Davidon-Herd-CapsSiam-ConvNet: A Novel Synergistic Model for Time-Series Wind Speed Forecasting

Accurate wind speed forecasting is essential for making wind energy generation more effective, keeping the grid stable and improving how things are run. Many models currently struggle with data that are noisy, features that are not well chosen and a lack of time-related modelling. To overcome these issues, this paper presents a new model called Davidon-Herd-CapsSiamConvNet (DHCSN) which combines advanced preprocessing, smart feature selection and effective deep learning architectures. The process starts by data collection, data smoothing and imputing the data and then extracting statistical, temporal, meteorological and frequency-domain features. To ensure robust feature selection, DHPO is used which combines Davidon-Fletcher-Powell (DFP) and Elephant Herding Optimization (EHO). For forecasting, the model uses CNNs, Deep Siamese LSTM and CapsNets to learn both the spatial and temporal relationships in wind patterns. The results of experiments on two benchmark datasets show that the model is highly accurate, with [Formula: see text] scores of 0.9895 and 0.9718, exceeding other models in both short-term and long-term forecasts. The results suggest that DHCSN can be used effectively and expanded for practical renewable energy applications.