Artificial neural network based cluster head selection process in wireless sensor networks

Abstract Wireless Sensor Networks (WSNs) face a significant challenge in extending network lifetime due to their limited energy capacity. This study proposes an energy-efficient, distributed cluster head selection method supported by artificial neural networks (ANNs) to enhance the energy efficiency of cluster-based sensor networks. The methodology allows each node to autonomously determine its optimal selection threshold, addressing the computational constraints of WSNs through a two-stage process: a Cascade-ANN is trained offline to model the complex relationship between a node’s energy status and the optimal threshold. Subsequently, its learned behavior is translated into lightweight mathematical equations via curve fitting, which eliminates the need for nodes to run the ANN directly. Under the evaluated simulation scenarios, the results suggest that the proposed method improves key performance metrics compared to the standard LEACH protocol, within the defined network parameters. Specifically, network lifetime was observed to extend by approximately 97% (from 1773 to 3497 rounds), and total data packets successfully transmitted to the sink showed an increase of approximately 57% (from 61,766 to 96,757 packets). Furthermore, a comparative analysis indicates that the proposed method performs favorably against similar recent studies, with a performance factor ranging from 3.63 to 6.26 in terms of network lifetime. This work aims to contribute a robust and scalable solution that may enhance network performance.