Multi-condition optimization of S-Type hydrokinetic turbine driven by surrogate modeling and ocean current LSTM forecasting

Full-factor and multi-condition optimization of S-type hydrokinetic turbines offers a novel way to improving their performance. However, existing optimization studies often consider only a limited set of parameters or single inflow velocity, which constrains their effectiveness under real marine currents. To address this limitation, this study proposes a multi-condition optimization framework that integrates surrogate model with LSTM-driven current prediction. For current prediction, three LSTM models were compared to forecast annual current at a representative South China Sea site. Coupling the predicted current data with CFD simulations enabled a two-stage optimization process. The optimized turbine exhibits an aspect ratio of 2.6, a twist angle of 180°, and an overlap ratio of 0.125. Compared with the baseline turbine, the optimized turbine achieved a 19% increase in annual power output. At a typical velocity of 0.5 m/s, the power coefficient improved by 20.2%, and the mean static torque coefficient increased by 13%. Flow field analysis revealed that the performance enhancement arises not only from the synergy of three-dimensional parameters but also from the blade profile, which strengthens the low-pressure region on the advancing blade and thereby increases the turbine torque. This work provides a new methodological framework for multi-condition optimization of S-type turbines and marine current energy exploitation.