Machine learning–based multi-objective optimisation of low-carbon and profitable hydrogen and diesel production from non-recycled municipal plastic waste: An integrated life cycle assessment and cost–benefit analysis

Bauyrzhan Biakhmetov, Г. Ж. Тасболат, Yue Li, Qunshan Zhao, Abay Mukhamediyarovich Dostiyarov, David Flynn, Peng Jiang, Siming You

Carbon Capture Science & Technology, 2026

Full text Open access: Gold Metadata (CSL-JSON)
DOI 10.1016/j.ccst.2026.100586
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

Sustainable plastic waste management is essential for net zero trajectory, potentially transforming the sector from an emissions source to a circular asset. MPWs (Municipal Plastic Wastes) that are not mechanically recycled can go through pyrolysis-based chemical recycling to produce hydrogen and diesel. There is limited understanding about the optimal configuration and design of pyrolysis-based chemical recycling of plastic waste. Associated attempts to optimise the recycling is rare. In this study, a reliable optimisation framework incorporating machine learning, life cycle assessment and cost-benefit analysis was developed for the design of the pyrolysis of Non-Recycled Municipal Plastic Waste (NMPW). Specifically, the global warming potential (GWP) and net-present value (NPV) of 900 diesel and hydrogen-producing scenarios for the pyrolysis of NMPW were calculated. Associated transportation and pyrolysis process were modelled using ArcGIS Pro and Aspen Plus, respectively. The long short-term memory recurrent neural network (LSTM-RNN) was applied to define temporal dependencies and dynamics of the system, which was integrated with Monte Carlo simulations to expand scenarios from 900 to 700,000. A Pareto curve was derived from the GWPs and NPVs, from which the optimal scenario in terms of environmental and economic performance was identified based on the comparison of two multi-criteria decision-making approaches, i.e., TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) and LINMAP (Linear Programming Technique for Multidimensional Analysis of Preference). The solutions by TOPSIS and LINMAP achieved GWPs of -2570.42 and -1025.28 kg CO 2 -eq per tonne NMPW, and NPVs of £300.32 and £-1402.92 per tonne NMPW, respectively. Thus, the TOPSIS scenario is preferable to the LINMAP scenario due to its lower carbon footprint and higher economic feasibility. This study showed that the proposed optimisation framework has the capacity to facilitate the design of pyrolysis-based processing of NMPW that is profitable and carbon-saving. Such systems could be deployed widely across the UK, where a large share of NMPW is currently either landfilled or incinerated.