Water End Use Clustering Using Hybrid Pattern Recognition Techniques - Artificial Bee Colony, Dynamic Time Warping and K-Medoids Clustering

Abstract

The smart water meter collected data has made a great progress for the categorization of residential water end use events, the efficiency and accuracy still need to be improved. In this paper, an advanced algorithm is proposed for clustering the end-use category of a mechanical appliance. For this study, the database of end use events was collected using smart meters from over 200 households located in South-east Queensland (SEQ), Australia. Firstly, the raw data is pre-processed and physical characteristics (e.g., volume, duration, max flowrate, etc.) are extracted. Due to the type of the dataset is water end used flow data, which based on time series, a K-Medoids clustering algorithm based on the Dynamic Time Warping algorithm is used for clustering. In addition, a swarm intelligence which is named Artificial Bee Colony algorithm brings the whole system into equilibrium. Numerical experiments are based on toilet flushing events. Results indicate that the hybrid technique improves the clustering accuracy from 82.85% to 95.71%, and it can be implemented to other mechanical water end use events such as clothes washers and dish washers.