Long-Term Energy Efficiency Prediction for Lithium-Ion Batteries Through Multi-Feature Fusion and Deep Learning

Abstract

The accurate prediction of battery energy efficiency is crucial for a wide range of critical applications, where reliable energy output is essential for system stability and mission success. Over time, lithium-ion battery energy efficiency declines in a nonlinear fashion due to complex electrochemical and physical changes that accumulate during repeated charge-discharge cycles. This study presents a direct multi-step prediction model for forecasting the long-term energy efficiency of lithium-ion batteries by integrating several aging-related features, including Voltage-dependent Capacity Variance, Initial CC Voltage, State of Health, and historical Energy Efficiency. The deep learning model, designed with specialized embedding and output layers, enables effective feature fusion and in-depth analysis of time-series data. Experimental results demonstrate that the model achieves high prediction accuracy across different battery lifespan stages, successfully capturing long-term energy efficiency degradation trends.

Keywords: Lithium-ion Battery, energy efficiency, long-term prediction model