Advanced integration strategies and machine learning-based superstructure optimization for Power-to-Methanol

Abstract

The Power-to-methanol (PtMe) process faces significant challenges, including high production costs, low energy efficiency, and a lack of systematic and applicable integrated design and superstructure optimization methods. This study proposes advanced integration and machine learning (ML)-based superstructure optimization approaches that aim to enhance the performance of the PtMe process. Alkaline water electrolyzer (AWE), polymer electrolyte membrane electrolyzer (PEM), and solid oxide electrolyzer (SOE) are chosen for investigation due to their high technology readiness levels. The validated mathematical models for these electrolyzers are integrated with other units to form 3 conventional and 12 advanced designs. The conventional designs comprise electrolyzer-based H${}_2$ and CO${}_2$-to-methanol sections. In contrast, the advanced designs integrate these sections with four waste-utility reutilization strategies, including heat (H), heat and steam (HS), heat and power (HP), and heat, steam, and power (HSP) generations. A techno-economic analysis demonstrates the pivotal role of electrolyzers in the PtMe process. Two deep neural networks (DNN) models are developed to represent the superstructure design of the PtMe process. With marginal training and test errors (0.28% and 1.03%), the one-hot vector-DNN (OHV-DNN) model is selected to formulate four optimization problems, identifying the PtMe-SOE-HSP and PtMe-AWE-HSP designs as optimal solutions for minimizing energy consumption and production cost considering carbon tax. The PtMe-AWE and PtMe-SOE designs are the best candidates among the conventional designs. Compared to the optimal conventional designs, the optimal advanced designs improve the techno-economic-environmental performance by 1.8–29.7%. Additionally, compared to the PtMe-AWE-HSP design, the PtMe-SOE-HSP design achieves a 4.3% reduction in net CO${}_2$ reduction and a 10.2% reduction in energy consumption. Then, an economic analysis reveals the PtMe-SOE-HSP design as the superior design under scenarios of reduced electrolyzer CAPEX and increased electrolyzer lifetime. These findings are valuable for improving the techno-economic-environmental performance of the PtMe process. Moreover, the proposed integration strategies and ML-based superstructure optimization approach hold the promise for enhancing other power-to-liquid processes.