Prediction of Indoor Climate and Long-term Air Quality Using a Building Thermal Transient model, Artificial Neural Networks and Typical Meteorological Year

The objective of this research was to develop a building thermal analysis and air quality predictive (BTA-AQP) model to predict indoor climate and long-term air quality (NH3, H2S and CO2 concentrations and emissions) for swine deep-pit buildings. The paper presents the development of the BTA-AQP model using a building thermal transient model, artificial neural networks, and typical meteorological year (TMY3) data in predicting long-term air quality trends. The good model performance ratings (MSE/S.D.<0.5, CRM≈0; IoA≈1; and Nash-Sutcliffe EF > 0.5 for all the predicted parameters) and the graphical presentations reveal that the BTA-AQP model was able to accurately forecast indoor climate and gas concentrations and emissions for swine deep-pit buildings. By comparing the air quality results simulated by the BTA-AQP model using the TMY3 data set with those from a five-year local weather data set, it was found that the TMY3-based predictions followed the long-term mean patterns well, which indicates that the TMY3 data could be used to represent the long-term expectations of source air quality. Future work is needed to improve the accuracy of the BTA-AQP model in terms of four main sources of error: (1) Uncertainties in air quality data; (2) Prediction errors of the BTA model; (3) Prediction errors of the AQP model, and (4) Bias errors of the TMY3 and its limited application.