Health and thermal focused feature modeling and optimal operation for Li-ion energy storage system under uncertainties

• Main Author: Xie, Jiahang
• Other Authors: Hung Dinh Nguyen
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/181569
• Institution: Nanyang Technological University

The increasing adoption of battery energy storage systems, particularly driven by the circular economy and the proliferation of electric vehicle applications, offers a promising solution to mitigate renewable energy penetration and distributed energy resource integration uncertainties and bring forward decarbonization in power systems. However, ensuring the safety and service lifespan of batteries while efficiently managing their operation remains a pressing challenge. The increasing level of charge/discharge current under highly concurrent energy usage and the exposure to extreme weather conditions continue to drive uncertainties in battery storage operations. Consequently, unprecedented risks in terms of insecure operations have been induced. The behavioral dynamics of Li-ion battery energy storage are generally complex and coupled among multi-physics. It is thereby difficult to estimate and control due to high nonlinearity. Considering the reduced applicability of traditional analysis and decision-making methods, this thesis presents a series of studies aimed to address the above issues, including battery storage system feature modeling, control design, and incorporation into cyber-physical system applications. This thesis focuses on developing novel systematic approaches for the evaluation of battery health status and optimal operation by exploring recent findings in Deep Learning. The proposed battery health-informed operation approach is developed for prolonging the battery’s remaining useful life based on data-driven techniques given the underlying aging characteristics. More specifically, the derived feasible regions with operational compliance facilitate computationally efficiency. The established framework can be further extended to power grid economic operations incorporating Li ion storage and to cover peer-to-peer energy procurement. Thereafter, physics-informed learning is leveraged to enhance the estimation of battery cell surface temperature and power trading for the aggregation of several battery storage members. The thermal safety of the battery can be better maintained. The individual energy storage is investigated, and also the hierarchical aggregator to regulate and optimize daily operation. Furthermore, the dual digital twin concept is demonstrated built on cloud-edge collaboration architecture to handle the time scale difference consisting of cloud full model analytics and edge lite model realtime control. The dual digital twin can be utilized in electric vehicles, which take a large share in transportation. The robustness of the algorithm to surrounding conditions is verified under various cases. This dissertation constitutes the foundation for intelligent real-time analysis of battery systems, spanning from individual devices to system-level applications. Its findings hold significant implications for safe and sustainable battery energy storage operation within power systems.