An Instantaneous Growing Stream Clustering Algorithm for Probabilistic Load Modeling/Profiling

With the large-scale adoption of Advanced Metering Infrastructure (AMI), power systems are now characterized by a wealth of information that can be exploited for better monitoring, management, and control. On the other hand, specific techniques have to be employed to face the challenges brought by this large amount of data (Big Data). Traditional load modeling methodologies do not use the streams of data generated by AMI, providing static load profiles. In this work, an adaptive streaming algorithm is described to model any load through a Markov Chain. The proposed algorithm is able to cluster the load curves with a minimal computational effort, allowing realtime load modeling. The presented procedure’s performance is evaluated by experimental validation and compared with two reference methodologies (Dynamical Clustering and k-Means) in terms of accuracy and computational time.