Spatio-temporal analysis of power demand in smart grid based on stacked autoencoder

Energy demand in modern cities is becoming more and more complex. Knowing its pattern can help managers make proper decisions. However, finding the pattern of large-scale spatial coverage fields, such as a whole city, is not easy. In this paper, a data-driven stacked autoencoder-based framework is proposed to capture the spatio-temporal pattern for electricity consumption. The data can be organized as a one-time series and the geography coverage of the analysis can be determined by the collected data. An experiment with collected data proves the feasibility of the framework.