Modification and use of a machine learning algorithm to the Z-SEP clustering in WSNs

Wireless Sensor Networks (WSNs) consist of thousands of sensor nodes randomly placed to monitor and observe the state of the physical environment. The limited power of batteries, which are often non-replaceable or non-rechargeable, is crucial to the operation of WSNs. To extend the lifespan of WSNs, researchers focus on reducing the energy consumption of sensor nodes and developing energy-efficient routing protocols. Research has shown that threshold-based cluster head (CH) selection approaches in hierarchical routing protocols are effective in prolonging network lifetime. For heterogeneous WSNs, where some sensor nodes interact directly with the base station (BS) while others rely on clustering mechanisms, the Modified Zonal Stable Election Protocol (MZ-SEP) and Machine Learning Zonal Stable Election Protocol (MLZ-SEP) are proposed. These algorithms start with strategic node initialization, positioning nodes throughout the network and configuring their energy levels based on predefined parameters. Nodes are classified as 'Advanced' or 'Normal' to reflect inherent energy disparities within the network. A distance-based cluster head selection approach is used to assign nodes a probability of becoming cluster heads, considering their proximity to the sink node and the network's dimensions. Results indicate that MZ-SEP and MLZ-SEP are more energy-efficient than the existing protocols, showing improvements in throughput, average residual energy, network lifetime, and packet delivery to the base station. Specifically, MZ-SEP and MLZ-SEP achieve a 13 % increase in network lifetime, a 15 % boost in packet transmission to the base station, and a 12 % enhancement in average residual energy compared to existing protocols. These improvements demonstrate the potential of MZ-SEP and MLZ-SEP to significantly extend the lifespan of WSNs while optimizing data transmission and energy consumption.