Hybrid Tunicate Swarm - Manta Ray Foraging Optimized Ensemble Learning for Photovoltaic Power Generation Forecasting

Predicting the amount of photovoltaic (PV) power generated accurately has been regarded as essential for maintaining grid stability and integrating renewables into electrical systems. To overcome such challenges, this article introduces a novel hybrid ensemble of Bidirectional gated recurrent unit (BiGRU), Bidirectional long short-term memory (BiLSTM), and Deep Neural Network (DNN) based on a Tunicate Swarm Algorithm (TSA) and Manta Ray Foraging Optimization (MRFO) hybrid algorithm. The proposed architecture addresses shortcomings inherent within standalone deep learning (DL) models that result from their inability to adequately weight each individual component's contribution towards overall accuracy, instead relying upon more dynamic techniques involving metaheuristically optimized weightings applied across various sources contributing toward final forecasts. In experiments performed against real world datasets representing actual conditions encountered within operational environments. The proposed approach achieves an impressive Mean Squared Error (MSE) of just 39729.9242, Root Mean Square Error )RMSE( of only199.3237 and a fantastic coefficient of determination (R²) value of 0.9939 which indicates extremely good performance levels relative even some top tiered alternatives currently available today. Such results indicate significant potential for applying similar strategies, paving the way for promising prospects regarding future developments in sustainable energy solutions.