A Long Horizon Data Augmentation Technique to Produce Reliable Synthetic Battery Parameters

Battery driven Electric Vehicles have created huge inroads into the conventional transportation sector over the past decade. Lowered cost and greater energy density numbers have encouraged many gasoline powered automakers to explore electrifying their future fleet of vehicles. State-of-Charge algo-rithm validation and inference has amassed a lot of commer-cial and research interest from the energy storage community. State-of-Charge essentially indicates the remaining range of the vehicle. Data driven techniques have enabled researchers to design accurate State-of-Charge computation algorithms but finite and restricted data availability is worrisome. This paper introduces a competent data augmentation technique which will help to generate error-free synthetic data. We show that such an approach greatly diminishes the limited data challenge. We release the code implementation of our work in the results section to allow reliable reproduction of our work by all battery data scientists. Code and data can be found at this URL: https://github.com/vageeshmaiya/Temporal-Fusion-Transformer