是提高吸附剂CO2吸附容量的有利因素,而O含量过高的基底材料会削弱其CO2吸附容量。这表明使用纯SiO2或Al2O3的基底材料对于提高CO2吸附容量而言可能并非最优选择,可以考虑采用金属有机框架[67]、共价有机框架、多孔有机硅等基底材料,或者对SiO2和Al2O3等材料进行表面改性来提升固体胺吸附剂的CO2吸附容量。例如,Liu等[68]通过合成改性的SBA‒15,在提供了更大的有机胺负载空间和CO2扩散通道的同时改善了有机胺的分散性,因此获得了更大的CO2吸附容量和更短的平衡时间。

2.6 讨论

当前固体胺DAC吸附剂的实验研究主要集中于在室温、干空气、静态、标准气体组分(CO2/N2)条件下的CO2吸附容量研究。而从DAC实际部署的情况来看,低于室温、潮湿条件下的吸附实验可能更具价值。由SHAP分析结果可知,降低温度有助于提高固体胺吸附剂的CO2吸附性能。但这一结论主要来自室温和高于室温条件下实验结果的外推,在更低温度下固体胺CO2吸附剂的吸附性能及变化趋势研究需要更多的实验数据来支撑。

湿度对于固体胺吸附剂的CO2吸附容量有着显著的影响[69]。例如,Sehaqui等[70]报道,在其他条件相同的情况下,PEI负载的纤维素纳米纤维(NFC)薄片在80%相对湿度的湿空气中的CO2吸附量提高了4.8倍。另一方面,湿度也有可能削弱吸附剂对CO2的吸附容量,如Goeppert等[71]报道,负载了PEI的气相SiO2薄片在67%相对湿度的湿空气中相比于干空气中的CO2吸附容量下降了约17%。由此可见,湿度对吸附剂CO2吸附容量的影响比较复杂。从当前有限的研究报道来看,湿度对CO2吸附的有利因素可能包括:1)水分子参与胺与CO2的反应提高了胺的利用率[70];2)水分强化了CO2在部分有机胺相中的传质[72];3)水分促进有机胺相的流动和表面更新[73]。而湿度对CO2吸附的不利因素可能包括:1)水分子与CO2分子发生竞争吸附[71];2)水分子在有机胺表面形成液膜,阻碍CO2的扩散及与胺接触[74]

然而,基于现有的数据样本,无论在单变量分析或者SHAP分析中,湿度对于固体胺吸附剂的重要影响效应都没有得到充分体现。这是因为不同相对湿度条件下的实验数据相对于干燥空气中的研究而言不够丰富,限制了通过机器学习方法获得具有统计代表性的分析结果。此外,气体流量、气体组分等因素也会对固体胺吸附剂的性能产生影响。通过机器学习方法研究这些影响因素以及它们之间的协同关系也将建立在更多的实验数据基础之上。

最后,理想的固体胺DAC吸附剂需要满足多重标准,除了CO2的吸附容量,吸附剂的传质效率、循环稳定性等也是非常重要的评价指标。从机器学习视角来看,这些问题仍然需要更多的实验数据支撑,以开展系统性的分析。基于新的实验数据,机器学习模型能够得到进化,帮助研究者更加全面和准确地进行DAC吸附剂的开发。

3 结论与展望

本文采用机器学习算法对决定固体胺DAC吸附剂的CO2吸附容量的一系列相关参数进行了量化解析。主要结论如下:

1)单变量分析结果表明,固体胺吸附剂的CO2吸附容量受到多重因素影响,24个特征描述符与CO2吸附容量之间表现出明显的非线性关系。

2)RF算法成功实现了对固体胺DAC吸附剂CO2吸附容量的准确预测,性能指标:训练集R2=0.964,测试集R2=0.823。尤其是对于训练样本数相对较少的高性能DAC吸附剂,RF模型也能做出准确的预测。

3)SHAP分析的结果显示,有机胺负载量是影响CO2吸附量最为重要的因素,平均对CO2吸附容量的影响为0.46 mmol/g且为正向贡献;负载有机胺的分子量也比较重要,具有相对低分子量的有机胺能够提升CO2的吸附容量。其余7个比较重要的特征描述符为:CO2实验浓度、实验温度、吸附剂的BET表面积、孔隙率和基底的O、C、F元素组成。

