Management of Zero Emissions Zones in Urban Transport by Means of Neuroevolution

Abstract

As cities strive to reduce carbon emissions in transportation, plugin hybrid electric buses (PHEB) stand out as an essential technology, proposing a solution that is more flexible than fully electric buses. Whileexisting studies focus on designing operational strategies for PHEBs, such as selecting the optimal propulsion method along a route to minimize emissions or maximize electric range, these strategies are often static and cannot dynamically adapt to unforeseen circumstances impacting bus performance (e.g., changes in traffic or the battery state of energy). In this study, we propose a novel approach to address thislimitation by modeling the optimization of PHEB operational strategiesas a classification problem. Our method, named ML-EPBO (for Ma-chine Learning Efficient PHE Bus Operation Model), leverages machine learning and takes into account both local information from the bus and route-specific data to dynamically adapt to unexpected traffic conditions.We employ neuroevolution [1] to fine-tune ML-EPBO for optimal performance. Results show how ML-EPBO rapidly generates competitive solu-tions, outperforming existing techniques that lack the ability to dynami-cally adapt to unforeseen events due to their reliance on global knowledgeand high computational complexity. The work was originally publishedin [2]. The model achieves an average accuracy exceeding 99.2%, with proposed strategies accurately predicted in over 70% of cases, surpassing state-of-the-art results by up to 4.7% in emission reduction and 1.82% in electric range improvement.