Rising global energy mandates innovation. The worlds reliance on limited fossil fuels is evident, however these are a limited resource. The integration of sustainable renewable sources in to existing energy networks is however challenging. Chef among these challenges being the sporadic and often unpredictable nature of energy production form renewable energy source that often rely on weather. This work seeks to provide a robust framework for solar energy production forecasting by applying metaheuristics optimizers to gated recurrent unit (GRU) networks to improve performance. A modified metaheuristic is introduced to improve hyperparameter tuning and simulations are conducted on a real-world solar energy dataset. The best performing models optimized by the proposed modified optimiser showcase admirable performance and suggest real-world applicability with the finest structures attaining a mean square error (MSE) of only 0.006951.