Abstract
Renewable energy sources (RESs) become increasingly popular for futuristic cities, and play a key role for the response to climate change. Hybidrizing the RSs, including solar energy and wind energy, improves the overall energy production, however, they are intermetent in their nature. This makes it challenging for energy scheduling and management due to unexpected weather conditions. The RSs forecasting is a critical tool for energy system planning, management, and operations. To this end, a demand-supply matching approach based on an accurate renewable energy forecasting and demand forecasting enhance the energy managment, and reduces the uncertainty. In this paper, we propose deep learning (DL) models, including recurrent neural networks, long short-term memory and gated recurrent unit, for their superior performance in the time series predictions. The DL models are trained and tested on real solar and wind datasets. The obtained forecasts are conducted for energy managment calculations towards a supply-demand matching approach that is applicable to any futuristic city with 100% renewables, e.g., NEOM in Saudi Arabia.