Development of Predictive Model for Building-integrated Rooftop Greenhouse using Artificial Neural Networks

The rapid growth of the global population and urbanization has significantly reduced the availability of arable land, threatening food supply chains. In response, urban agriculture has emerged as a viable solution to address future food security challenges while promoting sustainability in urban settings. Among various urban agriculture models, building-integrated rooftop greenhouses (BiRTGs) present a promising approach by utilizing underused building spaces and enhancing energy efficiency through dynamic energy exchanges between buildings and greenhouses. This study developed a predictive model to optimize the operational control of BiRTGs, focusing on how greenhouse control variables impact the heating and cooling loads of both the building and the greenhouse. An artificial neural network (ANN) was used to simulate the complex thermal interactions between the two structures. The ANN models were trained with simulation data from a validated TRNSYS model, and their predictive accuracy was assessed using performance metrics outlined in ASHRAE Guideline 14. The results indicated that the ANN models achieved high accuracy in predicting the heating and cooling loads of BiRTGs, establishing a strong foundation for the development of real-time optimization algorithms for these systems.