Designing a Robust Charging Infrastructure for Battery-Eelectric Buses with Sparse Energy Consumption Data

Abstract

This paper tackles the challenges of designing charging infrastructure for Battery Electric Buses (BEBs) with limited energy consumption data. Accurate energy consumption estimation is critical for cost-effective and reliable electrification but often requires costly field experiments, leading to limited data. To address uncertainty and data sparsity, we propose two models: a robust optimization model with box uncertainty, which relies solely on the data range, and a data-driven distributionally robust optimization model that uses observed data for more informed solutions. Our analysis of the Rotterdam bus network reveals three key insights: (1) Ignoring energy consumption variations can lead to unreliable designs. (2) Designing infrastructure based on worst-case energy consumption increases costs by 67% compared to using average estimates. (3) The datadriven distributionally robust optimization model is more cost-effective than the box uncertainty model while maintaining reliability, even when extreme energy consumption values are frequent.