Carbon, Cost and Capacity: Multi-objective Charging of Electric Buses

TL;DR

This paper tackles multi-objective scheduling for electric bus charging to curb CO2 emissions and grid peak demand under a deterministic, day-ahead planning regime. It formalizes a scheduling problem with decision variables $e_{i,j}$ and three objectives: $C(s)=\sum_i s(i)\textsc{CO}_2(i)$, $F(s)=\sum_i s(i)^2$, and a weighted combination $W(s,w_c,w_f)$ transformed for solver compatibility. Using a real Dutch bus-site dataset and 15-minute intervals, the authors show that CO$_2$-driven scheduling can reduce emissions by about 31% with modest peak reductions, while a pure flattening objective can cut peak loads by roughly 55% with ~20% emissions reduction; a weighted objective provides intermediate but substantial gains. They also demonstrate a notable flexibility potential when coordinating with larger baseloads, with synthetic baseload tests indicating a 57% peak-reduction opportunity, supporting grid-sharing with nearby assets.

Abstract

The public transport sector is in the process of decarbonizing by electrifying its bus fleets. This results in challenges if the high electricity demand resulting from battery charging demand is confronted with limited grid capacity and high synchronicity at bus charging sites. In this paper, we explore multi-objective scheduling for bus charging sites to minimize the emissions associated with charging processes and to aid the operation of the electricity grid by mitigating peak consumption. In particular, we discuss and validate optimization approaches for those objectives, as well as their weighted combination, based on data from a real-life bus charging site in the Netherlands. The simulation results show that compared to uncontrolled charging, power peaks can be reduced by up to 57%, while time-of-use emissions associated with the charging of electric buses are also reduced significantly. Furthermore, by using a synthetic baseload, we illustrate the flexibility potential offered by bus charging sites, and advocate that such sites should share a grid connection with other high-load assets.

Figures (4)

• Figure 1: Schematic of flow network structure for the $\textsc{CO}_2$ minimization for charging electric buses.
• Figure 2: The power profiles with data-based baseload.
• Figure 3: Metric results for the controlled scenarios for $w_c=1$ and various weights $w_f$.
• Figure 4: The power profiles with randomly sampled baseload.