DEVELOPMENT AND VALIDATION OF A THERMAL NETWORK MODEL TO PREDICT INDOOR OPERATIVE TEMPERATURES IN DRY ROOFPOND BUILDINGS.

摘要

This article presents the development and validation of a thermal network model that can be used to predict the average interior operative temperature of a roofpond building based on outdoor climatic data. The study uses data collected from a roofpond test cell during the cooling season of 2009. Measured data for a 15 day period is used to develop the unsteady state heat-transfer model for the cooling mode operation of the test cell. The thermal network model implements a transfer function method with a time lag (At) of one hour in order to calculate the indoor operative temperature. The thermal model presented in this article has a moderately strong correlation (R2 = 80.19%) with the actual measured data. Since the indoor operative temperature is strongly influenced by outdoor air temperature and solar radiation, which are correlated with the hours of the day, residuals of the linear regression between calculated and measured indoor temperatures yields a strong daily pattern. A time series model, on the other hand, yields an improved correlation (R~2 = 92.79%). Though significantly improved, the residuals were not fully de-trended and the Durbin-Watson statistic is found to be 0.321. To further de-trend the residuals of the time series model an Auto Correlation Factor (ACF) and a Partial Auto Correlation Factor (PACF) test is conducted. The test demonstrates that an Auto Regression (AR) model would be the most appropriate to be able to de-trend the seasonality of its residuals. The new seasonal + AR model yields an R~2= 97.69%, with an RMSE of 0.144458 and a Durbin-Watson statistic of 1.74. This model was therefore found to be a good fit in predicting the indoor operative temperature for a roofpond building.