Prediction of power demand in residential areas using the load profile clustering technique

The present-day advances in technologies provide the opportunities to pave a road from conventional power systems towards smart grids. As a result, smart grid features enable us to analyze the electricity usage data and identify electricity consumption patterns. This paper provides an analysis of half-hourly electricity consumption in domestic regions of the UK using clustering methods. To decrease the data dimensions and make it convenient to work with, unsupervised clustering methods such as k-means and Self-Organizing Maps are used for load profiling. The households are divided into several types and clusters, depending on the number of bedrooms and their daily electricity consumption patterns. Clustering is performed every day for different seasons providing intra-daily and seasonal variations. Probabilistic Neural Network is implemented to train the labeled dataset based on the clusters which identify load profile classes. The paper provides an investigation of the interconnection between house types and profile classes.