Litcius/Paper detail

A Multivariate Time Series Approach for Forecasting of Electricity Demand in Bangladesh Using ARIMAX Model

Fazly Rabbi, Shihab Uddin Tareq, Md. Monirul Islam, Mohammad Asaduzzaman Chowdhury, Mohammad Abul Kashem

202011 citationsDOI

Abstract

This research attempts to forecast the yearly electricity demand of Bangladesh using a multivariate time series model. As the univariate time series cannot include the external factors, so we introduced two exogenous variables including Population and GDP per capita as exogenous variables to get better performance. The model has been developed on the yearly data collected from1994 to 2018. For the tested dataset, the Autoregressive Integrated Moving Average with Exogenous ARIMAX (0, 1, 1) model shows comparatively better performance than the state-of-art model with the lowest Akaike Information Criterion (AIC) values. The model has validated using the data from 2014 to 2018. The model shows Mean Absolute Error (MAE) 591.07, Mean Absolute Percent Error (MAPE) 5.43 and Root Mean Square (RMS) 782.28. Using this model, we forecast the energy demand for the period 2019 to 2021 and we found that the demand for electricity will be increased for each of every year.

Topics & Concepts

UnivariateAkaike information criterionMean absolute percentage errorAutoregressive integrated moving averageMultivariate statisticsStatisticsAutoregressive modelEconometricsTime seriesMean absolute errorMean squared errorPer capitaSeries (stratigraphy)PopulationMathematicsElectricityEngineeringDemographyPaleontologyElectrical engineeringSociologyBiologyEnergy Load and Power ForecastingForecasting Techniques and ApplicationsMarket Dynamics and Volatility