Enhancement of Advanced Metering Infrastructure Performance Using Unsupervised K-Means Clustering Algorithm

Abstract

Data aggregation may be considered as the technique through which streams of data gathered from Smart Meters (SMs) can be processed and transmitted to a Utility Control Center (UCC) in a reliable and cost-efficient manner without compromising the Quality of Service (QoS) requirements. In a typical Smart Grid (SG) paradigm, the UCC is usually located far away from the consumers (SMs), which has led to a degradation in network performance. Although the data aggregation technique has been recognized as a favorable solution to optimize the network performance of the SG, the underlying issue to date is to determine the optimal locations for the Data Aggregation Points (DAPs), where network coverage and full connectivity for all SMs deployed within the network are achieved. In addition, the main concern of the aggregation technique is to minimize transmission and computational costs. In this sense, the number of DAPs deployed should be as minimal as possible while satisfying the QoS requirements of the SG. This paper presents a Neighborhood Area Network (NAN) placement scheme based on the unsupervised K-means clustering algorithm with silhouette index method to determine the efficient number of DAPs required under different SM densities and find the best locations for the deployment of DAPs. Poisson Point Process (PPP) has been deployed to model the locations of the SMs. The simulation results presented in this paper indicate that the NAN placement scheme based on the ageless unsupervised K-means clustering algorithm not only improves the accuracy in determining the number of DAPs required and their locations but may also improve the network performance significantly in terms of network coverage and full connectivity.