To improve the performance and extend the operational lifetime of wireless sensor networks (WSNs), this thesis presents a novel two-stage hybrid approach based on energy efficiency and resource constraints. In the first stage, a GWO-based optimization is applied to the cluster head selection, which is assisted by K-means clustering to provide better starting solutions and increase the convergence speed. This form of hybridization increases the energy consumption efficiency along with improving the overall energy efficiency. The second stage focuses on optimizing the multi-hop routing between cluster heads. In order to achieve energy-optimized routing, PSO is combined with Prim’s minimum spanning tree (MST) algorithm. Since the implementation of PSO is computationally very intensive, the optimal paths generated by PSO are learned by an LSTM neural network. The weighted routing prediction inferred by LSTM enables independent PSO inference during routing without comprehensive optimization, eliminating computational overhead, thereby saving time and energy. The computational efficiency of the proposed GWO-LSTM method comes at a negligible cost to routing performance.

Wireless sensor. networks, particle swarm optimization, gray wolf algorithm,