Feature selection and optimal neural network algorithm for the state of charge estimation of lithium-ion battery for electric vehicle application

M. S. Hossain Lipu, M. A. Hannan, A. Hussain

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

© Renewable Energy Research, 2017. This paper presents the estimation of the state of charge (SOC) for a lithium-ion battery using feature selection and an optimal NN algorithm. Principle component analysis (PCA) is used to select the most influencing features. Out of nine variables, five input variables are selected based on the value of eigenvectors. An optimal neural network (NN) is developed by selecting the hidden layer neurons and learning rate since these parameters are the most critical factors in constructing a NN model. The model is tested and evaluated by using US06 driving cycle at 25°C and 45°C respectively. In order demonstrate the effectiveness and accuracy of the proposed model, a comparative study is performed between proposed NN model and two different NN models (NN1 and NN2). The proposed NN model estimates SOC with lower mean squared error (MSE) and root mean squared error (RMSE) compared to two NN models which proves that the proposed model is competent and robust in estimating SOC. The simulation results show an improvement in proposed NN model accuracy over NN1 and NN2 models in minimizing RMSE by 26% and 22% and MSE by 45% and 39% respectively at 25°C.
Original languageEnglish
Pages (from-to)1701-1708
Number of pages1530
JournalInternational Journal of Renewable Energy Research
Publication statusPublished - 01 Jan 2017

Fingerprint

Electric vehicles
Feature extraction
Neural networks
Lithium-ion batteries
Eigenvalues and eigenfunctions
Neurons

All Science Journal Classification (ASJC) codes

  • Engineering(all)
  • Energy(all)
  • Energy Engineering and Power Technology
  • Materials Science(all)
  • Computer Science(all)
  • Signal Processing

Cite this

@article{6a4a26549bf74ffeb0fbd203dab2acce,
title = "Feature selection and optimal neural network algorithm for the state of charge estimation of lithium-ion battery for electric vehicle application",
abstract = "{\circledC} Renewable Energy Research, 2017. This paper presents the estimation of the state of charge (SOC) for a lithium-ion battery using feature selection and an optimal NN algorithm. Principle component analysis (PCA) is used to select the most influencing features. Out of nine variables, five input variables are selected based on the value of eigenvectors. An optimal neural network (NN) is developed by selecting the hidden layer neurons and learning rate since these parameters are the most critical factors in constructing a NN model. The model is tested and evaluated by using US06 driving cycle at 25°C and 45°C respectively. In order demonstrate the effectiveness and accuracy of the proposed model, a comparative study is performed between proposed NN model and two different NN models (NN1 and NN2). The proposed NN model estimates SOC with lower mean squared error (MSE) and root mean squared error (RMSE) compared to two NN models which proves that the proposed model is competent and robust in estimating SOC. The simulation results show an improvement in proposed NN model accuracy over NN1 and NN2 models in minimizing RMSE by 26{\%} and 22{\%} and MSE by 45{\%} and 39{\%} respectively at 25°C.",
author = "{Hossain Lipu}, {M. S.} and Hannan, {M. A.} and A. Hussain",
year = "2017",
month = "1",
day = "1",
language = "English",
pages = "1701--1708",
journal = "International Journal of Renewable Energy Research",
issn = "1309-0127",
publisher = "International Journal of Renewable Energy Research",

}

TY - JOUR

T1 - Feature selection and optimal neural network algorithm for the state of charge estimation of lithium-ion battery for electric vehicle application

AU - Hossain Lipu, M. S.

AU - Hannan, M. A.

AU - Hussain, A.

PY - 2017/1/1

Y1 - 2017/1/1

N2 - © Renewable Energy Research, 2017. This paper presents the estimation of the state of charge (SOC) for a lithium-ion battery using feature selection and an optimal NN algorithm. Principle component analysis (PCA) is used to select the most influencing features. Out of nine variables, five input variables are selected based on the value of eigenvectors. An optimal neural network (NN) is developed by selecting the hidden layer neurons and learning rate since these parameters are the most critical factors in constructing a NN model. The model is tested and evaluated by using US06 driving cycle at 25°C and 45°C respectively. In order demonstrate the effectiveness and accuracy of the proposed model, a comparative study is performed between proposed NN model and two different NN models (NN1 and NN2). The proposed NN model estimates SOC with lower mean squared error (MSE) and root mean squared error (RMSE) compared to two NN models which proves that the proposed model is competent and robust in estimating SOC. The simulation results show an improvement in proposed NN model accuracy over NN1 and NN2 models in minimizing RMSE by 26% and 22% and MSE by 45% and 39% respectively at 25°C.

AB - © Renewable Energy Research, 2017. This paper presents the estimation of the state of charge (SOC) for a lithium-ion battery using feature selection and an optimal NN algorithm. Principle component analysis (PCA) is used to select the most influencing features. Out of nine variables, five input variables are selected based on the value of eigenvectors. An optimal neural network (NN) is developed by selecting the hidden layer neurons and learning rate since these parameters are the most critical factors in constructing a NN model. The model is tested and evaluated by using US06 driving cycle at 25°C and 45°C respectively. In order demonstrate the effectiveness and accuracy of the proposed model, a comparative study is performed between proposed NN model and two different NN models (NN1 and NN2). The proposed NN model estimates SOC with lower mean squared error (MSE) and root mean squared error (RMSE) compared to two NN models which proves that the proposed model is competent and robust in estimating SOC. The simulation results show an improvement in proposed NN model accuracy over NN1 and NN2 models in minimizing RMSE by 26% and 22% and MSE by 45% and 39% respectively at 25°C.

UR - https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85043487236&origin=inward

UR - https://www.scopus.com/inward/citedby.uri?partnerID=HzOxMe3b&scp=85043487236&origin=inward

M3 - Article

SP - 1701

EP - 1708

JO - International Journal of Renewable Energy Research

JF - International Journal of Renewable Energy Research

SN - 1309-0127

ER -