Experimental and data-driven optimization of vortex-induced vibration for marine energy harvesting

Marine current energy, particularly harnessed through vortex-induced vibration (VIV), offers a cost-effective method for renewable energy generation, especially in low-flow velocity environments. In practical applications, VIV harvester performance is influenced by a free surface, making its dynamic behavior challenging to predict, which differs from the extensively studied VIV phenomenon in an infinite water depth. In this study, we propose an alternative approach to predict the optimal VIV working conditions using an artificial neural network (ANN). Physical model tests with varying parameters were conducted, and an ANN algorithm established a nonlinear mapping between these parameters and the energy conversion ratio. Utilizing the established ANN model, we identify the optimal combinations of working parameters that maximize the conversion ratio of the VIV. Experimental validation of the predicted optimal working conditions is conducted, and our predictions agree favorably with experimental measurements. Our results demonstrate that the neural network model, trained with the experimental data, exhibits high prediction accuracy and strong fitting ability. It offers a practical direction for optimizing VIV working parameters and predicting vibration performance at low flow velocities with a free surface, enhancing the potential for efficient marine current energy harvesting. • Physical model tests explored marine energy harvesting via VIV under free-surface effects at low velocities. • An ANN model analyzed how working parameters affect energy harvesting efficiency. • Predicted optimal conditions align closely with experimental measurements. • Combining experiments and data-driven models aids VIV device design optimization.