Deep learning based real-time energy extraction system modeling for flapping foil

Kinematic Optimization of Energy Extraction Efficiency for Flapping Airfoil by using Response Surface Method and Genetic Algorithm

“…Zheng et al (2020a) reported that the optimal kinematic configuration parameters of the flapping airfoil can be efficiently detected with specific aerodynamic performance using the multifidelity Gaussian process regression and Bayesian optimization. Li et al (2022) have indicated that the aerodynamic characteristics and the physical fields of a flapping wing energy harvester are predicted with accuracy at a minimum computational cost by using deep learning based on real-time model founded in two modular convolutional neural networks. Zheng et al (2020b) have carried out a framework optimization method created on the datainformed self-adaptive quasi-steady model.…”

Wave power forecasting using an effective decomposition-based convolutional Bi-directional model with equilibrium Nelder-Mead optimiser

“…However, the superiority of that work [23] compared just with a few traditional machine learning methods. Recently, a two-modular convolutional deep learning model was hybridised by a gradient-based optimisation [24] in order to predict the aerodynamic attributes and the physical fields in a flapping foil, and the modelling results revealed that the two-modular CNN model was more effective than other predictive methods, especially in terms of computational cost.In another practical study, Zou et al [25] proposed a Deep Reinforcement Learning (DRL) model and simulated a point absorber with a direct-drive PTO for predicting the wave power. The authors reported a considerable improvement in the prediction accuracy from 24% up to 152% compared with other model-based controls.…”