Accuracy-Enhanced Multi-Variable LSTM-Based Sensorless Temperature Estimation for Marine Lithium-Ion Batteries Using Real Operational Data for an ORC–ESS

Abstract

Driven by increasingly stringent carbon emission regulations from the International Maritime Organization (IMO), the maritime industry increasingly requires eco-friendly power systems and enhanced energy efficiency. Lithium-ion batteries, a core component of these systems, necessitate precise temperature management to ensure safety, performance, and longevity, especially under high-temperature conditions owing to the inherent risk of thermal runaway. This study proposes a sensorless temperature estimation method using a long short-term memory network. Using key parameters, including state of charge, voltage, current, C-rate, and depth of discharge, a MATLAB-based analysis program was developed to model battery dynamics. The proposed method enables real-time internal temperature estimation without physical sensors, demonstrating improved accuracy via data-driven learning. Operational data from the training vessel Hannara were used to develop an integrated organic Rankine cycle–energy storage system model, analyze factors influencing battery temperature, and inform optimized battery operation strategies. The results highlight the potential of the proposed method to enhance the safety and efficiency of shipboard battery systems, thereby contributing to the achievement of the IMO’s carbon reduction goals.