Chemical and Process Engineering 2014, 36 (2), 714-731 Special Issue 714 THE APPLICATION OF K-MEANS CLUSTERING ALGORITHM IN OPTIMIZATION OF ENERGY CONSUMPTION IN WIRELESS SENSOR NETWORKS Amin Ebrahimi Gandomani * , Farshad Kiyoumarsi, Siavash Mahmoudy Department of Computer, Isfahan Science and Research branch, Islamic Azad University, Isfahan, Iran Department of Computer, Shahrekord Branch, Islamic Azad University, Iran Department of Computer, Isfahan Science and Research branch, Islamic Azad University, Isfahan, Iran Abstract. A wireless sensor network is composed of a large number of small units called sensor nodes. Sensor nodes are generally equipped with sensor, processing, and communicative capabilities. The major duty of sensor nodes is collection of data at regular intervals, conversion of such data into electronic signals, and transmission of the signals into a sink node or Base Station. The most important reason for the emergence and development of wireless sensor networks is constant monitoring of the applications related to the environments where permanent access and presence of human beings is difficult or impossible. Consequently, dead nodes (i.e. nodes that have lost their efficiency due to the ending of energy resources) are impossible to re-charge or replace. Thus, two points bear significant importance regarding the efficiency of sensor networks. One is the lifetime of sensor networks and the other is the rage of coverage for such networks. The solution put forward to cope with such challanges is reducing the rate of energy consumption in sensor nodes, while at the same time keeping energy consumption in network nodes steady. In this study, two approaches (i.e., the data aggragation approach and the duty cycling approach) have been adopted to achive this goal. Following the data aggregation approach, the k-means hiararchical clusterning method is used to organize sensor nodes in such a way as to substantially reduce the degree of sending individual data to the central station, thereby reducing energy consumption in sensor nodes. In addition to the above-mentioned clustering method, a new parameter has been defined as the criterion of the similarity of the area covered for each node. Accordingly, the number of sensor nodes which are estimated to cover a common area, will be clustered evenly and thus, only one node in evenly distibuted sets will remain active at a time. In the end, * Corresponding author