Cooling load prediction model during the design of mosques in Madinah

Abstract

The building sector accounts for almost 40% of the total global energy consumption. Saudi Arabia, along with other developed countries have expressed their concern on the increasing energy demand and established several related policies focusing on the building sector. Mosques are one category of buildings that consume huge amounts of energy above other public sector buildings such as hospitals. An extensive review of literature has revealed that there is an increased demand to build new mosques. The majority of previous research works focused on operational and maintenance stages. In term of energy reduction, not much can be done to existing mosques as the solutions are both costly and time consuming. The importance of making right design at the design stage has been stressed which may save up to 70% of total energy consumption. The literature also revealed that there is a gap and absence of design stage integration for mosque projects, due to the complexity of the stage, the lack of information, and limited support tools. This research aims to develop a prediction model known as the Mosque Cooling Load Prediction Model (MCLPM) to assist designers and local authorities in reducing the energy consumption of mosques during the design stage. The process began by identifying significant structural and architectural design parameters that influence the energy consumption of mosques, using a three rounds of Delphi approach with 33 local experts. Thirteen significant parameters were identified, of which mosque orientation was found to be the most significant. Integration between Rhinoceros/Grasshopper parametric model, EnergyPlus™ simulation of selected medium-sized mosques, and optimization through Genetic Algorithm (GA) and Galapagos were made to generate the dataset required for developing mosque cooling load prediction model based on the Artificial Neural Network (ANN) approach. Two thousand five hundred simulations were performed to achieve the optimum (approximately 58%) of total energy reduction, and 23 non-repetitive design alternatives with the least demand for cooling load were generated. The Mean Square Error (MSE) and correlation coefficient (R) were obtained for the developed ANN prediction model. Based on the findings, the least MSE and R values were at 6.27 * 1 0 _ 9 and 0.99888, respectively. Validation of the results revealed that the back-propagation strategy and Levenberg-Marquardt algorithm have the highest accuracies in predicting the exact total cooling load, in comparison to the actual values, and the absolute difference is less than 1%. The comparison with other methods and algorithms showed that the proposed prediction model has the highest accuracy, effectiveness, and least time required, to complete a given task. Hence, the developed prediction model act as a powerful tool to support the decision-making process that helps mosque designers provide a range of lowest cooling load design alternatives, thus facilitating the design process, and easy-quick estimation of total cooling load.