Energy Storage Technologies in Grid Modernization

6Machine Learning–Based SoC Estimation: A Recent Advancement in Battery Energy Storage System

Prerana Mohapatra*, Venkata Ramana Naik N. and Anup Kumar Panda

Dept. of Electrical Engineering, National Institute of Technology, Rourkela, Odisha, India

Abstract

An energy management system has become indispensable for a microgrid to manage different distributed energy resources in order to have a modern grid-connected system. To compensate for the intermittent nature of renewables and to ensure continuity in supply to the load, energy storage systems (ESS) especially battery energy storage (BES) have emerged for grid applications. The repeated charging/discharging cycles of the battery adversely affect its operational life which decreases the overall system reliability in the long term. This chapter concentrates on the management of the BES by estimating its state of charge (SoC). SoC estimation is an imperative metric to accurately estimate the available battery capacity. Recently, machine learning (ML) based estimation techniques have gained much attention as they can solve nonlinear modeling problems and their state estimation with great accuracy. In this study, ML techniques like support vector regression (SVR), and extreme learning machine (ELM) are investigated. To further improve the performance of ELM, regression analysis is also performed by using a penalty factor that reduces the error coefficient estimate toward zero. In this chapter, an extension of linear regression, ...