A wavelet based hybrid SARIMA-ETS model to forecast electricity consumption A wavelet based hybrid SARIMA-ETS model to forecast electricity consumption
Electronic Journal of Applied Statistical Analysis EJASA, Electron. J. App. Stat. Anal. Electronic Journal of Applied Statistical Analysis
This work is copyrighted by Università del Salento, and is licensed under a Creative Commons Attribuzione-Non commerciale-Non opere derivate 3.0 Italia License. For more information see: http://creativecommons.org/licenses/by-nc-nd/3.0/it/ Energy plays a fundamental role in the process of economic growth of a nation. Growth in the demand and consumption of energy is linked to economic output of a country as measured by gross domestic product. In view of formulating sustainable strategies in the
... energy service market, modelling and forecasting future demand of electricity is an integral part of decision support system of energy production in developed and developing world. Recent years have witnessed an increasing interest in providing prediction models of electrical energy consumption with greater accuracy. Besides the use of logistic and Harvey logistic growth curve models, the stochastic forecasting has been carried out using Box-Jenkins ARIMA, Holt-Winter, SARIMA and time series ANN models etc. However, their performance is far from perfect and it is especially true when the data contain complex nonlinear pattern and volatility. In this article, we propose a hybrid model that splits a time series into an approximate and a detailed component via discrete wavelet transform and then SARIMA and ETS models are used to fit and forecast the wavelet approximation and the detailed component respectively. Using the real monthly electricity consumption data from eight north eastern states of India during 2004-2015, we have developed the proposed model. Results of our investigation successfully demonstrates the higher degree of prediction accuracy of the proposed model than a data driven Box-Jenkins ARIMA model in terms of various performance accuracy measurement statistics and produces a substantial reduction in forecast errors.