面向新型DAC固体胺吸附剂的开发,本文提出以下设计建议:使用具有高孔隙率的杂化材料基底、低聚合度高负载量的有机胺,同时在吸附过程中选择更高的CO2浓度和较低的操作温度等。以上策略有助于固体胺吸附剂DAC性能的进一步提升,加快DAC这一关键负碳技术的开发与部署。

参考文献

[1]

United Nations Environment Programme.Emissions gap report 2023:Broken record-temperatures hit new highs,yet world fails to cut emissions (again)[R].Nairobi:United Nations Environment Programme,2023. doi:10.59117/20.500.11822/43922

[2]

Fuhrman J, McJeon H, Patel P,et al.Food-energy-water implications of negative emissions technologies in a +1.5 ℃ future[J].Nature Climate Change,2020,10(10):920‒927. doi:10.1038/s41558-020-0876-z

[3]

Zhu Xuancan, Xie Wenwen, Wu Junye,et al.Recent advances in direct air capture by adsorption[J].Chemical Society Reviews,2022,51(15):6574‒6651. doi:10.1039/d1cs00970b

[4]

Pradhan S, Shobe W M, Fuhrman J,et al.Effects of direct air capture technology availability on stranded assets and committed emissions in the power sector[J].Frontiers in Climate,2021,3:660787. doi:10.3389/fclim.2021.660787

[5]

Butnar I, Broad O, Rodriguez B S,et al.The role of bioenergy for global deep decarbonization:CO2 removal or low-carbon energy?[J].GCB Bioenergy,2020,12(3):198‒212. doi:10.1111/gcbb.12666

[6]

Bistline J E T, Blanford G J.Impact of carbon dioxide removal technologies on deep decarbonization of the electric power sector[J].Nature Communications,2021,12:3732. doi:10.1038/s41467-021-23554-6

[7]

Hou Zhengmeng, Luo Jiashun, Cao Cheng,et al.Development and contribution of natural gas industry under the goal of carbon neutrality in China[J].Advanced Engineering Sciences,2023,55(1):243‒252.

[8]

侯正猛,罗佳顺,曹成,.中国碳中和目标下的天然气产业发展与贡献[J].工程科学与技术,2023,55(1):243‒252.

[9]

Ren Shihua, Xie Yachen, Jiao Xiaomiao,et al.Characteristics of carbon emissions during coal development and technical approaches for carbon neutral development[J].Advanced Engineering Sciences,2022,54(1):60‒68.

[10]

任世华,谢亚辰,焦小淼,.煤炭开发过程碳排放特征及碳中和发展的技术途径[J].工程科学与技术,2022,54(1):60‒68.

[11]

Chen Bin, Xie Heping, Liu tao,et al.Principles and progress of advanced hydrogen production technologies in the context of carbon neutrality[J].Advanced Engineering Sciences,2022,54(1):106‒116.

[12]

陈彬,谢和平,刘涛,.碳中和背景下先进制氢原理与技术研究进展[J]. 工程科学与技术,2022,54(1):106‒116.

[13]

Chao Cheng, Hou Zhengmeng, Xiong Ying,et al.Technical routes and action plan for carbon neutral for Yunnan Province[J].Advanced Engineering Sciences,2022,54(1):37‒46.

[14]

曹成,侯正猛,熊鹰,.云南省碳中和技术路线与行动方案[J].工程科学与技术,2022,54(1):37‒46.

[15]

Hou Zhengmeng, Xiong Ying, Liu Jianhua,et al.Strategy, Technical route and action plan towards carbon peak and carbon neutrality in Henan Province[J].Advanced Engineering Sciences,2022,54(1):23‒36.

[16]

侯正猛,熊鹰,刘建华,.河南省碳达峰与碳中和战略、技术路线和行动方案[J].工程科学与技术,2022,54(1):23‒36.

