K-means based load estimation of domestic smart meter measurements

A load estimation algorithm based on k-means cluster analysis was developed. The developed algorithm applies cluster centres – of previously clustered load profiles – and distance functions to estimate missing and future measurements. Canberra, Manhattan, Euclidean, and Pearson correlation distances were investigated. Several case studies were implemented using daily and segmented load profiles of aggregated smart meters. Segmented profiles cover a time window that is less than or equal to 24 hours. Simulation results show that Canberra distance outperforms the other distance functions. Results also show that the segmented cluster centres produce more accurate load estimates than daily cluster centres. Higher accuracy estimates were obtained with cluster centres in the range of 16-24 hours. The developed load estimation algorithm can be integrated with state estimation or other network operational tools to enable better monitoring and control of distribution networks. This dataset provides details to the Input load profiles; Output load profiles; and Cluster centres which comprise the average active power demand (measured in kilo-Watts) at each half-hourly time step during a day. The dataset also includes the values of the Mean Absolute Percentage Error (MAPE) between the actual and the estimated values of the active power demand. A readme.txt file has been included in each folder to help the reader trace the type of information provided within the folders.