Asynchronous Distributed IoT-Enabled Customer Characterization in Distribution Networks: Theory and Hardware Implementation

Abstract

This paper proposes and implements an asynchronous distributed IoT-enabled customer characterization framework to classify customers’ load consumption behaviors in electric distribution networks. More specifically, the proposed framework enables robust distributed clustering of customers’ electricity consumption habits in a highly scalable and interoperable framework. The proposed clustering method eliminates the need for hefty synchronization efforts typical to other distributed clustering algorithms. The theoretical foundations of designing the proposed framework are introduced. The orchestration of the application layers integrated in the proposed framework is demonstrated in an experimental implementation on a real-time network adopting a data-centric databus architecture. The results of the experimental implementation on IoT development boards demonstrate that the proposed framework can characterize customer categories with a 95% accuracy compared to classical centralized <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$k$ </tex-math></inline-formula> -means clustering, while ensuring seamless and interoperable peer-to-peer information exchange. This proved to be highly scalable and applicable to the real-world.