Intelligent Renewable Energy Prediction Using AE-CAM-BiLSTM with Elk Herd Optimizer

Introduction: Given that solar photovoltaic (PV) technology is highly dependent on weather and climate, its recent fast and unexpected growth poses a future test for electricity sector agents accountable for electricity distribution and organisation. In order to make effective policy decisions, maintain grid stability, and manage renewable energy sources sustainably, accurate consumption forecasts are required. Problems with feature extraction, learning from sequential dependencies, and data imbalance are common in conventional forecasting models. Objectives: The goal is to solve these issues. While CAM improves the selection of relevant features, the AE component efficiently extracts representations of high-level features. BiLSTM enhances sequential pattern recognition by capturing complicated temporal dependencies within data on energy consumption. Methods: Using the Elk Herd Optimiser (EHO) for optimisation and a modified Synthetic Minority Over-sampling Technique (SMOTE) for data balancing, this study presents a new hybrid model that combines Autoencoder (AE), Channel Attention Mechanism (CAM), besides Bidirectional Long Short-Term Memory (BiLSTM). To improve predictive accuracy, the EHO-based optimisation adjusts model parameters, and to lessen the impact of class imbalances, the modified SMOTE guarantees balanced training data. In comparison to more conventional ML and DL models, the suggested AE-CAM-BiLSTM model achieves much better accuracy, stability, and generalisability, according to the experimental results. Conclusions: This model shows great promise as a tool for renewable energy forecasting, as it can effectively learn from historical data on energy consumption. This will help decision-makers with energy planning and optimising resources. To further improve forecasting performance in dynamic energy environments, future research could investigate hybrid optimisation techniques, real-time adaptive learning, and the incorporation of external factors like weather and economic variables.