Session: 14-01-02: Smart and Sustainable Maritime Systems

Submission Number: 156104

Lithium-Ion Battery Degradation Forecasting Using Data-Driven Time Series Models 

The maritime industry faces significant challenges as it adapts from a major carbon emitter to a low-emission sector, with the goal of eventually achieving zero emissions. This transition requires innovative solutions for both new and old vessels,  lithium-ion batteries show promise in achieving these goals. However, their integration demands extensive research for ensuring safe and reliable operations of the vessels, especially when introducing new advanced energy storage technologies.

Battery management systems improve battery reliability and safety by gathering data such as voltage, current, and temperature through sensors to track the system's health and performance. These parameters enable prediction of remaining usable life, allowing for prompt maintenance and replacement before failure occurs. Remaining useful life prediction frequently uses time series models such as AutoRegressive Integrated Moving Average and its Seasonal - AutoRegressive Integrated Moving Average variants. These models predict trends and seasonalities accurately but fall short for non-linear battery degradation estimation, yet provide a strong foundation for predictive modeling. When integrated with Physics-Inspired Neural Networks, these models can help in understanding hidden patterns and trends, allowing for even more accurate battery health forecasts. Such predictive insights can help to transition from risk-based and corrective maintenance to predictive maintenance strategies that allow scheduling based on real battery conditions.

Lithium-ion battery applications in the maritime sector remain limited due to lack of long-term degradation data from ships, varying load profiles and harsh operational conditions.  However, publicly accessible lithium battery degradation datasets for a range of applications provide a useful starting point for model development. This study investigates time series modeling methodologies for Lithium-ion battery degradation, utilizing NASA’s battery degradation dataset to recreate observed patterns and forecast future states based on historical data using time series models.

Presenting Author: Kishan Patel German Aerospace Center (DLR)

Presenting Author Biography: Kishan Patel is a first-year Ph.D. candidate at the Technische Universität Berlin and a researcher at the German Aerospace Center (DLR) in Germany. He has a bachelor's degree in Mechanical Engineering and a master's in Simulation and System Design. His research focuses on component degradation study in the maritime industry, with particular focus on analyzing sensor data from lithium-ion batteries. Kishan is currently working with open source sensor-based degradation data and data-driven models combined with machine learning algorithms, to improve component life cycle predictions.