Systems and methods for accurately predicting state of charge in battery powered systems

Abstract

In one aspect, a computer-implemented may include receiving, from sensors, measurements comprising battery cell voltage, current, temperature, or some combination thereof. The method may include determining, via first computer-implemented models using the measurements, a state of charge prediction, wherein the first computer-implemented models are initially trained to determine the state of charge prediction using a first training dataset comprising data pertaining to a plurality of battery cells and are additionally trained to determine the state of charge prediction using a second training dataset comprising data pertaining to a target battery cell. The method may include transmitting parameters of the first computer-implemented models to an edge processing device to cause the edge processing device to execute at least a portion of the first computer-implemented models to predict the state of charge prediction and to display the state of charge prediction on a user interface.