Advancing medium- and heavy-duty electric vehicle adoption models with novel natural language processing metrics
Date
2024
Journal Title
Journal ISSN
Volume Title
Abstract
The transportation sector must rapidly decarbonize to meet its emissions reduction targets. Medium- and heavy-duty decarbonization is lagging behind the light-duty industry due to technical and operational challenges and the choices made by medium- and heavy-duty fleet operators. Research investigating the procurement considerations of fleets has relied heavily on interviews and surveys, but many of these studies need higher participation rates and are difficult to generalize. To model fleet operators' decision-making priorities, this thesis applies a robust text analysis approach based on latent Dirichlet allocation and Bi-directional Encoder Representations of Transformers to two broad corpora of fleet adoption literature from academia and industry. Based on a newly developed metric, this thesis finds that the academic corpus emphasizes the importance of suitability, familiarity, norms, and brand image. These perception rankings are then passed to an agent-based model to determine how differences in perception affect adoption predictions. The results show a forecast of accelerated medium- and heavy-duty electric vehicle adoption when using the findings from the academic corpus versus the industry corpus.
Description
Rights Access
Subject
BERTopic
latent Dirichlet allocation
natural language processing
electrification
agent-based modeling
medium- and heavy-duty transportation