Abstract
Ocean waves are a reliable and high energy density renewable resource in coastal regions; however, their utilization is still very limited by constraints of wave energy converters (WECs) and variable wave climate. Biomimetic wave energy converters (BWECs), influenced by the adaptive architectures of marine organisms, possess the capacity to address structural and efficiency challenges. Nonetheless, their design and optimization require highly precise predictions of wave conditions and device performance, which conventional numerical and statistical methods are incapable of achieving. Artificial intelligence (AI) and machine learning (ML) methods have shown great promise for overcoming these constraints, with studies investigating ensemble models, deep learning structures, time-series architectures, and reinforcement learning for forecasting and converter control. However, data-driven predictive modeling in relation to BWEC design optimization for the low-energy monsoon-affected Malaysian coastline is not extensively studied. This review systematically analyzes 231 articles published in Scopus, Web of Science, and IEEE Xplore (2010–2026) on predictive models for wave resource estimation, power production, hydrodynamics, and adaptive control. Input features, benchmark datasets, and current research gaps are also analyzed. Ensemble stacking models achieve R^2 ≥ 0.89–0.99 on benchmark WEC farm and buoy datasets; hybrid LSTM architectures achieve R^2 > 0.99 under single-site stationary conditions; and RL–MPC improves simulated power output by up to 185% over passive baselines — though generalizability to Malaysia’s monsoon-modulated climate remains unvalidated. Critical gaps persist in Malaysian model validation, Transformer architectures, and physics-informed neural networks. This review integrates BWEC design, AI/ML predictive modeling, and Malaysian coastal constraints as a unified framework.