Utilizing Variance Inflation Factor for Electricity Demand Forecasting

Abstract

The impact of climate change on electricity demand is an influential factor regarding load forecasting in Middle-East countries including Iraq. Selection of appropriate weather variables for prediction of electricity demand is crucial as it affects the accuracy and reliability of the forecasting. Recently, the slight upward temperature campaign with aired weather leads the trend of rising electricity demand as the dominate factors in Iraq. This is almost associated with air-conditioning loads. The factors relevant to the temperature such as maximum and minimum temperature and the average temperature are investigated in this paper. The paper introduces an efficient methodology of forecasting that consider the correlation among the different parameters involved in the forecasting. A statistical analysis is essential to reduce the data and retaining the independent variable sets that contribute substantially to the model of load forecasting. Hence, this paper discusses the potential problem of collinearity and multicollinearity among the variable sets that may create inflation in data which, in turn, creates a biased forecasting model. The variance inflation factor (VIF) and the variance–decomposition proportion (VDP) are financial tools used in economic studies and utilized in this study to build efficient forecasting models. This paper utilizes a real data set for 12 months in year 2018 in Baghdad city, central Iraq. A statistical analysis is implemented using MATLAB and Microsoft Excel to identify the sources of multicollinearity and validate the proposed methodology.