State of Charge Prediction for Electric Aircraft Lithium-Ion Batteries with Limited Data by Stochastic Quantization and Informer Network

With the ongoing global energy transition, new energy electric aircraft offer a promising technological pathway for achieving green aviation. As the core power source of electric aircraft, lithium-ion batteries have their state of charge (SOC) as one of the most critical parameters throughout the battery lifecycle. Accurate SOC estimation directly impacts flight endurance, performance, and operational safety. Since new energy electric aircraft are still in the research and development stage in China and have not yet achieved large-scale industrial deployment, the availability of real flight data is limited. This data scarcity hampers deep learning models from fully capturing the dynamic feature patterns of battery systems under diverse and complex operating scenarios, thereby directly affecting SOC estimation accuracy. To address the issue of limited real-flight monitoring data, this study proposes a battery SOC prediction method with stochastic quantization-based data augmentation and Informer network. The approach enhances the original training samples through a stochastic quantization (SQ) algorithm, thereby expanding the data distribution, while leveraging the Informer network's strengths in time-series modeling to improve prediction accuracy under small-sample conditions. Experimental results demonstrate that this method achieves significant advantages in both data augmentation effectiveness and model generalization capability, providing a feasible solution to the problem of scarce battery data.