Performance of artificial neural network-based predictive controllers for thermal comfort in typical prefabricated movable building

Prefabricated Movable Buildings (PMBs) are gaining great attention in several applications, such as accommodations, offices in construction sites, disaster-reliefs, etc. A simple on-off strategy is often used for controlling air-to-air Electric Heat Pumps (EHPs) usually serving PMBs, leading to significant ‘‘overheating” (indoor air temperature exceeding desirable thermal comfort level) and energy waste. In this study, a reference PMB, intended for 3-person office use in construction sites, has been identified as representative of the PMBs available on the Italian market. The performance of the reference PMB served by an EHP have been dynamically simulated via the software TRNSYS 18 while operating under 4 different EHP control logics during heating season of Naples (Italy). In particular, a traditional on-off logic has been compared with 3 different strategies based on the prediction (over a period of 30 minutes) of indoor air temperature via Artificial Neural Networks (ANNs). The analyses have been performed with the main aims of assessing the capability of the proposed ANN-based predictive controls in improving thermal comfort by limiting overheating phenomena and reducing EHP electric consumption. The simulation results highlighted that the ANN-based strategies can reduce both overheating period up to 10.5% and EHP electric demand up to 5.3%.