Forecasting and surrogate models-based MINLP for long-term integrated design and planning of power-to-methanol

Abstract

The power-to-methanol (PtMe) process faces critical challenges, including the absence of efficient forecasting models, the lack of viable frameworks for integrated design and planning (IDP) optimization, and insufficient long-term evaluations of economic and operational flexibility. To address these research gaps, this study proposes a novel approach that integrates forecasting and surrogate models with mixed-integer nonlinear programming (MINLP) to optimize the long-term IDP of the PtMe process, with the aim of reducing production costs and enhancing operational flexibility. First, we develop two forecasting models to predict renewable energy availability over the next four years and two surrogate models to accurately represent the methanol production section. These models are then integrated with models of other sections to formulate four system models for the PtMe process. The system models are integrated with a superstructure design and MINLP to formulate four optimization problems, aiming to minimize methanol production costs. The optimization results indicate that incorporating the transformer and polynomial models with MINLP is the most effective approach, with the lithium-ion battery (LiB)-Grid-PtMe configuration emerging as the optimal design. Using the transformer model reduces the required LiB storage capacity and methanol production cost by 5.7 % and 41.01 %, respectively, while selecting an efficient design and integrating the grid reduces methanol production costs by up to 68.2 %. The findings are applicable to the long-term IDP of the PtMe process. Additionally, the proposed forecasting models and solution approach can be extended to the IDP of other power-to-liquid processes.