<p>The Internet of Underwater Things is a rapidly emerging technology that enhances a diverse range of underwater applications. In IoUT, UNs have batteries that are difficult to remove and recharge, which requires a method to maximise energy efficiency. Thus, to provide energy-efficient communication. Energy-aware optimisation algorithms ensure resilient and efficient network operations. The purpose of this study is to model an efficient energy-aware approach using meta-heuristics and reinforcement learning approaches driven by quality of service for IoUT networks. The proposed framework addresses various energy optimisation problems in hierarchical solutions by optimising clustering management based on stable election protocol computations to develop an energy-aware stable election protocol optimised by integrated artificial bee colony and lightweight Q-learning approaches known as (ABCQL-EASE). Furthermore, the proposed approach is improved by a multi-criteria decision-making scheme to optimise inter-cluster routing decisions by ensuring that the CHs successfully relay data to the surface base station. The primary novelty of this study lies in three key contributions. First, a hybrid ABC-QL approach for dynamic cluster head management. Second, integration of an MCDM-based inter-cluster routing scheme that optimises data forwarding and QoS. Third, a fully integrated and self-adaptive architecture that balances exploration and exploitation while minimising computational overhead. The proposed approach is evaluated and compared to other previously developed methods, providing better performance in reducing energy consumption by 45%. It provides a reliable and sustainable method for obtaining energy-efficient data in IoUT networks compared with other existing approaches.</p>