Clustering techniques for data network planning in Smart Grids

Smart Grid, a modern approach to electricity distribution, requires innovation on various fronts. Communication is a key component of Smart Grid applicability. To satisfy Quality of Service (QoS) needs when deciding on network structure and topology, especially in urban areas, artificial intelligence techniques may be applied. Techniques such as clustering methods or genetic algorithms are useful to resolve this optimization problem. Choice of network topology for a specific electrical grid is also important. This choice influences the resulting QoS and network implementation price. This paper uses graph theory formulation to create a model. This model is designed to optimize topology of data network while accounting for delay constrains. Since this problem belongs to class NP-hard, this paper indicates appropriate clustering methods that best suit a given Smart Grid scenario in terms of QoS. A typical wireless network planning scenario of electric energy distribution is used as a case study. Each resulting cluster will contain a base station to attend the needs of Smart Grid Intelligent electronic devices in its cluster area. The algorithms K-medoids and K-means had the best performance with K-medoids bringing financial benefits regarding base station deployment.