Probabilistic Approaches to Enhance Safety and Energy Management of Energy Transition Components and Systems

Global initiatives to achieve net-zero carbon emissions by 2050 have spurred efforts to reduce reliance on fossil fuels, marking the beginning of a new era in energy transition. Advancements in renewable energy, zero-emission urban mobility, and smart manufacturing are crucial for climate action and sustainability. However, the widespread adoption of these innovations is challenged by several problems, including energy intermittency, complex degradation behavior, safety concerns, and lack of historical monitoring data. The goal of this research is to enhance probabilistic approaches for health and energy management of energy transition components and systems under uncertainty and limited data conditions, answering the research question: Can probabilistic approaches enhance safety and energy management for critical energy transition components and systems under limited data conditions? This problem is tackled in four different complexity levels of the energy transition process: Component, system, product, and factory. At the component level (1st Objective), a framework for lithium-ion batteries' remaining useful life (RUL) prediction is proposed. A machine learning model is trained to estimate reliability from small samples of battery sensor data, which are then converted into RUL values using entropy-based reliability models. Unlabeled sensor data is considered, introducing the concept of lack of temporal identifiers for battery RUL prediction. This approach minimizes data requirements while maintaining high prognosis accuracy. At the system level (2nd Objective), a safety-aware battery thermal management system (BTMS) for electric vehicles (EVs) is developed using physics-based probabilistic modeling to predict thermal-electrical behavior and validated with real-world EV sensor data. A novel safety metric quantifies battery-safe operating area, enabling uncertainty-resilient BTMS for EV safety in various environments. At the product level (3rd Objective), a health management strategy for electric vertical take-off landing aircraft (eVTOL) batteries is proposed in the context of product lifecycle management. This approach addresses the uncertainty associated with health performance metrics and RUL prediction, leading to more robust and safer eVTOL designs. At the factory level (4th Objective), a reinforcement learning-based energy management system is proposed to optimize production schedules for manufacturers integrating PV energy. The model considers time-of-use pricing, solar energy generation, and demand forecasts, ensuring cost-efficient operations while mitigating the intermittency challenges of solar energy. In summary, critical challenges in energy transitions were addressed through innovative frameworks for data-constrained RUL prediction, uncertainty-resilient battery safety management, and optimized solar energy integration. Together, these contributions pave the way for sustainable and reliable technologies in energy and transportation.