Please use this identifier to cite or link to this item:
|Title:||Data-driven health estimation of Li-ion battery energy storage systems||Authors:||Yang, Yesen||Keywords:||Engineering::Electrical and electronic engineering::Electric power||Issue Date:||2021||Publisher:||Nanyang Technological University||Source:||Yang, Y. (2021). Data-driven health estimation of Li-ion battery energy storage systems. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/149564||Abstract:||Lithium-ion batteries are widely used in aerospace, electric vehicles, renewable energy, and other fields due to their excellent performance in various aspects such as energy density and cycle life. However, batteries would experience degrading and aging during the operation, which would affect the performance and safety. The estimation of state of health (SOH) of lithium-ion battery cells has become increasingly important. Generally, there are three types of methods to estimate the health status of Li-ion batteries: experimental methods[4, 5], model-based estimation methods and data-driven estimation methods[7, 8]. With the breakthrough of computing power and modern data measurement and storage capacity, the data-driven approach is highlighting more advantages and becoming popular in SOH estimations[9, 10]. Currently, many health estimation methods concentrate on the operations under certain current mode, for example, the constant current (CC), and extract the information from the CC curve instead of the original data. However, SOH estimation under dynamic currents is rarely mentioned. Aiming at the SOH estimation under dynamic operation profile, this thesis proposed a novel SOH estimation method, which contains 2 steps: health indicators (HIs) extraction and SOH estimation. For the first step, two potential extraction methods are studied, which are ECM-based and learning-based extractions. For the second step, multi-layers perceptron and model transferring are applied to improve the accuracy and generalization of the estimation. The dataset containing randomized walk operation from NASA is employed to train and test the performance.||URI:||https://hdl.handle.net/10356/149564||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Updated on Jan 18, 2022
Updated on Jan 18, 2022
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.