This dissertation aims to explore some of the key barriers to realizing the full emission reduction potential of Plug-in Electric Vehicles (PEVs). Specifically, it explores the tradeoffs between Battery Electric Vehicles (BEVs) and Plug-in Hybrid Electric Vehicles (PHEVs) at reducing tailpipe emissions in the presence of an imperfect Electric Vehicle (EV) charging network. BEVs are true Zero Emission Vehicles (ZEVs) since they rely solely on energy from a battery (i.e., a rechargeable energy storage system) for propulsion, emitting no tailpipe emissions. In contrast, PHEVs can propel themselves using a combination of battery and internal combustion engine (ICE) energy. As such, driver behavior that determine the extent to which PHEVs’ electric range is used, have a strong influence on their energy use and emission potential. The first few chapters of this dissertation specifically explore the impact of the interaction between driver behavior and technical vehicle parameters on the energy consumption and Green House Gas (GHG) emissions of PHEVs in order to inform policy about the true emission potential of these low emission vehicles. Chapter 2 presents a study that aims to characterize the engine start activity profiles and emission potential of various PHEV models by examining the characteristics associated with engine starts, identifying the travel conditions that trigger engine starts, and determining the frequency of different types of starts. The study ultimately finds that long range PHEVs with high battery capacity such as the Chevrolet Volt are ideal for both curbing start emissions via initializing few engine starts and maximizing fuel displacement. Chapter 3 presents two studies that aim to understand the motivations and implications of driver mode, user-selectable drivetrain configuration setting, usage in PHEVs. In addition to comprehensively defining and classifying various drive modes, the first study examines the motivations for drive mode usage using a survey of over 26,000 PEV drivers in California. The second study quantifies the energy use and emission impacts of drive mode usage using on-road vehicle data from 81 Chevy Volts driven in California. Since BEVs aren’t equipped with ICEs, they are far superior to PHEVs at curbing tailpipe emissions. However, given the vehicles are solely powered by electricity, the adoption and acceptance of BEVs is tightly coupled with the quality of the EV charging infrastructure. As such, the scarcity of reliable and functional EV charging stations presents a significant barrier to the widespread adoption of BEVs. This dissertation aims to complement and expand the limited literature on EV charging reliability by examining the impact of EV charger reliability on BEV driver experience and developing a tool to help charging networks effectively meet impending reliability standards. Chapter 4 presents a study that focuses on understanding the impact of EV charger reliability on driver experience. It uses real-world EV charging data to simulate the level of disruption that would’ve occurred to EV drivers had their successful charging sessions been unsuccessful. Additionally, it quantifies how many charging sessions were actually unsuccessful and qualifies how disruptive those unsuccessful charging sessions were to drivers. By quantifying and qualifying the level of disruption associated with both real and hypothetical charge failures, it finds that EV chargers are not all equally important to EV drivers, highlighting the need for more nuanced charging reliability standards to more effectively meet consumer charging needs. Chapter 5 develops a tool enabling EV charging service providers to swiftly detect charge failures that cannot be detected by standard monitoring protocols. By analyzing habitual charging patterns of EV drivers, the tool identifies unexpected gaps in charger usage, indicating potential charger faults. The tool incorporates two anomaly detection models: a naive probability distribution-based technique and a LSTM for complex pattern modeling. Depending on the tool’s preferred confidence level, CPOs could’ve detected potential charging faults 1.5 to 3 times faster with the naive method and 1.5 to 2.4 times faster with the LSTM method.   The Risk and Resilience of Plug-in Vehicles in the Presence of an Imperfect Charging Network