New Method to Improve Energy Savings in Wireless Sensor Networks by Using SOM Neural Network

Abstract

A new routing protocol for wireless sensor network is presented using energy clustering based on self-organizing map (ECSOM). In the process of neural network clustering, a self-organizing plan is used. Network input, three dimensions of energy and spatial coordinates of each node are denoted by X, Y and E, respectively. Connectional weights are the coordinates X, Y and E, which correspond to every unit of map, i.e., energetic nodes. The self-organizing neural network forms high-energy clusters based on the nodes in the network right after the training and reorganization of weights. In fact, each cluster is a combination of a high-energy node and the closest low energy nodes. Therefore, all clusters have almost the same energy level.