State of Charge Estimation of Lithium Iron Phosphate Battery Using Bidirectional Long Short-Term Memory

Abstract

Lithium iron phosphate batteries are currently popular in the electric vehicle market due to their high reliability and low price. However, due to the strong non-linearity of lithium iron phosphate open circuit voltage, it is difficult to estimate the state of charge with the traditional method. In this paper, a bidirectional long short-term memory model is used to accurately estimate the state-of-charge of a lithium iron phosphate battery in a usage environment such as an electric vehicle. A lithium iron phosphate battery charge/discharge test applying an electric vehicle driving cycle was preceded, and the state of charge estimation error was confirmed in the bidirectional long short-term memory model through the charge/discharge data. The mean absolute error of the bidirectional long short-term memory model was 1.80%, confirming the best performance among the deep learning models evaluated in this paper.