Short-Term Forecasting of Total Energy Consumption for India-A Black Box Based Approach

Abstract

Continual energy availability is one of the prime inputs requisite for the persistent growth of any country. This becomes even more important for a country like India, which is one of the rapidly developing economies. Therefore electrical energy’s short-term demand forecasting is an essential step in the process of energy planning. The intent of this article is to predict the Total Electricity Consumption (TEC) in industry, agriculture, domestic, commercial, traction railways and other sectors of India for 2030. The methodology includes the familiar black-box approaches for forecasting namely multiple linear regression (MLR), simple regression model (SRM) along with correlation, exponential smoothing, Holt’s, Brown’s and expert model with the input variables population, GDP and GDP per capita using the software used are IBM SPSS Statistics 20 and Microsoft Excel 1997–2003 Worksheet. The input factors namely GDP, population and GDP per capita were taken into consideration. Analyses were also carried out to find the important variables influencing the energy consumption pattern. Several models such as Brown’s model, Holt’s model, Expert model and damped trend model were analysed. The TEC for the years 2019, 2024 and 2030 were forecasted to be 1,162,453 MW, 1,442,410 MW and 1,778,358 MW respectively. When compared with Population, GDP per capita, it is concluded that GDP foresees TEC better. The forecasting of total electricity consumption for the year 2030–2031 for India is found to be 1834349 MW. Therefore energy planning of a country relies heavily upon precise proper demand forecasting. Precise forecasting is one of the major challenges to manage in the energy sector of any nation. Moreover forecasts are important for the effective formulation of energy laws and policies in order to conserve the natural resources, protect the ecosystem, promote the nation’s economy and protect the health and safety of the society.

A Comparative Study of Time Series Forecasting Methods for Short Term Electric Energy Consumption Prediction in Smart Buildings

“…In [24], Samarasinghe and Al-Hawani compared multiple linear regression with Gaussian Processes on power consumption data from 2008 to 2010 to forecast the values in the next 24 h in Norway. In a more recent work, Rahman et al [25] proposed a high-precision methodology that included multiple linear regression and simple regression model along with other techniques to forecast the total energy consumption in India.…”

Energy Consumption and Trade Openness Nexus in Egypt: Asymmetry Analysis

“…After discussions of pollution literature, we review the literature with focus of determinants of energy consumption. For example, Rahman et al [21] investigate the determinants of electricity consumption in India and conclude that GDP is a better forecaster of electricity consumption comparing with to population and GDP per capita. Gomez et al [22] find that energy consumption is causing the economic growth in both linear and nonlinear causality analyses in Mexico.…”

Sequential Learning-Based Energy Consumption Prediction Model for Residential and Commercial Sectors

“…Kim and Cho [18] extracted features for energy consumption data using CNN and then forwarded these features to state expendable autoencoder for future consumption predictions based on 15-, 30-, 45-, and 60-min resolutions. Recently, Sajjad et al [28] proposed a hybrid sequential learning model for energy forecasting by integrating CNN and gated recurrent units (GRU) into a unified framework for accurate energy consumption prediction.…”

“…All the methods in comparison extract features using simple CNN and then forward the extracted features to different sequential learning models for energy consumption predictions. For instance, LSTM [23], auto encoder (AE) [18], Multi-layer Bidirectional LSTM [26], Bidirectional LSTM [25], LSTM followed by AE [17], GRU [28], and CNN with multilayer bidirectional gated recurrent unit (CNN-MB-GRU) [59]. For this comparison, we select the best-proposed model from Section 4.3, i.e., ConvLSTM-BiLSTM, which uses ConvLSTM as an encoder and bidirectional LSTM as a decoder.…”