Energy analysis with deep learning in next generation buildings

As people's needs for lighting, heating, and comfort increase, so does the global demand for electricity. As technology advances in residential buildings and buildings serving as common areas, various types of electronic devices respond to this demand for comfort, and electricity consumption is increasing year by year. Studies on electricity consumption indicate that nearly half of the world's total electricity consumption comes from buildings. Public service buildings, shopping malls, hospitals, and hotels are typical examples of these high-energy-consuming structures. Hospitals are a particularly important building type because they house a wide variety of high-load devices and are critical to human health. While energy consumption in spaces like hotels and shopping malls can be assessed based on comfort requirements, in hospitals, it is a necessity rather than a desire for comfort. Therefore, analyzing the energy consumed by hospitals is a key step toward reducing the environmental impact of electricity consumption. This study conducted an energy analysis of a 55,000-square-meter green building in Istanbul. Deep learning techniques were used in the analysis. The existing dataset was analyzed using various methods used in building energy analysis, and classification and prediction were performed. Analyses of prominent parameters in the dataset were conducted using several methods, primarily the long short-term memory (LSTM) model. It was discussed which of these methods would be more useful for analysis in green building hospitals. Error rates for each method were compared. The results showed that the vanilla LSTM model yielded acceptable results in building energy analysis. The model proposed in this thesis was compared with previous studies in the literature. This comparison demonstrated that the proposed method is valid for green building types.