[17]

Kwon H T, Sakwa‒Novak M A, Pang S H,et al.Aminopolymer-impregnated hierarchical silica structures:Unexpected equivalent CO2 uptake under simulated air capture and flue gas capture conditions[J].Chemistry of Materials,2019,31(14):5229‒5237. doi:10.1021/acs.chemmater.9b01474

[18]

Darunte L A, Walton K S, Sholl D S,et al.CO2 capture via adsorption in amine-functionalized sorbents[J].Current Opinion in Chemical Engineering,2016,12:82‒90. doi:10.1016/j.coche.2016.03.002

[19]

Panda D, Kulkarni V, Singh S K.Evaluation of amine-based solid adsorbents for direct air capture:A critical review[J].Reaction Chemistry & Engineering,2023,8(1):10‒40. doi:10.1039/d2re00211f

[20]

Boyd P G, Chidambaram A, García‒Díez E,et al.Data-driven design of metal-organic frameworks for wet flue gas CO2 capture[J].Nature,2019,576(7786):253‒256. doi:10.1038/s41586-019-1798-7

[21]

Avci G, Velioglu S, Keskin S.High-throughput screening of MOF adsorbents and membranes for H2 purification and CO2 capture[J].ACS Applied Materials & Interfaces,2018,10(39):33693‒33706. doi:10.1021/acsami.8b12746

[22]

Leperi K T, Chung Y G, You Fengqi,et al.Development of a general evaluation metric for rapid screening of adsorbent materials for postcombustion CO2 capture[J].ACS Sustainable Chemistry & Engineering,2019,7(13):11529‒11539. doi:10.1021/acssuschemeng.9b01418

[23]

Yuan Xiangzhou, Suvarna M, Low S,et al.Applied machine learning for prediction of CO2 adsorption on biomass waste-derived porous carbons[J].Environmental Science & Technology,2021,55(17):11925‒11936. doi:10.1021/acs.est.1c01849

[24]

Situ Yizhen, Yuan Xueying, Bai Xiangning,et al.Large-scale screening and machine learning for metal-organic framework membranes to capture CO2 from flue gas[J].Membranes,2022,12(7):700. doi:10.3390/membranes12070700

[25]

Wang Shihui, Cheng Min, Luo Lei,et al.High-throughput screening of metal-organic frameworks for hydrogen purification[J].Chemical Engineering Journal,2023,451:138436. doi:10.1016/j.cej.2022.138436

[26]

Cheng Min, Wang Shihui, Zhang Zhiyuan,et al.High-throughput virtual screening of metal-organic frameworks for xenon recovery from exhaled anesthetic gas mixture[J].Chemical Engineering Journal,2023,451:138218. doi:10.1016/j.cej.2022.138218

[27]

Zhang Zhiyuan, Cheng Min, Xiao Xinyi,et al.Machine-learning-guided identification of coordination polymer ligands for crystallizing separation of Cs/Sr[J].ACS Applied Materials & Interfaces,2022,14(29):33076‒33084. doi:10.1021/acsami.2c05272

[28]

Zhang Zhiyuan, Dong Yue, Qiu Yuqing,et al.Deep-learning-guided High-throughput evaluation of ligands for selective Sr/Cs coordination[J].Journal of Nuclear and Radiochemistry,2023,45(5):456‒465.

[29]

张智渊,董越,邱雨晴,.深度学习引导的高通量分子筛选用于锶铯的选择性配位[J].核化学与放射化学,2023,45(5):456‒465.

[30]

Cheng Min, Zhang Zhiyuan, Wang Shihui,et al.A large-scale screening of metal-organic frameworks for iodine capture combining molecular simulation and machine learning[J].Frontiers of Environmental Science & Engineering,2023,17(12):148. doi:10.1007/s11783-023-1748-3

[31]

Wang Bingbing, Zhang Zhiyuan, Dong Yue,et al.Machine-learning-enabled ligand screening for Cs/Sr crystallizing separation[J].Inorganic Chemistry,2023,62(33):13293‒13303. doi:10.1021/acs.inorgchem.3c01564

[32]

Wassertheil S, Cohen J.Statistical power analysis for the behavioral sciences[J].Biometrics,1970,26(3):588. doi:10.2307/2529115

[33]

Pedregosa F, Varoquaux G, Gramfort A,et al.Scikit-learn: Machine Learning in Python[J].Journal of Machine Learning Research,2011,12:2825‒2830.

