{"title":"通过机器学习优化多孔结构,促进太阳能热化学燃料生产","authors":"Da Xu , Lei Zhao , Meng Lin","doi":"10.1016/j.pnsc.2024.07.024","DOIUrl":null,"url":null,"abstract":"<div><div>Porous reactant is the key component in solar thermochemical reactions, significantly affecting the solar energy conversion and fuel production performance. Triply periodic minimal surface (TPMS) structures, with analytical expressions and predictable structure-property relationships, can facilitate the design and optimization of such structures. This work proposes a machine learning-assisted framework to optimize TPMS structures for enhanced reaction efficiency, increased fuel production, and reduced temperature gradients. To mitigate the computational cost of conventional high-throughput optimization, neural network regression models were used to for performance prediction based on input features. The training dataset was generated using a three-dimensional multiphysics model for the thermochemical reduction driven by concentrated solar energy considering fluid flow, heat and mass transfer, and chemical reacions. Both uniform and gradient structures were initially assessed by the three-dimensional model showing gradient design in <em>c</em> and <em>ω</em> were necessary for performance enhancement. Further, with our proposed optimization framework, we found that structures with parameters <em>c</em><sub>1</sub> = <em>c</em><sub>2</sub> = 0.5 (uniform in <em>c</em> ) and <em>ω</em><sub>1</sub> = 0.2, <em>ω</em><sub>2</sub> = 0.8 (gradient in <em>ω</em>) achieved the highest relative efficiency (<em>f</em><sub>chem</sub><em>/f</em><sub>chem,ref</sub>) of 1.58, a relative fuel production (Δ<em>δ</em>/Δ<em>δ</em><sub>ref</sub>) of 7.94, and a max relative temperature gradient (<em>dT/dy)/</em>(<em>dT/dy</em>)<sub>ref</sub> of 0.26. Kinetic properties, i.e., bulk diffusion and surface exchange coefficient, were also studied showing that for materilas with slow kinetics, the design space in terms of <em>c</em> and <em>ω</em> were highly limited compared to fast kinetics materials. Our framework is adaptable to diverse porous structures and operational conditions, making it a versatile tool for screening porous structures for solar thermochemical applications. This work has the potential to advance the development of efficient solar fuel production systems and scalable industrial applications in renewable energy technologies.</div></div>","PeriodicalId":20742,"journal":{"name":"Progress in Natural Science: Materials International","volume":"34 5","pages":"Pages 895-906"},"PeriodicalIF":4.8000,"publicationDate":"2024-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Optimization of porous structures via machine learning for solar thermochemical fuel production\",\"authors\":\"Da Xu , Lei Zhao , Meng Lin\",\"doi\":\"10.1016/j.pnsc.2024.07.024\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Porous reactant is the key component in solar thermochemical reactions, significantly affecting the solar energy conversion and fuel production performance. Triply periodic minimal surface (TPMS) structures, with analytical expressions and predictable structure-property relationships, can facilitate the design and optimization of such structures. This work proposes a machine learning-assisted framework to optimize TPMS structures for enhanced reaction efficiency, increased fuel production, and reduced temperature gradients. To mitigate the computational cost of conventional high-throughput optimization, neural network regression models were used to for performance prediction based on input features. The training dataset was generated using a three-dimensional multiphysics model for the thermochemical reduction driven by concentrated solar energy considering fluid flow, heat and mass transfer, and chemical reacions. Both uniform and gradient structures were initially assessed by the three-dimensional model showing gradient design in <em>c</em> and <em>ω</em> were necessary for performance enhancement. Further, with our proposed optimization framework, we found that structures with parameters <em>c</em><sub>1</sub> = <em>c</em><sub>2</sub> = 0.5 (uniform in <em>c</em> ) and <em>ω</em><sub>1</sub> = 0.2, <em>ω</em><sub>2</sub> = 0.8 (gradient in <em>ω</em>) achieved the highest relative efficiency (<em>f</em><sub>chem</sub><em>/f</em><sub>chem,ref</sub>) of 1.58, a relative fuel production (Δ<em>δ</em>/Δ<em>δ</em><sub>ref</sub>) of 7.94, and a max relative temperature gradient (<em>dT/dy)/</em>(<em>dT/dy</em>)<sub>ref</sub> of 0.26. Kinetic properties, i.e., bulk diffusion and surface exchange coefficient, were also studied showing that for materilas with slow kinetics, the design space in terms of <em>c</em> and <em>ω</em> were highly limited compared to fast kinetics materials. Our framework is adaptable to diverse porous structures and operational conditions, making it a versatile tool for screening porous structures for solar thermochemical applications. This work has the potential to advance the development of efficient solar fuel production systems and scalable industrial applications in renewable energy technologies.