ABSTRACT: Energy is a prime and compulsory source for development of any nation. Energy demand of Assam is increasing at an alarming rate and to compensate these loss energy is generated from many non-renewable resources, as a result Assam fails to predict and generate the future energy demand sustainably. This is due to many factors like rapidly rising population, literacy rate, industrialization, GDP and standard of living. Therefore predicting the future energy demand accurately will help in better decision making in implementing energy policies and provide information regarding future energy requirements and how to generate it sustainably. In this paper an attempt has been made to assess the effectiveness of statistical methods for forecasting energy demand in Assam. The data collected consist of several attributes that influence the energy consumption like Per capita income of Assam and Natural growth rate of Assam. Regression matrix plot is used to check the relationship between the Per capita energy requirement of Assam (demand) and the selected attributes. Traditional statistical forecasting methods like MLR, ARIMA, Winter’s model, Double exponential, Time series decomposition, linear trend, exponential trend and quadratic trend analysis are used to predict the future per capita energy requirement of Assam for a period of 10 years (2005-2016). The accuracy of the forecasted data is compared with the actual energy consumption data of Assam during the FY 2015-2016 and FY 2018-2019. The predicted results of ARIMA, Quadratic trend analysis and double exponential smoothing showed good accuracy for short term forecast whereas the predicted results of winter’s model with varying constants, time series decomposition, linear trend analysis and exponential trend analysis showed better accuracy for medium term forecast (2022-2025).
Keywords: Energy, Forecasting, Statistical methods, Accuracy