[34]

Belmabkhout Y, Serna‒Guerrero R, Sayari A.Adsorption of CO2-containing gas mixtures over amine-bearing pore-expanded MCM-41 silica:Application for gas purification[J].Industrial & Engineering Chemistry Research,2010,49(1):359‒365. doi:10.1021/ie900837t

[35]

Kumar D R, Rosu C, Sujan A R,et al.Alkyl-aryl amine-rich molecules for CO2 removal via direct air capture[J].ACS Sustainable Chemistry & Engineering,2020,8(9):10971‒10982.

[36]

Rim G, Kong Fanhe, Song Mingyu,et al.Sub-ambient temperature direct air capture of CO2 using amine-impregnated MIL-101(Cr) enables ambient temperature CO2 recovery[J].JACS Au,2022,2(2):380‒393. doi:10.1021/jacsau.1c00414

[37]

Didas S A, Kulkarni A R, Sholl D S,et al.Role of amine structure on carbon dioxide adsorption from ultradilute gas streams such as ambient air[J].ChemSusChem,2012,5(10):2058‒2064. doi:10.1002/cssc.201200196

[38]

Miao Yihe, He Zhijun, Zhu Xuancan,et al.Operating temperatures affect direct air capture of CO2 in polyamine-loaded mesoporous silica[J].Chemical Engineering Journal,2021,426:131875. doi:10.1016/j.cej.2021.131875

[39]

Potter M E, Cho K M, Lee J J,et al.Role of alumina basicity in CO2 uptake in 3-aminopropylsilyl-grafted alumina adsorbents[J].ChemSusChem,2017,10(10):2192‒2201. doi:10.1002/cssc.201700115

[40]

Choi S, Drese J H, Eisenberger P M,et al.Application of amine-tethered solid sorbents for direct CO2 capture from the ambient air[J].Environmental Science & Technology,2011,45(6):2420‒2427. doi:10.1021/es102797w

[41]

Sujan A R, Pang S H, Zhu Guanghui,et al.Direct CO2 capture from air using poly(ethylenimine)-loaded polymer/silica fiber sorbents[J].ACS Sustainable Chemistry & Engineering,2019,7(5):5264‒5273. doi:10.1021/acssuschemeng.8b06203

[42]

Wang Jitong, Huang Haihong, Wang Mei,et al.Direct capture of low-concentration CO2 on mesoporous carbon-supported solid amine adsorbents at ambient temperature[J].Industrial & Engineering Chemistry Research,2015,54(19):5319‒5327. doi:10.1021/acs.iecr.5b01060

[43]

Pang S H, Lively R P, Jones C W.Oxidatively-stable linear poly(propylenimine)-containing adsorbents for CO2 capture from ultradilute streams[J].ChemSusChem,2018,11(15):2628‒2637. doi:10.1002/cssc.201800438

[44]

Sujan A R, Kumar D R, Sakwa-Novak M,et al.Poly(glycidyl amine)-loaded SBA-15 sorbents for CO2 capture from dilute and ultradilute gas mixtures[J].ACS Applied Polymer Materials,2019,1(11):3137‒3147. doi:10.1021/acsapm.9b00788

[45]

Wurzbacher J A, Gebald C, Steinfeld A.Separation of CO2 from air by temperature-vacuum swing adsorption using diamine-functionalized silica gel[J].Energy & Environmental Science,2011,4(9):3584‒3592. doi:10.1039/c1ee01681d

[46]

Kong Yong, Jiang Guodong, Wu Ye,et al.Amine hybrid aerogel for high-efficiency CO2 capture:Effect of amine loading and CO2 concentration[J].Chemical Engineering Journal,2016,306:362‒368. doi:10.1016/j.cej.2016.07.092

[47]

Goeppert A, Czaun M, May R B,et al.Carbon dioxide capture from the air using a polyamine based regenerable solid adsorbent[J].Journal of the American Chemical Society,2011,133(50):20164‒20167. doi:10.1021/ja2100005

[48]

Wang Xiaoxing, Ma Xiaoliang, Schwartz V,et al.A solid molecular basket sorbent for CO2 capture from gas streams with low CO2 concentration under ambient conditions[J].Physical Chemistry Chemical Physics,2012,14(4):1485‒1492. doi:10.1039/c1cp23366a

[49]

Sarazen M L, Sakwa-Novak M A, Ping E W,et al.Effect of different acid initiators on branched poly(propylenimine) synthesis and CO2 sorption performance[J].ACS Sustainable Chemistry & Engineering,2019,7(7):7338‒7345. doi:10.1021/acssuschemeng.9b00512

