A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models

Electrical energy demand forecasting plays a pivotal role as a decision support tool in the modern power industry. The focus of the paper is to propose a hybrid approach for the selection of the most influential input variables for the training and testing of neural network based hybrid models. The...

Full description

Main Authors: ul Islam, B., Baharudin, Z.
Format: Article
Institution: Universiti Teknologi Petronas
Record Id / ISBN-0: utp-eprints.19700 /
Published: UK Simulation Society 2017
Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85017241870&doi=10.5013%2fIJSSST.a.17.41.04&partnerID=40&md5=f2b4655edc51bd0191ea547eb4e4565e
http://eprints.utp.edu.my/19700/
Tags: Add Tag
No Tags, Be the first to tag this record!
id utp-eprints.19700
recordtype eprints
spelling utp-eprints.197002018-04-20T07:32:30Z A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models ul Islam, B. Baharudin, Z. Electrical energy demand forecasting plays a pivotal role as a decision support tool in the modern power industry. The focus of the paper is to propose a hybrid approach for the selection of the most influential input variables for the training and testing of neural network based hybrid models. The combined influence of the genetic algorithm and correlation analysis are used in this technique. The significance of the selected input variable vectors is studied to analyze their effects on the prediction process. Another objective of the study is to develop and compare the prediction models for electrical energy demand of one day-ahead. These models are developed by integrating multilayer perceptron neural network and evolutionary optimization techniques. Genetic algorithm and simulated annealing techniques are used to optimize the control parameters of the neural network. The results show that the neural network optimized with genetic algorithm and trained with an optimally and intelligently selected input vector containing historical load and meteorological variables produced the best prediction accuracy. Keywords - artificial neural network; mean absolute percentage error; genetic algorithm; simulated annealing; correlation analysisAbstract — Electrical energy demand forecasting plays a pivotal role as a decision support tool in the modern power industry. The focus of the paper is to propose a hybrid approach for the selection of the most influential input variables for the training and testing of neural network based hybrid models. The combined influence of the genetic algorithm and correlation analysis are used in this technique. The significance of the selected input variable vectors is studied to analyze their effects on the prediction process. Another objective of the study is to develop and compare the prediction models for electrical energy demand of one day-ahead. These models are developed by integrating multilayer perceptron neural network and evolutionary optimization techniques. Genetic algorithm and simulated annealing techniques are used to optimize the control parameters of the neural network. The results show that the neural network optimized with genetic algorithm and trained with an optimally and intelligently selected input vector containing historical load and meteorological variables produced the best prediction accuracy. © 2017, UK Simulation Society. All rights reserved. UK Simulation Society 2017 Article PeerReviewed https://www.scopus.com/inward/record.uri?eid=2-s2.0-85017241870&doi=10.5013%2fIJSSST.a.17.41.04&partnerID=40&md5=f2b4655edc51bd0191ea547eb4e4565e ul Islam, B. and Baharudin, Z. (2017) A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models. International Journal of Simulation: Systems, Science and Technology, 17 (41). 4.1-4.5. http://eprints.utp.edu.my/19700/
institution Universiti Teknologi Petronas
collection UTP Institutional Repository
description Electrical energy demand forecasting plays a pivotal role as a decision support tool in the modern power industry. The focus of the paper is to propose a hybrid approach for the selection of the most influential input variables for the training and testing of neural network based hybrid models. The combined influence of the genetic algorithm and correlation analysis are used in this technique. The significance of the selected input variable vectors is studied to analyze their effects on the prediction process. Another objective of the study is to develop and compare the prediction models for electrical energy demand of one day-ahead. These models are developed by integrating multilayer perceptron neural network and evolutionary optimization techniques. Genetic algorithm and simulated annealing techniques are used to optimize the control parameters of the neural network. The results show that the neural network optimized with genetic algorithm and trained with an optimally and intelligently selected input vector containing historical load and meteorological variables produced the best prediction accuracy. Keywords - artificial neural network; mean absolute percentage error; genetic algorithm; simulated annealing; correlation analysisAbstract — Electrical energy demand forecasting plays a pivotal role as a decision support tool in the modern power industry. The focus of the paper is to propose a hybrid approach for the selection of the most influential input variables for the training and testing of neural network based hybrid models. The combined influence of the genetic algorithm and correlation analysis are used in this technique. The significance of the selected input variable vectors is studied to analyze their effects on the prediction process. Another objective of the study is to develop and compare the prediction models for electrical energy demand of one day-ahead. These models are developed by integrating multilayer perceptron neural network and evolutionary optimization techniques. Genetic algorithm and simulated annealing techniques are used to optimize the control parameters of the neural network. The results show that the neural network optimized with genetic algorithm and trained with an optimally and intelligently selected input vector containing historical load and meteorological variables produced the best prediction accuracy. © 2017, UK Simulation Society. All rights reserved.
format Article
author ul Islam, B.
Baharudin, Z.
spellingShingle ul Islam, B.
Baharudin, Z.
A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
author_sort ul Islam, B.
title A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
title_short A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
title_full A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
title_fullStr A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
title_full_unstemmed A meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
title_sort meta-heuristics based input variable selection technique for hybrid electrical energy demand prediction models
publisher UK Simulation Society
publishDate 2017
url https://www.scopus.com/inward/record.uri?eid=2-s2.0-85017241870&doi=10.5013%2fIJSSST.a.17.41.04&partnerID=40&md5=f2b4655edc51bd0191ea547eb4e4565e
http://eprints.utp.edu.my/19700/
_version_ 1741196249015517184
score 11.62408

Similar Items