ENERGY CONSUMPTION FORECASTING BY GREY PREDICTION MODELS AND ARTIFICIAL NEURAL NETWORKS IN A GREY SYSTEM: A COMPARATIVE ANALYSIS

Abstract

Forecasting energy consumption for short-term to make better decisions is becoming one of the basic needs for all industries. For this purpose many Machine Learning techniques are being used. More prominent and widely used techniques are Artificial Neural Networks and Grey Prediction Models. This research targets to benchmark both techniques for same system in order to provide the guidelines for selecting a suitable Machine Learning technique for short-term energy consumption forecasting. In this regard the energy consumption dataset from Remote Monitoring System installed at a facility is acquired and popular computation models from both techniques are implemented. An experiment with different variations is designed in order to explore multiple aspects of capabilities for both techniques. From experimentation results it is observed that the Artificial Neural Network is more suited technique if sufficient data for training is available as the case in this research. It is also observed that the performance of data is comparable while using small amount of data, which suggests that in absence of sufficient amount of data for training Grey Prediction Models should be preferred.