(371ar) Advanced Integration Strategies and Machine Learning-Based Superstructure Optimization for Power-to-Methanol

Power-to-X is well-known as a promising technology for converting renewable electricity into chemicals and fuels [1]. Among studied chemicals, methanol (MeOH) has emerged as one of the most viable products due to high demand, with the global methanol market reaching approximately 91 billion USD by 2026 [2]. Additionally, Power-to-Methanol (PtMe) utilizes a significant amount of carbon dioxide (CO₂) as a reactant. Consequently, the PtMe process has received much attention to contribute to achieving net zero emissions by 2050.

The PtMe process consists of green hydrogen (H₂) production and CO₂-to-Methanol (CTM) sections. Because the hydrogenation method is the most commonly used technology for the CTM section [3], previous studies have primarily investigated PtMe designs, which integrate various electrolysis technologies for green H₂ production with CO₂ hydrogenation [4], [5]. Additionally, due to the highly complex level of process integration, several studies have utilized specific electrolysis technologies and arbitrary integration strategies to improve the performance of the PtMe process [6], [7]. Meanwhile, the PtMe process can utilize different electrolysis technologies that generate significant quantities of waste heat, off-gas, O₂, and water and consume huge amounts of electricity, heat, and steam. Therefore, it is necessary to propose applicable integration strategies for selecting proper electrolysis technologies and reutilizing efficiently waste utilities to enhance the techno-economic-environmental (TEE) performance of the PtMe process.

Using various integration strategies leads to the demand for solving a superstructure design optimization to determine the most feasible one. Previous studies utilized algebraic equations to formulate a superstructure modeling and optimization framework for the PtMe process [8], [9]. However, the PtMe process typically consists of algebraic equations, ordinary algebraic equations, and differential algebraic equations to elucidate physical properties, transportation phenomena, and reaction kinetics. Consequently, solving superstructure design optimization, which is formulated from the integrated process simulator, becomes an urgent task for the PtMe process design. Solving the superstructure optimization problem, which directly connects with a simulator of the integrated PtMe process, faces the singificant challenge of the high computational cost for each simulation. Therefore, we forecast that ML-based superstructure optimization can identify the optimal design among the feasible PtMe designs.

To close the mentioned critical research gaps, we propose feasible integration strategies and the practical ML-based superstructure optimization approach to enhance the TEE performance of the PtMe process. Firstly, mathematical models of three electrolyzers, including AWE, PEM, and SOE, are developed and validated with various experimental references. The validated models are then integrated with the CTM section and different strategies of waste-utilities reutilization to formulate 15 PtMe designs. Secondly, an ML model is developed to represent 15 PtMe designs and then used to formulate optimization problems to determine the optimal designs in various aspects of TEE performance. Finally, from the optimal designs, an analysis is conducted to determine the main factors affecting the TEE performance and suggest a guideline for selecting the proper PtMe design under uncertainty.

Reference

• J. Palys, P. Daoutidis, Power-to-X: A review and perspective, Computers & Chemical Engineering, vol. 165, p. 107948, 2022.
• D. Nguyen, J. Chang, S.H. Hong, C.H. Lee, Performance and ANN-based optimization of an advanced process for wet CO₂-to-Methanol using a catalytic fluidized bed reactor integrated with separators, Fuel, vol. 343, p. 128045, 2023.
• D. Vo, M. Oh, C.H. Lee, Design guideline for CO₂ to methanol conversion process supported by generic model of various bed reactors, Energy Conversion and Management, vol. 269, p. 116079, 2022.
• Lee, C. Choe, H. Kim, A. Kim, Y.U. Shin, J. Haider, H. Lim, Economic Parity Analysis of Green Methanol Synthesis Using Water Electrolysis Based on Renewable Energy, ACS Sustainable Chemistry & Engineering, vol. 9, p. 15807–15818, 2021.
• A. Adnan, M.G. Kibria, Comparative techno-economic and life-cycle assessment of power-to-methanol synthesis pathways, Applied Energy, vol. 278, p. 115614, 2020.
• Battaglia, G. Buffo, D. Ferrero, M. Santarelli, A. Lanzini, Methanol synthesis through CO2 capture and hydrogenation: Thermal integration, energy performance and techno-economic assessment, Journal of CO2 Utilization, vol. 44, p. 101407, 2021.
• Szima, C.C. Cormos, Improving methanol synthesis from carbon-free H2 and captured CO2: A techno-economic and environmental evaluation, Journal of CO2 Utilization, vol. 24, p. 555–563, 2018.
• Kenkel, T. Wassermann, C. Rose, E. Zondervan, A generic superstructure modeling and optimization framework on the example of bi-criteria Power-to-Methanol process design, Computers & Chemical Engineering, vol. 150, p. 107327, 2021.
• Svitnič, K. Sundmacher, Identifying standard and simple designs of Power-to-Methanol processes: The costs of complexity reduction, Energy Conversion and Management, vol. 307, p. 118325, 2024.