Predicting Vehicle Refuelling Trips through Generalised Poisson Modelling

Abstract

This paper presents a model to predict the number of refuelling trips by vehicles on any given day considering weather conditions and time of the year. The predicted refuelling trips were founded on count-based data, i.e., data that contain events that occur at a certain rate. The paper presents an algorithm developed using Python programming language and the statsmodels module to achieve this. The results indicate that the GP-1 model developed in this paper is statistically significant at the 95% confidence level as it was able to converge—however, precipitation and high ambient temperature conditions are considered statistically insignificant in this model. The viability of the model was further tested on the remaining 20% of the data. Sensitivity tests indicate that there is a good correlation between the actual trips and predicted trips when 70% of the data are used to train the model. Overall, the model presented can be used to predict the number of trips taken by vehicles to refuel as well as model future trends, accurately. This model, can in the future, be applied to predict the refuelling behaviour of alternative fuel vehicles such as hydrogen fuel vehicles, when such data become available.