Author: Alexander Böhle
The idea for my Bachelor Thesis “Multi-Period Optimization of the Refuelling Infrastructure for Alternative Fuel Vehicles” originated from combining two of my favourite topics – The future of mobility and Operations Research.
While many consider alternative fuel vehicles (AFV) as an important mean to reduce greenhouse gas emissions, the lack of a sufficient refuelling infrastructure is a major barrier to the widespread adoption of AFVs. Although it is expected that an adequate number of alternative fuel stations (AFS) will eventually be constructed, due to the high resource-intensity of infrastructure development, an optimal step-by-step construction plan is needed. For such a plan to be actionable, it is necessary, that the underlying model considers realistic station sizes and budgetary limitations.
To address this issue, I have developed a new formulation of the flow-refuelling location model, that combines multi-periodicity and node capacity restrictions (MP-NC FRLM). For this purpose, the models of Capar et al., 2013 and Kluschke et al., 2020 have been extended, and the pre-generation process of sets and variables has been improved.
While adding a temporal component to a model might seem easy at first glance, it is far from obvious whether a multi-period model leads to a performance improvement that would justify the additional complexity. Hence, I adapted and applied the two evaluation concepts Value of the Multi-Period Solution (VMPS) and Value of Multi-Period Planning (VMPP) to assess the model’s relative additional benefit over its static counterparts. To better identify situations respectively parametric constellations where the multi-period model would be worth applying, I looked for potential drivers of the additional value of the multi-period model within the scope of a numerical experiment.
While the MP-NC FRLM has proven to provide additional benefit over static counterparts, it comes at the cost of a higher solving time. For determining realistic station sizes, the model could be further developed by adding deviation paths (= the drivers’ willingness to deviate from the optimal path for refuelling) and the stochasticity of demand.
Acknowledgement:
I would like to thank my mentor Hannah Bakker for the great support and Professor Stefan Nickel (both Karlsruhe Institute of Technology) for his openness to supervise my thesis project, even though it did not quite match the chair’s research focus. Finally, I would like to thank Professor Youssef Maknoon (Technische Universiteit Delft) for inspiring me with his excellent lectures during my exchange to write my thesis in Operations Research.