Hybrid energy systems, which integrate diverse energy sources including solar power plants, supercapacitors, UPS batteries, generators, hydrogen cells, and the grid, represent sophisticated yet highly promising approaches to enhancing energy efficiency, reducing operational costs, and supporting renewable and grid-independent initiatives. The inherent complexity of these systems necessitates the energy management strategy (EMS) capable of judiciously allocating resources in line with demand forecasts. A critical component of devising an effective task scheduling system within this framework is the ability to generate precise forecasts of energy production from renewable sources, solar power in this case. This paper showcases the deployment and comparative evaluation of two advanced deep learning models, Long Short-term Memory Recurrent Neural Networks (LSTMs) and Bidirectional Long Short-term Memory Networks (BiLSTMs), and our proposed Ensemble model, which averages the forecasts from LSTM and BiLSTM models, developed at our Laboratory for Energy Management ({LabE). Our primary goal is to predict solar power output for three days at 15-min intervals. Incorporating thirteen weather features, our findings reveal that the proposed models perform well in predicting energy production data, with the Ensemble predictions showing the best performance for 15-min interval forecasts spanning three days.