</div></div>\",\"PeriodicalId\":20742,\"journal\":{\"name\":\"Progress in Natural Science: Materials International\",\"volume\":\"34 5\",\"pages\":\"Pages 895-906\"},\"PeriodicalIF\":4.8000,\"publicationDate\":\"2024-10-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Progress in Natural Science: Materials International\",\"FirstCategoryId\":\"88\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S1002007124001710\",\"RegionNum\":2,\"RegionCategory\":\"材料科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"MATERIALS SCIENCE, MULTIDISCIPLINARY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Progress in Natural Science: Materials International","FirstCategoryId":"88","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1002007124001710","RegionNum":2,"RegionCategory":"材料科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"MATERIALS SCIENCE, MULTIDISCIPLINARY","Score":null,"Total":0}
Optimization of porous structures via machine learning for solar thermochemical fuel production
Porous reactant is the key component in solar thermochemical reactions, significantly affecting the solar energy conversion and fuel production performance. Triply periodic minimal surface (TPMS) structures, with analytical expressions and predictable structure-property relationships, can facilitate the design and optimization of such structures. This work proposes a machine learning-assisted framework to optimize TPMS structures for enhanced reaction efficiency, increased fuel production, and reduced temperature gradients. To mitigate the computational cost of conventional high-throughput optimization, neural network regression models were used to for performance prediction based on input features. The training dataset was generated using a three-dimensional multiphysics model for the thermochemical reduction driven by concentrated solar energy considering fluid flow, heat and mass transfer, and chemical reacions. Both uniform and gradient structures were initially assessed by the three-dimensional model showing gradient design in c and ω were necessary for performance enhancement. Further, with our proposed optimization framework, we found that structures with parameters c1 = c2 = 0.5 (uniform in c ) and ω1 = 0.2, ω2 = 0.8 (gradient in ω) achieved the highest relative efficiency (fchem/fchem,ref) of 1.58, a relative fuel production (Δδ/Δδref) of 7.94, and a max relative temperature gradient (dT/dy)/(dT/dy)ref of 0.26. Kinetic properties, i.e., bulk diffusion and surface exchange coefficient, were also studied showing that for materilas with slow kinetics, the design space in terms of c and ω were highly limited compared to fast kinetics materials. Our framework is adaptable to diverse porous structures and operational conditions, making it a versatile tool for screening porous structures for solar thermochemical applications. This work has the potential to advance the development of efficient solar fuel production systems and scalable industrial applications in renewable energy technologies.
期刊介绍:
Progress in Natural Science: Materials International provides scientists and engineers throughout the world with a central vehicle for the exchange and dissemination of basic theoretical studies and applied research of advanced materials. The emphasis is placed on original research, both analytical and experimental, which is of permanent interest to engineers and scientists, covering all aspects of new materials and technologies, such as, energy and environmental materials; advanced structural materials; advanced transportation materials, functional and electronic materials; nano-scale and amorphous materials; health and biological materials; materials modeling and simulation; materials characterization; and so on. The latest research achievements and innovative papers in basic theoretical studies and applied research of material science will be carefully selected and promptly reported. Thus, the aim of this Journal is to serve the global materials science and technology community with the latest research findings.
As a service to readers, an international bibliography of recent publications in advanced materials is published bimonthly.