Uncertainty management in prognosis of electric vehicle energy system

Abstract

The body of work described here seeks to understand uncertainties that are inherent in the system prognosis procedure, to represent and propagate them, and to manage or shrink uncertainty distribution bounds under long-term and usage-based prognosis for accurate and precise results. Uncertainty is an inherent attribute of prognostic technologies, in which we estimate the End-Of-Life (EOL) and Remaining-Useful-Life (RUL) of a failing component or system, with the time evolution of the incipient failure increasing the uncertainty bounds as the fault horizon also increases. In the given testbed case, the life of the electric vehicle energy system is not measurable. It is only estimated, thereby increasing the importance of uncertainty management. Therefore, methods are needed to handle this uncertainty appropriately in order to improve the accuracy and precision of prognosis via shrinking the uncertainty bounds. To this end, this thesis introduces novel methodologies for the RUL prognosis then the enabling technologies build upon a three-tiered architecture that aims to shrink EOL/RUL bounds: uncertainty representation, uncertainty propagation, and uncertainty management.