State of Health Estimation of Lithium Ion Batteries using Recurrent Neural Network and its Variants

Numerous internal and external factors affect performance and capacity degradation of batteries over a period of time. SOH prediction of batteries becomes challenging task owing to unpredictable and unknown features which influence battery's health. This paper proposes a data-driven approach for SOH estimation by using the battery ageing datasets of Prognostic Center of Excellence (PCoE) of NASA. SOH estimation requires tracking of long sequential and temporal data of battery aging which exhibit dynamic states. The state of the art algorithm, Recurrent Neural Networks (RNN), due to its internal memory isappropriate for processing and predicting battery SOH. Hence this work employs different RNN techniques to build battery SOH prediction model, and the results of different techniques are compared and analyzed. The internal modeling parameters are trained by NASA battery datasets, where discharge cycles are introduced for SOH estimation. Experimental results show that RNN techniques can accurately estimate battery SOH.

I. Introduction

Lithium-ion batteries (LIBs) are widely utilized in many applications such as electric vehicles, energy storage systems, smart phones, home appliances and many more owing to high energy density, low self-discharge, high electromotive force, low voltage drop, high output voltage and relatively simple management [1]–[2]. However, irreversible chemical and physical changes occur during the use of LIBs leading to battery degradation and the batteries need be replaced once the battery reaches its End of useful Life (EOL) condition [3]. Accurate State of Health (SOH) prediction of Lithium-ion batteries can measure the reduction of capacity and growth of internal resistance of the battery. SOH estimation can improve the control performance of the battery management system which can assist in extending battery life and increasing safety of Lithium-ion batteries. Remaining useful life (RUL) and state of health (SOH) of the battery are critical issues of the battery management system (BMS) [4]. Thus, prediction of SOH and remaining battery life becomes significant as these parameters can largely determine the performance, safety and stability of lithium-ion batteries [5]. The SOH of battery suggests the correlation between aging and its internal parameters, like reduction of the capacity and growth of internal resistance [6].