Robust Battery Data Augmentation Techniques for Reliable State-of-Charge Estimation using Contrastive Learning

The energy storage market has expanded rapidly during the past ten years. Lithium-ion batteries, which have been at the centre of this tremendous advancement in storage systems, has been the driving force behind the electrification revolution in the automotive industry. The development of exact algorithms for calculating the State-of-Charge of various energy storage devices has been the main goal of research on energy storage. Calculating the remaining device usage time and range in electric vehicles is made easier by the use of state-of-charge. All State-of-Charge estimation algorithms have until now relied on carefully selected, large-scale labelled datasets. Researchers do not, however, have access to battery manufacturer datasets because of privacy restrictions. The process of recording battery measurements is also time-consuming and costly. The Contrastive learning mechanism is employed in this paper to produce synthetic battery datasets. This strategy is particularly helpful when there is a lack of data and researchers have little access to it. We employ Mean Squared Error (MSE) and Mean Absolute Error (MAE) metrics to evaluate the effectiveness of this method. The code related to this work can be found in the following URL https://github.con/vageeshmaiya/Contrastive-learning. We have released the code and data employed in this approach to help battery researchers reproduce and build on our results to produce reliable synthetic battery datasets.