Spatio temporal methods for prediction of power consumption in Indian railways

Abstract

Load forecasting is a challenging problem and many Machine Learning approaches have been used for the estimation of power consumption in the recent years. The Indian Railways is a vast railway network. It is divided into 18 zones and each zone is further divided into various divisions. The railway network is spread over a span of 1,21,407 kilometers all over the country. Due to the large spread of the railway network in the country, it is a challenging problem to estimate the power consumption for each zone. By applying a few traditional machine learning techniques such as Linear Regression, Neural Networks and LSTM; we model the power consumption of the Indian Railways. We further evaluate the performance of these models on real time data of the power consumption of the Indian Railways in a certain spatio-temporal zone.