Deep Learning Based Short- Term Load Forecasting for Urban Areas

Abstract

This paper proposes a short-term load forecasting for residential applications. Deep learning is considered to be a powerful method in forecasting electricity load. As the present state of deep learning is still in progress, new updates towards improving the accuracy of the load forecasting are significantly important. Therefore, this paper proposes a restricted Boltzmann pre-training method and a rectifier linear unit method to enhance the current structure of Deep Neural Network (DNN) method. Moreover, making a priority list of factors that influence residential electricity consumption based on a location (London) is performed and analyzed in the paper.