Search

The increasing prevalence of EV charging poses challenges for power grid stability and quality due to high charging load demands. Without effective energy management strategies for EV charging, the simultaneous power demand from numerous EVs can strain the electric grid, impacting power quality and the wholesale electricity market. To address these challenges, this dissertation presents a comprehensive framework comprising five critical tasks: analyzing EV charging behavior, optimizing charging schedules, developing predictive models, analyzing aggregated impacts, and evaluating implications of predicted user behavior on scheduling. By examining EV charging behavior at household and public charging stations, this study aims to understand patterns and variations in charging sessions. The framework introduces a centralized scheduling approach for household charging stations to reduce peak demand and costs, relying on accurate knowledge of EV charging behavior. Machine learning and linear regression models are utilized to predict session charging parameters, with Random Forest models outperforming other methods, yet uncertainties persist in the predictions. The study also investigates aggregate demand and connectivity of multiple EV users, revealing the potential to predict aggregate trends by incorporating session predictions. Evaluating the day-ahead scheduling framework implemented with predicted charging data using actual data highlights challenges in meeting user demand due to prediction errors. This research provides valuable insights into EV charging behavior, emphasizing the significance of accurate data for strategic scheduling. While machine learning models show potential in predicting EV charging behavior, limited correlation between session variables and available information at plug-in is observed. Challenges arise from errors in session predictions when scheduling EV charging for a group of users, suggesting the need for a more decentralized approach. Advisor: Mahmoud Alahmad

Plug-in electric vehicles (PEVs) play a significant role in the development of green cities since they generate less pollution than conventional vehicles. To promote PEV adoption and mitigate range anxiety, charging infrastructure should be deployed at strategic locations that are readily accessible to the public. Nebraska is working on the expansion of charging infrastructure around the state; however, stakeholders face several difficulties in trying to minimize irregular charging behaviors. Most electric vehicle users plug in and leave their vehicles for an extended time at public parking lots designated for PEVs. Some users even leave their vehicles for longer than 24 hours. Prolonged idle time is a concern for other PEV users who need to charge their vehicles to complete their planned trip. This thesis proposes several well-known regression methods to predict the idle time to help policymakers minimize the impact of irregular charging behaviors. In addition, PEV user charging behavior has a significant influence on the distribution network and its reliability. In addition, to increase efficiency in management of the electric grid, this thesis also proposes several well-known regression methods to predict the energy consumption of a charging session. The performance of different regression methods for predicting the idle time as well as energy consumption are characterized using established statistical metrics. Adviser: Mahmoud Alahmad