K-Means++ Clustering Approach to Energy Efficient Data Aggregation in Wireless Sensor Network

• Main Authors: TING, SSU-JUNG, 丁思榕
• Other Authors: LIU, FANPYN
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/9knuc4

碩士 === 國防大學理工學院 === 資訊工程碩士班 === 108 === A wireless sensor network (WSN) has been designed and deployed in harsh environments to monitor contaminated, dangerous or hazardous areas. In this case, it’s difficult to replace or charge batteries for thousands of nodes, especially for the dangerous or hazardous areas. Therefore, extending lifetime of wireless sensor networks is one of the most critical issues in WSNs. The scalability of the clustering architecture is the ability of a wireless sensor network to its own expansion while improving energy efficiency. Since the cluster-head has a heavy workload than other nodes in the clustering of sensor networks, for the purpose of avoiding the node with low residual energy is elected as a cluster-head, which take residual energy and the spatial correlation between nodes into consideration. In this study, we leverage K-Means++ clustering algorithm to group the nodes and to select a cluster-head with respect to high residual energy, that is in charge of gathering the data from all the nodes in the cluster. In addition, we proposed an algorithm by considering the spatial correlation with nodes to select the working nodes, and the rest of the nodes of the cluster into hibernation save energy for prolonging the lifetime of the overall network. Through experimental for finding an optimal threshold value of residual energy of cluster-head is proposed for selecting a new cluster-head. This study first compares the results of an experimental result of two clustering techniques which are K-Means and K-Means++. Moreover, we also further carry out the network lifetime comparison on LEACH, LEACH-C and DEEC protocols with different round times: a) redundant data transfers from cluster-head to the sink node, b) the average residual energy consumption of the overall network, c) the energy variance of the overall network, and d) the lifetime of the overall network. We demonstrate that effectiveness of the proposed algorithm significantly outperform existing approaches especially in low energy consumption, and redundant data transfers which is able to dramatically prolong network lifetime.