Experimental study supported by ANN of new fabricated wall integrated with PCMs for energy saving and temperature uniformity of air-conditioned buildings

Abstract

• New Prefabricated Wall Integrated with PCM for Energy Saving and Temperature Uniformity. • Investigating several configurations of various PCM thickness and locations in walls. • Locating a thicker PCM layers close to wall external surface is best configuration. • Reduction of outer and inner surface temperatures in PCM-Prefabricated integrated wall. • 20 % daily energy saving in buildings with PCM-Prefabricated integrated wall. • Propesd the ANN model with high accuracy for predicting temperature profiles and energy savings without including ambient temperature in the training data set. The building of walls comprises numerous layers to protect against external environmental variations. Increasing layers through the integration of PCM into the wall as a separate layer improves daily temperature regulation inside buildings but reduces floor area, resulting in no economic justification. This study establishes a design approach for PCM walls by creating two wall configurations. Then, experimental testing and predictive modeling were conducted using Artificial Neural Network (ANN) trained by Feed Forward Back Propagation (FFBP) algorithm to assess thermal performance under different environmental conditions and various locations within the prefabricated wall. The thermal behavior of the new fabricated walls is tested and evaluated using external surface temperature, internal surface temperature, and average daily heat fluxes, in contrast to a reference wall. The results show that the configuration (2) is the best configuration for reducing outer surface temperature by 3 °C and reducing inner surface temperature by 0.53 °C at various outdoor temperatures. The location of PCM in prefabricated wall near to the outside wall is the optimum location in the summer season rather than the inside wall. The average daily energy saving decreases gradually with increasing the ambient temperature. in configuration (2), the energy saving is about 20 % higher than in configuration (1) at different ambient temperatures. Also, the artificial neural network model demonstrates high accuracy in predicting temperature profiles and energy savings without including ambient temperature in the training data set. It turns out that the predictive model agrees with the experimental tests, with the minimum and maximum relative errors for indoor and outdoor temperatures being 0.5 to 2 % and 2 % to 4 %, respectively.