Intelligent Approach of Event Detection with Efficient Energy Consumption in Wireless Sensor Networks

Abstract

In the context of environmental protection against fire, this work presents a hybrid system of decision making and early warning applied in wireless sensor networks. This system, also, integrates an efficient data routing technique, based on the clustering of the near-event nodes, ensuring judicious network energy consumption. Data fusion technique, based on sensors data aggregated by the Cluster Head node (CH) within a defined analysis area, is processed by K-medoids, the latter will mainly contribute to increase the system's performances by decreasing the intra-cluster noise parameter () conducting to improve the probability of detection. This step, therefore, will distinguish and merge only the correct and useful samples. On the basis of the fused data, the estimated alert by K-Nearest Neighbours (KNN) can be directly triggered based on a minimum number of sensor nodes detecting fire; this will affect in advantage on the rapidity of detection, which leads to limit the spread of fire quickly. The alert is transmitted from the CH node to the base station via an intermediate node (IN) elected intelligently outside the cluster. This proposed approach proves, through its simulation results, a remarkable improvement of system performances in terms of information reliability, rapidity of detection and alert, avoiding false and redundant information, and also it improves extending the network lifetime.