Prediction of Thermal Environment in a Large Space Using Artificial Neural Network

Abstract

Abstract:Since the thermal environment of large space buildings such as stadiums can vary depending on the location of the stands, it is important to divide them into different zones and evaluate their thermal environment separately. The thermal environment can be evaluated using physical values measured with the sensors, but the occupant density of the stadium stands is high, which limits the locations available to install the sensors. As a method to resolve the limitations of installing the sensors, we propose a method to predict the thermal environment of each zone in a large space. We set six key thermal factors affecting the thermal environment in a large space to be predicted factors (indoor air temperature, mean radiant temperature, and clothing) and the fixed factors (air velocity, metabolic rate, and relative humidity). Using artificial neural network (ANN) models and the outdoor air temperature and the surface temperature of the interior walls around the stands as input data, we developed a method to predict the three thermal factors. Learning and verification datasets were established using STAR CCM+ (2016.10, Siemens PLM software, Plano, TX, USA). An analysis of each model's prediction results showed that the prediction accuracy increased with the number of learning data points. The thermal environment evaluation process developed in this study can be used to control heating, ventilation, and air conditioning (HVAC) facilities in each zone in a large space building with sufficient learning by ANN models at the building testing or the evaluation stage.

Performance and energy optimization of building automation and management systems: Towards smart sustainable carbon-neutral sports facilities

“…Optimization algorithms were utilized mainly for design (≈55%), and component sizing optimization (≈45%) (See Figure 13) as they promote solving for optimized settings or parameters of the evaluated problem given a defined objective. While ML and AI algorithms were primarily deployed for modeling (47%) such as developing prediction and estimation models for energy consumption/demand (e.g., [74,80]), thermal process (e.g., [79,84]), etc. Around 35% of the AI and ML-based works were developed for achieving energy savings by optimizing the facility management system.…”

“…Estimation, prediction, and modeling: In[79], a NN-based prediction model for the thermal environment in an ample space was presented for a stadium utilizing the information about the indoor environment and users. It facilitates the control of HVAC systems in stadiums and similar facilities.…”

Optimization of energy consumption and thermal comfort for intelligent building management system using genetic algorithm

“… The around room temperature value is (19) C˚.  The heat lost by infiltration is calculated by using the standards [UNI-7979/79], Where the coefficients (c_a) is the window topology assumed as (0.32) with designed model, and (c_b) is the local window condition, which is used in terraced houses as (14) [39].…”

Neural network-based model predictive control system for optimizing building automation and management systems of sports facilities

“…In contrast, in [22], a hybrid swimming pool's thermal model was developed combining thermodynamic laws and NNs. Additionally, a NN-based prediction model for the thermal environment of a stadium was developed in [23] utilizing the details of the indoor conditions and the users. In [14], an intelligent sports facility management system was developed employing a hybrid model using support vector machine-based backpropagation (SVM-BP) algorithm and a NN for predicting the passenger flow for an improved facility operation management.…”