Analysis of Refueling Behavior Models for Hydrogen-Fuel Vehicles: Markov versus Generalized Poisson Modeling

Abstract

This paper presents an analysis of two methodologies that can be used to predict refueling behavior. Both models aim to provide insights into hydrogen-fuel vehicle users’ refueling patterns and behaviors. The first model leverages probabilistic transitions between refueling states to simulate and predict the refueling behavior of hydrogen vehicle users. In contrast, the GP-1 model employs Gaussian processes to capture the underlying patterns and uncertainties in hydrogen-fuel vehicle refueling behavior, taking into consideration additional factors such as weather conditions and the time of the year. The model demonstrates statistical significance and accuracy in predicting trips while identifying the insignificance of precipitation and high ambient temperatures. The methodologies, findings, strengths, and limitations of the two models were tested and compared to identify their relevant contributions. By contrasting their methodologies and evaluating their predictive performance, using performance metrics such as accuracy, precision, and recall values, this study provides valuable insights into the strengths and limitations of each approach. Limitations include assuming a stationary refueling process and excluding external factors and limitations related to data availability, as well as the absence of a specific focus on hydrogen-fuel vehicles. By understanding the differences and similarities between these two models, this paper aims to provide a unique perspective on gaps and further requirements for accurate prediction and modeling of refueling behavior to guide policymakers, infrastructure planners, and stakeholders in making informed decisions regarding the design and optimization of hydrogen refueling infrastructure.