[50]

Holewinski A, Sakwa-Novak M A, Jones C W.Linking CO2 sorption performance to polymer morphology in aminopolymer/silica composites through neutron scattering[J].Journal of the American Chemical Society,2015,137(36):11749‒11759. doi:10.1021/jacs.5b06823

[51]

Rosu C, Pang S H, Sujan A R,et al.Effect of extended aging and oxidation on linear poly(propylenimine)-mesoporous silica composites for CO2 capture from simulated air and flue gas streams[J].ACS Applied Materials & Interfaces,2020,12(34):38085‒38097. doi:10.1021/acsami.0c09554

[52]

Pang S H, Lee Lichen, Sakwa-Novak M A,et al.Design of aminopolymer structure to enhance performance and stability of CO2 sorbents:Poly(propylenimine) vs Poly(ethylenimine)[J].Journal of the American Chemical Society,2017,139(10):3627‒3630. doi:10.1021/jacs.7b00235

[53]

Kuwahara Y, Kang D Y, Copeland J R,et al.Dramatic enhancement of CO2 uptake by poly(ethyleneimine) using zirconosilicate supports[J].Journal of the American Chemical Society,2012,134(26):10757‒10760. doi:10.1021/ja303136e

[54]

Darunte L A, Oetomo A D, Walton K S,et al.Direct air capture of CO2 using amine functionalized MIL-101(Cr)[J].ACS Sustainable Chemistry & Engineering,2016,4(10):5761‒5768. doi:10.1021/acssuschemeng.6b01692

[55]

Stuckert A N, Yang R T.CO2 capture from the atmosphere and simultaneous concentration using zeolites and amine-grafted SBA-15[J].Environmental Science & Technology,2011,45(23):10257‒10264. doi:10.1021/es202647a

[56]

Liu Faqian, Wang Lei, Huang Zhaoge,et al.Amine-tethered adsorbents based on three-dimensional macroporous silica for CO2 capture from simulated flue gas and air[J].ACS Applied Materials & Interfaces,2014,6(6):4371‒4381. doi:10.1021/am500089g

[57]

Lin Yichao, Yan Qiuju, Kong Chunlong,et al.Polyethyleneimine incorporated metal-organic frameworks adsorbent for highly selective CO2 capture[J].Scientific Reports,2013,3:1859. doi:10.1038/srep01859

[58]

Sakwa-Novak M A, Yoo C J, Tan Shuai,et al.Poly(ethylenimine)-functionalized monolithic alumina honeycomb adsorbents for CO2 capture from air[J].ChemSusChem,2016,9(14):1859‒1868. doi:10.1002/cssc.201600404

[59]

Sakwa-Novak M A, Jones C W.Steam induced structural changes of a poly(ethylenimine) impregnated γ-alumina sorbent for CO2 extraction from ambient air[J].ACS Applied Materials & Interfaces,2014,6(12):9245‒9255. doi:10.1021/am501500q

[60]

Chaikittisilp W, Kim H J, Jones C W.Mesoporous alumina-supported amines as potential steam-stable adsorbents for capturing CO2 from simulated flue gas and ambient air[J].Energy & Fuels,2011,25(11):5528‒5537. doi:10.1021/ef201224v

[61]

Liu Lina, Chen Jian, Tao Lin,et al.Aminopolymer confined in ethane-silica nanotubes for CO2 capture from ambient air[J].ChemNanoMat,2020,6(7):1096‒1103. doi:10.1002/cnma.201900742

[62]

Kuwahara Y, Kang D Y, Copeland J R,et al.Enhanced CO2 adsorption over polymeric amines supported on heteroatom-incorporated SBA-15 silica:Impact of heteroatom type and loading on sorbent structure and adsorption performance[J].Chemistry-A European Journal,2012,18(52):16649‒16664. doi:10.1002/chem.201203144

[63]

Park S J, Lee J J, Hoyt C B,et al.Silica supported poly(propylene guanidine) as a CO2 sorbent in simulated flue gas and direct air capture[J].Adsorption,2020,26(1):89‒101. doi:10.1007/s10450-019-00171-w

[64]

