Abstract
Electric vehicles are poised to replace diesel and gasoline-based vehicles all over the developed world. The benefits are many, from reducing greenhouse gas emissions to reducing overall design complexity leading to lower repair costs. One central area of investigation in this regard is battery design and efficient operation. Estimating the battery’s health is essential in successfully adapting electric vehicles. Two critical parameters of interest in the battery management system are the state of charge (SOC) and a cell’s state of health (SOH). The state of charge depicts the battery capacity percentage. It informs the end user about the remaining performance in the current cycle. The state of health suggests the battery’s remaining life before it must be replaced. In this paper, a third-order resistor-capacitor equivalent circuit model is used to model a Lithium-ion battery’s dynamic behavior. This circuit model has been shown to accurately describe the behavior of the lithium-ion battery, with high concordance between simulation results and reality. The parameters of the model are estimated using an extended Kalman filter. Subsequently, utilizing the simulation thus created, a method is introduced to perform the estimation of state of charge and state of health using a recurrent neural network model with long short-term memory architecture. The paper then presents comparison results and identifies avenues for future work.
