Day-Ahead Load Forecasting of a Mosque using Data-Driven Models with Novel Predictors

Energy is the soul of modern societies as everyone deals with it when using phones, household appliances and transportation. Nevertheless, energy sources are finite and people need to use them efficiently with proper planning and regulations. In the last three decades, energy consumption increased dramatically due to population growth and the high usage of Air Conditioners (ACs). Saudi Arabia is one of the countries that had a massive increase in energy consumption in the last decade. Therefore, there is an urgent need at the state level to provide solutions that conserve energy and to lower the peak load during the day. These solutions usually include both Energy Storage Systems (ESS) and renewable energy integration. ESS would be more efficiently utilized when it is integrated with a load forecasting model, that's basically to predict the amount of energy to be charged during the off-peak hours in order to meet the demand during the peak hours. Thus, load forecasting models are of great importance for building operators and utilities. Building load forecasting has been studied extensively in the literature, however, there is lack of studies covering the load forecasting for the Mosque sector, which accounts for 2.5 GWh of annual energy usage across Saudi Arabia, with $3 Billion in annual cost. Typically, the load profile of the Mosque is highly variable and differs significantly from other typical load profiles. In this work, we propose a load forecasting model using new set of features, and targeted mainly for the Mosques. The proposed model was tested under five machine learning models and achieved an accuracy of 94%.