A Data-Driven Algorithm for Estimating Battery State of Charge via Smart Plug-Based PCC Measurements

Estimating the State of Charge (SoC) of a battery is a complex task influenced by various chemical and environmental factors. Traditional SoC estimation methods often rely on high-frequency measurements recorded directly at the battery terminals and require detailed knowledge of battery models and parameters. These obstacles make them difficult to implement in practical scenarios. In this paper, we propose a novel algorithm that estimates SoC using measurements from the Point of Common Coupling (PCC), which is more readily accessible. Although PCC-based SoC estimation is inherently challenging—due to interference from chargers and other connected devices—our approach overcomes these limitations by treating the electric vehicle as a black box. The proposed method is model- and age-independent, making it adaptable across different battery types and usage scenarios. Our method explicitly uses a third-party measuring device named smart plug without requiring any information about the internal battery model or characteristics. We present a Convolutional Neural Network (CNN)-based algorithm that uses eight extracted features from smart plug measurements to estimate SoC with low error, offering a scalable and generalizable solution for real-world applications.