Deep COLA: A Deep COmpetitive Learning Algorithm for Future Home Energy Management Systems

A smart grid ecosystem requires intelligent Home Energy Management Systems (HEMSs) that allow the adequate monitoring and control of appliance-level energy consumption in a given household. They should be able to: i) profile highly non-stationary and non-linear measurements and ii) conduct correlations of such measurements with diverse inputs (e.g. environmental factors) in order to improve the end-user experience, as well as to aid the overall demand-response optimisation process. However, traditional approaches in HEMS lack the ability to capture diverse variations in appliance-level energy consumption due to unpredictable human behavior and also require high computation to process large datasets. In this article, we go beyond current profiling schemes by proposing Deep COLA; a novel Deep COmpetitive Learning Algorithm that addresses the limitations of existing work in terms of high dimensional data and enables more efficient and accurate clustering of appliance-level energy consumption. The proposed approach reduces human intervention by automatically selecting load profiles and models variations and uncertainty in human behavior during appliance usage. We demonstrate that our proposed scheme is far more computationally efficient and scalable data-wise than three popular conventional clustering approaches namely, K-Means, DBSCAN and SOM, using real household datasets. Moreover, we exhibit that Deep COLA identifies per-household behavioral associations that could aid future HEMSs.