Wijesiri R P, Knowles G P, Yeasmin H,et al.CO2 capture from air using pelletized polyethylenimine impregnated MCF silica[J].Industrial & Engineering Chemistry Research,2019,58(8):3293‒3303. doi:10.1021/acs.iecr.8b04973

[65]

Zhu Xuancan, Ge Tianshu, Yang Fan,et al.Efficient CO2 capture from ambient air with amine-functionalized Mg‒Al mixed metal oxides[J].Journal of Materials Chemistry A,2020,8(32):16421‒16428. doi:10.1039/d0ta05079b

[66]

Breiman L.Random forests[J].Machine Learning,2001,45(1):5‒32. doi:10.1023/a:1010933404324

[67]

Rumelhart D E, Hinton G E, Williams R J.Learning representations by back-propagating errors[J].Nature,1986,323(6088):533‒536. doi:10.1038/323533a0

[68]

Hearst M A, Dumais S T, Osuna E,et al.Support vector machines[J].IEEE Intelligent Systems and Their Applications,1998,13(4):18‒28. doi:10.1109/5254.708428

[69]

McDonald G C.Ridge regression[J].WIREs Computational Statistics,2009,1(1):93‒100. doi:10.1002/wics.14

[70]

Cogswell C F, Xie Z, Wolek A,et al.Pore structure-CO2 adsorption property relations of supported amine materials with multi-pore networks[J].Journal of Materials Chemistry A,2017,5(18):8526‒8536. doi:10.1039/c7ta01616f

[71]

Heydari-Gorji A, Belmabkhout Y, Sayari A.Polyethylenimine-impregnated mesoporous silica:Effect of amine loading and surface alkyl chains on CO2 adsorption[J].Langmuir,2011,27(20):12411‒12416. doi:10.1021/la202972t

[72]

Wilson M, Barrientos-Palomo S N, Stevens P C,et al.Substrate-independent epitaxial growth of the metal-organic framework MOF-508a[J].ACS Applied Materials & Interfaces,2018,10(4):4057‒4065. doi:10.1021/acsami.7b16029

[73]

Liu Lina, Chen Jian, Tao Lin,et al.Aminopolymer confined in ethane-silica nanotubes for CO2 capture from ambient air[J].ChemNanoMat,2020,6(7):1096‒1103. doi:10.1002/cnma.201900742

[74]

Wijesiri R P, Knowles G P, Yeasmin H,et al.CO2 capture from air using pelletized polyethylenimine impregnated MCF silica[J].Industrial & Engineering Chemistry Research,2019,58(8):3293‒3303. doi:10.1021/acs.iecr.8b04973

[75]

Sehaqui H, Gálvez M E, Becatinni V,et al.Fast and reversible direct CO2 capture from air onto all-polymer nanofibrillated cellulose—Polyethylenimine foams[J].Environmental Science & Technology,2015,49(5):3167‒3174. doi:10.1021/es504396v

[76]

Goeppert A, Czaun M, May R B,et al.Carbon dioxide capture from the air using a polyamine based regenerable solid adsorbent[J].Journal of the American Chemical Society,2011,133(50):20164‒20167. doi:10.1021/ja2100005

[77]

Knöfel C, Martin C, Hornebecq V,et al.Study of carbon dioxide adsorption on mesoporous aminopropylsilane-functionalized silica and titania combining microcalorimetry and in situ infrared spectroscopy[J].The Journal of Physical Chemistry C,2009,113(52):21726‒21734. doi:10.1021/jp907054h

[78]

Chang A C C, Chuang S S C, Gray M,et al. In-situ infrared study of CO2 adsorption on SBA-15 grafted with γ-(aminopropyl)triethoxysilane[J].Energy & Fuels,2003,17(2):468‒473. doi:10.1021/ef020176h

[79]

Liu Qing, Xiong Bitao, Shi Junjie,et al.Enhanced tolerance to flue gas contaminants on carbon dioxide capture using amine-functionalized multiwalled carbon nanotubes[J].Energy & Fuels,2014,28(10):6494‒6501. doi:10.1021/ef501614m

基金资助

中国石油化工股份有限公司技术开发项目(323048)

PDF (2776KB)

346

访问

0

被引

详细

导航
相关文章

AI思维导图

/