Energy flows: An algorithmic scheduling approach to electric vehicle charging

Our energy system is transitioning away from a top-down fossil-fueled system towards a renewable-centric bidirectional system with volatile supply and demand that incorporates large synchronized loads. On top of that, the demand for electric energy is increasing as a result of among others the electrification of mobility and heating. While the need for grid expansion is undisputed, it is more sustainable, financially efficient, and timely to also invest into unlocking flexibility to mitigate peaks and match the volatile energy supply.

This also applies to settings where loads and local generation share a grid connection. In such settings, an energy management system (EMS) is required to unlock the flexibility of for example the charging demand of an electric vehicle (EV) fleet to mitigate power peaks, reduce asset degradation and increase self-consumption of locally generated energy.

The work presented in this thesis is motivated by one such emergent use case, namely an office environment with large-scale EV charging infrastructure in Utrecht, the Netherlands. The EV parking lot is integrated into an environment that also hosts a solar rooftop installation, a visitor's parking lot with another 50 charging spots and an office building with an electric heating system. Our analysis of the use case identifies information gaps and user acceptance as two major challenges for the successful deployment of energy management for such EV parking lots. While there are work-arounds to overcome those challenges in practice, they often rely on frequent re-optimization. Therefore, efficient optimization subroutines are crucial for an EMS to work properly.

Based on these insights gained from the field, we develop a mathematical formulation of the underlying (offline) EV scheduling problem. The main contributions of this thesis are the development of (constructive) algorithms for this problem, their theoretical analysis, and their validation using data collected at the mentioned real-world parking lot. On top of that, in this thesis we relate the EV scheduling problem to (i) processor speed scaling and (ii) majorized flows, and evaluate the relevance of the developed algorithms to these other mathematical problems.