Deep Neural Network Model for Monthly Natural Gas Prediction

Abstract

Deep-learning networks have emerged as a powerful tool in the area of forecasting, greatly outperforming traditional methods in many applications, and have completely revolutionized some fields. In this paper, a deep-learning model has been proposed for forecasting monthly consumption of natural gas. Real-time natural gas dataset from an organization is used for validation purpose. A sliding window method for forecasting is used to feed the data into the network. We find definite advantages of using past observations as separate time steps of input feature. The performance of the model is compared against a second model that contains past observation as input features. Root-mean-square error (RMSE) and mean absolute percentage error (MAPE) accuracy measures are used to compare the performance of both models.