Spatial-temporal modeling with DL-based hybrid cheetah hippopotamus optimizer framework for wind energy forecasting and turbine performance enhancement

An effective prediction and performance optivectormization of wind energy systems require advanced techniques to manage and analyze complex datasets. This study proposes a novel Integrated Deep Learning -based Hybrid Cheetah Hippopotamus optimizer (IDHCHO) framework that integrates data pre-processing, feature extraction, hybrid optimization-based feature selection, and deep learning models to enhance wind energy prediction and turbine performance. The process begins with pre-processing steps such as noise reduction, missing value imputation, and normalize the raw data for analysis. For feature extraction, both statistical methods (mean, variance, skewness) and Spearman's rank correlation for spatial features are employed. A novel hybrid optimization approach is based on the combination of feature selection using Cheetah optimization Algorithm (COA) with Hippopotamus optimization Algorithm (HOA). A combination of COA and HOA is used, namely, COA for local search and fast convergence, as well as HOA in order to explore diverse solution spaces efficiently. These features are utilized in a deep learning model using simultaneous feature extraction and temporal dependency capturing of prediction inputs by Convolutional Neural Network (CNN), Capsule Network and Attention-Recurrent Neural Networks (RNN), respectively. Their usefulness for energy production forecasting, anomaly detection, and system optimization via real-world wind turbine performance data proves the effectiveness of the proposed technique. The proposed method integrates the Cheetah and Hippopotamus optimization algorithms for feature selection in conjunction with advanced models of deep learning that provide a powerful solution to enhance the efficiency and resilience of the wind energy systems with the lowest value of 0.0254 & 0.0224 Root Mean Square Error (RMSE) error metrics on different datasets.