Please use this identifier to cite or link to this item:
|Title:||Data-driven health monitoring of energy storage systems||Authors:||Udayakumar Ashwin Kumar||Keywords:||DRNTU::Engineering::Electrical and electronic engineering::Electric power||Issue Date:||2019||Abstract:||Lithium-ion batteries have become an integral part of energy storage systems in modern electrical grids and the latest transportation systems. Battery health information is essential for the system decisions on energy storage’s optimal operation, control and maintenance. Usually, internal resistance and maximum available capacity are used for degradation modelling and remaining useful life estimation. However, for on-line applications, maximum available capacity is difficult to estimate in complex operating field conditions. Also, internal resistance measurement is too expensive to be implemented for on-line applications. The conventional methods also suffer from low accuracy and robustness under varying working conditions. In this dissertation, battery health is estimated using only the measurements available in the battery management system such as voltage or current. State of health (SOH) of a battery is estimated using data-driven methods. Two machine learning algorithms, Random Vector Functional Link and Extreme learning machine, were used for estimating the battery state of health, and their performance is compared. This machine learning framework includes raw feature extraction, box-cox transformation and correlation analysis to achieve enhanced performance. A battery dataset from NASA is used to illustrate the high efficiency in estimating the battery degradation.||URI:||http://hdl.handle.net/10356/78681||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Theses|
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.