A rule-based digital twin strategy for optimizing hydrogen production from biogas under dynamic feed conditions

Hydrogen production from biogas offers a promising pathway for decarbonizing both the energy and industrial sectors. However, fluctuations in biogas feed composition can significantly affect the thermodynamics of steam methane reforming (SMR), leading to inefficiencies, increased emissions, and higher operational costs. This study proposes a novel digital twin-based smart multi-objective optimization (SMOO) strategy designed to adaptively regulate SMR operations under variable biogas feed conditions. The approach combines a k-nearest neighbors (KNN) classifier to identify feed scenarios with a deep neural network (DNN) model that simulates SMR performance. Optimization is carried out using the non-dominated sorting genetic algorithm II (NSGA-II) to minimize unit production cost (UPC) and net carbon emissions (NCE), while maximizing energy efficiency (EEF). The rule based-SMOO strategy is evaluated across three operational scenarios: fixed, moderate variation, and intensive variation in biogas composition. Results show that the moderate mode led to reductions of 19.25 % in UPC and 20.31 % in NCE, along with a 21.28 % increase in EEF. Although the intensive mode yielded slightly higher performance, it lacked operational stability due to the severity of fluctuations in SMR conditions. The moderate mode was therefore selected as the optimal balance for assigning the regulation rules. When validated to real-world biogas data, the SMOO strategy resulted in a 30.59 % decrease in UPC and a 22.92 % improvement in efficiency. Thus, the proposed rule-based SMOO strategy offers a practical and effective solution for optimizing hydrogen production under dynamic feed conditions, ensuring both economic and environmental sustainability.