Reducing transportation-related energy consumption and greenhouse gas (GHG) emissions have been a common goal of public agencies and research institutes for years. In 2013, the total energy consumed by the transportation sector in the United States was as high as 24.90 Quadrillion BTU. U.S. Environmental Protection Agency (EPA) reported that nearly 27% GHG emissions resulted from fossil fuel combustion for transportation activities in 2013. From a vehicle perspective, innovative powertrain technologies, such as hybrid electric vehicles (HEVs), are very promising in improving fossil fuel efficiency and reducing exhaust emissions. Plug-in hybrid electric vehicles (PHEVs) attracted most of the attention due to their ability to also use energy off of the electricity grid, through charging their batteries, thereby achieving even higher overall energy efficiency. At the heart of the PHEV technologies, the energy management system whose functionality is to control the power streams from both the internal combustion engine (ICE) and the battery pack based on vehicle and engine operating conditions, has been studied extensively. In the past decade, a large variety of EMS implementations have been developed for HEVs and PHEVs, whose control strategies may be well categorized into two major classes: a) rule-based strategies which rely on a set of simple rules without a priori knowledge of driving conditions. Such strategies make control decisions based on instant conditions only and are easily implemented, but their solutions are often far from being optimal due to the lack of consideration of variations in trip characteristics and prevailing traffic conditions; and b) optimization-based strategies which are aimed at optimizing some predefined cost function according to the driving conditions and vehicle’s dynamics. The selected cost function is usually related to the fuel consumption or tailpipe emissions. Based on how the optimization is implemented, such strategies can be further divided into two groups: 1) off-line optimization which requires a full knowledge of the entire trip to achieve the global optimal solution; and 2) short-term prediction-based optimization which takes into account the predicted driving conditions in the near future and achieves local optimal solutions segment by segment within an entire trip. However, major drawbacks of these strategies include: 1) heavy dependence on the a priori knowledge of future driving conditions; and 2) high computational costs that are difficult to implement in real-time.
