Wsn Cluster Head Selection for Energy Balancing Using a Hybrid Deep Learning Model

Abstract

Energy efficiency is a critical design consideration for Wireless Sensor Networks (WSNs), particularly in heterogeneous environments. To optimize energy utilization and enhance network longevity, this study integrates deep learning models into the cluster head (CH) selection process. The proposed methodology employs a Convolutional Neural Network (CNN) to analyze spatial attributes such as residual energy, node density, and proximity to sinks, and a Recurrent Neural Network with Long Short-Term Memory (RNN-LSTM) to capture temporal variations in energy depletion and workload distribution. The outputs of these models are combined through a hierarchical decision framework, enabling dynamic and adaptive CH selection. Comparative analysis with I-SEP and OptGACHE demonstrates that the proposed methodology maintains a residual energy of 9.89 J at a node density of 150, surpassing I-SEP and OptGACHE by 1.34% and 1.33%, respectively. Cluster lifetime is extended to 0.63 ms for 20 clusters, reflecting improvements of 16.7% over OptGACHE and 57.5% over I-SEP. Additionally, the packet delivery rate (PDR) reaches 450,000 kbps at a node density of 150, representing enhancements of 28.6% over OptGACHE and 50% over I-SEP. These findings validate the proposed methodology as a robust solution for energy-efficient CH selection, improving the sustainability and performance of WSNs in diverse operational conditions.