Surrogate-assisted nature-inspired optimization of tapered Savonius hydrokinetic turbine system using hybrid EPCO–HHO

River canals with established irrigation infrastructure offer substantial potential for decentralized renewable energy generation using hydrokinetic turbines. Despite the Savonius turbine's simplicity and self-starting nature, its low efficiency necessitates strategic design modifications. This study introduces a tapered-blade Savonius turbine integrated with a scoop-type deflector, numerically simulated in a section of the Tapi River canal using ANSYS CFX 2022 R1. The tapered geometry enhances interaction with deeper, slower flow regions, improving energy capture while maintaining cost-effectiveness. To model the turbine's torque output under varying conditions, a random forest regression (RFR) surrogate model is developed based on six key parameters: blade taper, minor and major radii of deflector curvature, angular orientations of the deflector sections, and its longitudinal position relative to the rotor. The surrogate model is coupled with a hybrid nature-inspired optimization algorithm, combining Emperor Penguin colony optimization (EPCO) for exploration and Harris Hawks optimization (HHO) for exploitation. Compared with conventional single-algorithm optimization or direct CFD-based parameter searches, the proposed surrogate-assisted hybrid framework significantly reduces computational cost while maintaining high prediction accuracy and enabling efficient exploration of the design space. This hybrid approach efficiently identifies the optimal turbine-deflector configuration aimed at maximizing torque. Final numerical validation confirms a torque output of 237.20 Nm (achieving 173.2 W power), with only 1.13% error, demonstrating the robustness and reliability of the proposed surrogate-assisted hybrid optimization framework for hydrokinetic applications in low-head water channels.