The purpose of this dissertation is to investigate how intelligent algorithms can support electricity customers in their complex decisions within the electricity grid. In particular, we focus on how electric vehicle (EV) owners can be supported in their charging and discharging decisions, benefiting from the information available. We examine the problem from different standpoints and show the benefits for each involved stakeholder dependent on the market conditions.
In Chapter 2, we take the perspective of an individual EV owner and design an intelligent algorithm which learning from her preferences and driving and consumption information, proposes optimized charging and discharging recommendations. These recommendations are tailored to each individual EV owner and strive to satisfy her own preferences, while at the same time ensure financial benefits on the electricity bill. We observed that besides the EV owners, the proposed algorithm creates benefits for the electricity grid in the form of peak demand reduction. Specifically, when the preference heterogeneity increases, the peak demand is reduced significantly. This leads to an emergent charging coordination which results from different preferences and driving schedules.
In Chapter 3, we extend Chapter 2 by incorporating the EV within a smart home with a photovoltaic panel. The main goal of this study is to examine how accurate solar generation forecasting can be useful for charging the EV and make the best out of renewable sources. We propose a supervised learning algorithm which estimates the solar generation output from the weather conditions. We observe that the algorithm is capable of reducing the electricity costs on the customer side, since significant amount of EV charging is done with renewable energy and capable of increasing the levels of sustainability on the grid.
In Chapter 4, we examine the problem from the grid operator’s point of view, taking a top-down approach. We propose an auction mechanism that has as its main goal to service as many EV owners as possible, given a certain grid capacity. We show that an important parameter in scheduling the EV charging is the cost that each customer incurs from any potential delay. We prove the properties of the optimal EV charging scheduling and show that using the proposed mechanism reduces both the peak electricity demand and the overall delay in the grid.
In Chapter 5, we propose a hybrid mechanism which combines benefits from top-down and bottom-up approaches. This mechanism is based on dynamic price functions that are able to incentivize EV customers to delay their charging duration when there is no urgency. We show that the overall peak demand is reduced and that the herding effects, that might appear in traditional pricing schemes, are mitigated. Furthermore, the proposed mechanism is dynamic and learns from the EV customer portfolio, resulting to fast adaptability when the market conditions change.

Overall, this dissertation contributes to the academic literature with new algorithms that can leverage the power of data available and personalize EV charging recommendations. It also contributes to practice by providing useful insights to the involved stakeholders such as grid operators, energy utility companies, individual customers and automotive companies with respect to creating the right incentives for EV adoption. Finally, it adds to the very important discussion about sustainability, since it proposes ways to reduce carbon footprint and benefit the most from the available renewable sources.

Additional Metadata
Promotor H.W.G.M. van Heck (Eric) , W. Ketter (Wolfgang)
Publisher Erasmus University Rotterdam
ISBN 978-90-5892-450-6
Persistent URL hdl.handle.net/1765/93018
Series ERIM Ph.D. Series Research in Management
Grant This work was funded by the European Commission 7th Framework Programme; grant id fp7/261795 - Common assessment and analysis of risk in global supply chains (CASSANDRA)
Citation
Valogianni, K. (2016, June 30). Sustainable Electric Vehicle Management using Coordinated Machine Learning (No. EPS-2016-387-LIS). ERIM Ph.D. Series Research in Management. Erasmus University Rotterdam. Retrieved from http://hdl.handle.net/1765/93018