Predicting Industrial Sector's Energy Consumption: Application of Support Vector Machine

Abstract

To follow in Europe's footstep in minimizing emission rate, much concentration should be on Africa's industrial energy consumption. South Africa's industrial sector is contributing immensely to the country's economic growth of which energy plays significant role. To reduce its consumption will mean planning accurately. Forecasting techniques have gained grounds when it comes to planning accessibility to energy demand to prevent incessant increase in emission rate. These techniques include traditional and machine learning techniques. This study applied support vector machine (SVM) of machine learning technique compared to regression analysis of traditional technique. The SVM is applied to forecast South Africa's energy consumption of five subsectors (manufacturing, basic nonferrous metals, basic iron and steel, non-metallic minerals and basic chemicals), with activity, structure and intensity as inputs whereas energy consumed was the output from 1970 to 2016. In contrast to the traditional technique, results confirmed SVM to be a better modelling system in terms of visual inspection (Figures 2 and 3) and statistical measures of performance in Table 1 (correlation coefficient, RMSE and RRSE).