Buildings’ internal heat gains prediction using artificial intelligence methods

• Published in: Energy and buildings Vol. 258; p. 111794
• Main Authors: Liang, Rui, Ding, Wangfei, Zandi, Yousef, Rahimi, Abouzar, Pourkhorshidi, Sara, Khadimallah, Mohamed Amine
• Format: Journal Article
• Language: English
• Published: Lausanne Elsevier B.V 01-03-2022; Elsevier BV
• Subjects: Air conditioning; Artificial intelligence; Back propagation; Back propagation networks; Building; Buildings; Electric equipment; Energy conservation; Energy consumption; HVAC; HVAC equipment; Internal heat gains; LM-BP neural network model; Multiple regression models; Neural networks; Office; Office buildings; Parameters; Prediction models; Predictive control; Regression analysis; Residential areas; Residential buildings; Residential energy; Similar day model; System effectiveness; Thermal comfort; Office; Building; Similar day model; LM-BP neural network model; Internal heat gains
• ISSN: 0378-7788; 1872-6178
• DOI: 10.1016/j.enbuild.2021.111794

A large part of energy consumption in homes, offices and commercial spaces is related to Heating, Ventilation and Air-conditioning (HVAC) devices. The effective parameter on the consumption of HVAC systems is internal heat gains that arise from occupants, electric equipment and lighting. In order to reduce the energy consumption of these systems, internal heat gains should be predicted accurately. Since there are few investigations performed on the prediction of internal heat gains, in this paper, three predictive models, namely multiple regression model, Levenberg–Marquardt back-propagation (LM-BP) model and similar days method based on combined weights, have been deployed. By assessing the influential factors on internal heat gains, fundamental theories, structures, equations and parameters of these models are thoroughly proposed. To examine the prediction techniques, an office building in China was considered. It was found that all the proposed models have high accuracy; however, the LM-BP neural network showed the most precision among other models with RMSE = 15.59, MAE = 10.16 and MAPE = 6.35. This model had a higher agreement with the actual internal heat gains compared to the predetermined working programs in the ASHRAE standard 90.1. The proposed models used in this study can lead to providing a theoretical base for scholars and engineers to improve the predictive control of HVAC systems, which plays an important role in enhancing thermal comfort, saving energy of residential buildings.