Optimizing energy supply superstructure for plastic waste gasification systems: minimizing life cycle environmental impacts with AI models

This study optimizes the energy supply superstructure for plastic waste-to-energy system through path programming using machine learning models. Multiple fuel alternatives, carbon capture technologies, renewable energy driven water electrolysis techniques are incorporated into a mixed-integer nonlinear programming model. Surrogate model-based optimization strategy, which utilizes a machine learning model for regression, was applied to solve the path programming problem. The objective is to minimize the life cycle environmental impacts, with concurrent optimization of the surrogate model's hyperparameters. Feature importance analysis identifies the selection of carbon dioxide usage pathways as the most significant feature in determining the environmental impacts. 93.39 % of the CO₂ is favorable for compression for storage, while only 6.61 % is utilized for methanol synthesis. Negative global warming potential value is obtained for the optimal energy supply superstructure. Additionally, the study explores the interconnections between different midpoint environmental categories. Minimizing global warming potential and fossil resource scarcity impacts synergistically leads to significant increases in other environmental indicators. Conversely, minimizing human carcinogenic toxicity results in trade-offs with global warming potential. This research provides valuable insights into the environmental optimization of plastic waste valorization processes, highlighting the intricate balance required between different environmental objectives.