In a significant stride towards enhancing electric vehicle (EV) battery management, researchers have developed a novel framework that promises more accurate and reliable state of charge (SOC) estimations for lithium-ion batteries throughout their lifecycle. This advancement, published in the journal *Energy and AI* (translated to English), could have substantial implications for the maritime sector, particularly as shipping increasingly turns to electrification and hybrid solutions.
The study, led by Jinwei Xue from the School of Automotive Studies at Tongji University in Shanghai, addresses a critical challenge in EV battery management: the degradation of battery performance over time. As batteries age, their internal resistance and electrochemical properties change, leading to inaccuracies in SOC estimations. Traditional methods, whether data-driven or model-driven, often fail to adapt to these changes, resulting in uncontrolled estimation errors.
Xue and his team propose a closed-loop SOC estimation framework that integrates active learning with uncertainty-aware correction into deep learning networks. This innovative approach allows for real-time feedback on SOC prediction confidence levels without the need for additional sensors or reference data. “Our model-disagreement-based active learning correction strategy maintains robustness throughout the battery lifecycle,” Xue explains.
The researchers quantified the performance degradation of mainstream data-driven methods, including long short-term memory (LSTM) networks and Gaussian process regression (GPR), under complex aging paths. They demonstrated that their method reduces the average SOC estimation error to below 1.5%, and the maximum cycle-based average error to below 2% with only four active retraining sessions over the full aging process.
For the maritime industry, this research presents several opportunities. As ships increasingly adopt electric and hybrid propulsion systems, accurate battery management becomes crucial for safety, efficiency, and longevity. The proposed framework could enhance the reliability of battery systems on board, reducing the risk of failures and extending the lifespan of these critical components.
Moreover, the data-efficient nature of the method could lower maintenance costs and improve operational efficiency. “This work establishes a path toward uncertainty-informed, lifecycle-resilient, and data-efficient SOC estimation,” Xue notes, highlighting the potential for significant advancements in battery management systems for real-world applications.
The integration of active learning and deep learning networks offers a promising solution for the maritime sector, ensuring that battery systems remain accurate and reliable throughout their lifecycle. As the industry continues to embrace electrification, such innovations will be key to overcoming the challenges of battery management and paving the way for a more sustainable and efficient future.