Leveraging Regression Techniques to Optimize Lithium-Ion Battery Lifespan Prediction
Students & Supervisors
Student Authors
Supervisors
Abstract
Accurately predicting the lifespan of lithium-ion batteries is critical for effective battery management systems, ensuring reliability and timely maintenance. However, existing prediction methods often fail to deliver precise results during the initial phases of battery capacity degradation. This study introduces a predictive framework for estimating lithium-ion battery lifespan by leveraging advanced regression techniques to enhance battery performance and sustainability. Utilizing a dataset of 15,065 samples from NMC-LCO 18650 batteries, the study emphasizes robust data preprocessing, including feature extraction, normalization, and dimensionality reduction via PCA, to improve model accuracy. Regression models such as Linear Regression, Random Forest, and Dense Neural Networks (Dense NN) were trained and evaluated using Mean Absolute Error (MAE) and R-squared (R²) as performance metrics. Among these, k-Nearest Neighbors (KNN) and Random Forest models achieved the highest accuracy, demonstrating their reliability in lifespan estimation. The practical deployment of a Dense NN model using TensorFlow Serving underscores the real-world applicability of the proposed approach for predicting Remaining Useful Life (RUL). This framework facilitates proactive decision-making in battery lifecycle management, supporting energy efficiency and sustainability while advancing battery health management technologies.
Keywords
Publication Details
- Type of Publication: Conference
- Conference Name: IEEE Region 10 Symposium 2025 (TENSYMP2025)
- Date of Conference: 07/07/2025 - 07/07/2025
- Venue: University of Canterbury, Christchurch, New Zealand.
- Organizer: IEEE Region 10