Abstract
Estimating a battery's state of charge (SoC) can become more difficult as a device, for example a mobile phone or vehicle, ages. This difficulty may be partly due to a potential circular dependency where SoC calculations may depend on the battery's state of health (SoH), and SoH estimation may in turn rely on an accurate SoC. A technique is described to address this challenge by decoupling these two estimations. A system can use a machine learning model to generate an independent SoH estimate based on operational data features, such as differential charge with respect to voltage. This SoH estimate can then be used to compensate or configure a separate, characterized battery model. A computation engine, which could use an algorithm such as a Kalman filter, can synthesize predictions from this compensated model with real-time sensor measurements to generate a refined SoC value. This approach may improve the long-term accuracy and consistency of battery charge reporting by mitigating potential compounding errors over a battery's lifespan.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Kang, Kang; Wang, Yu; Pelosi, AleX; Chen, Emil; and Pradhan, Ishu, "Battery State Estimation", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9422