Load forecasting using deep neural networks

Abstract

Short-term electricity demand prediction is of great importance to power companies since it is required to ensure adequate capacity when needed and, in some cases, it is needed to estimate the supply of raw material (e.g., natural gas) required to produce the required capacity. The deregulation of the power industry in many countries has magnified the importance of this need. Research in this area has included the use of shallow neural networks and other machine learning algorithms to solve this problem. However, recent results in other areas, such as Computer Vision and Speech Recognition, have shown great promise for Deep Neural Networks (DNN). Unfortunately, far less research exists on the application of DNN to short-term load forecasting. In this paper, we apply DNN as well as other machine learning techniques to short-term load forecasting in a power grid. The data used is taken from periodic smart meter energy usage reports. Our results indicate that DNN performs quite well when compared to traditional approaches. We also show how these results can be used if dynamic pricing is introduced to reduce peak loading.