Towards enabling predictive optimal energy management systems for hybrid electric vehicles with real world considerations
Date
2021
Authors
Rabinowitz, Aaron, author
Quinn, Jason, advisor
Bradley, Thomas, advisor
Windom, Bret, committee member
Pasricha, Sudeep, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
In the pursuit of greater vehicle fleet efficiency, Predictive Optimal Energy Management Systems (POEMS) enabled Plug-in Hybrid Electric Vehicles (PHEV) have shown promising theoretical results. In order to enable the practical development of POEMS enabled PHEV technology, if must first be determined what method and what data is needed is for providing optimal predictions. Research performed at Colorado State University and partner institutions in 2019 and 2020 pursued a novel course in considering the widest range of possible data and methods of prediction currently available including a survey of all feasible Vehicle to Infrastructure (V2I), Vehicle to Vehicle (V2V), Advance Driver Assistance Systems (ADAS), and Ego vehicle CAN data streams with classical and novel machine learning methods. Real world vehicle operation data was collected in Fort Collins Colorado, processed, and used in the development of optimal prediction methods. From the results of this research, concrete conclusions on the relative value of V2I, V2V, and ADAS information for prediction, and high fidelity predictions were obtained for 10 second horizons using specialized Artificial Neural Networks.
Description
Rights Access
Subject
HEV
LSTM
PHEV
Kalman filter
ANN
optimal EMS