COMPARATIVE ANALYSIS OF ARIMA, ANN, ANFIS, LSTMAND THEIR HYBRID METHODS FOR PREDICTION OF SMALL DATA SET LONG TERM ELECTRICITY DEMAND OF ASSAM

Abstract

The demand for energy in Assam is rising at an exponential rate, and a lot of non-renewable resources are utilised in the process of producing it. Because of this, Assam is unable to predict its future energy demands and provide them in a way that is sustainable. This may be attributed to a number of factors, such as the rapidly expanding population, the literacy rate, industrialisation, GDP, and standard of living. Therefore, estimating future power demand accurately would provide information on sustainable energy generation and future electricity requirements, as well as improve decision-making throughout the implementation of energy policies. The data utilised for forecast in this article came from the Assamese statistics department and archives. The small amount of data that has been collected includes a number of factors that affect the state of Assam's electricity consumption, including the state's natural growth rate, consumer price index, GSDP, and per capita income. Traditional time series method ARIMA and Machine learning techniques likeANN, ANFIS, LSTMand their combination of hybrid models are used to predict the future per capita Electricity requirement of Assam. The performance of these methods is compared in order to determine which method is most useful for predicting future small data energy consumption.