ABSTRACT Wireless sensor networks (WSNs) are essential for modern data‐acquisition applications, but their long‐term operation is severely limited by the energy capacity of individual sensor nodes. Clustering algorithms are a key strategy for increasing energy efficiency. However, traditional algorithms, such as k‐means and low‐energy adaptive clustering hierarchy (LEACH), frequently ignore data redundancy, resulting in unnecessary energy consumption. We propose info‐kmeans, an information‐aware protocol that saves energy by dynamically scheduling sensor activity. Info‐kmeans quantifies the informational value of a node's data by combining historical readings and conditional entropy, relying on key parameters such as a historical sliding window, the number of spatial neighbours, data discretisation bins and a tunable information threshold. If the data are deemed predictable and redundant based on these parameters, the node is temporarily deactivated, saving significant energy. We evaluated info‐kmeans in a custom simulation platform against k‐means and LEACH. We demonstrate that info‐kmeans increases network lifetime by approximately twofold. Critically, it reduces cumulative reconstruction error by 15%–25% when compared to traditional clustering, demonstrating its ability to save energy whilst maintaining data integrity.
Abolghasemi et al. (2026) studied this question.