Prediction and optimization models for integrated renewable-storage energy systems in mixed-use buildings

Mixed-use buildings contribute to the sustainable development of cities by providing economic, environmental, and social benefits. However, energy management of these buildings remains a challenge due to their unpredictable energy consumption characteristics and the lack of design guidelines for energy efficiency and sustainability solutions. This study aimed to develop and prototype a prediction model to characterize and forecast the energy consumption of mixed-use buildings and demonstrate its application by developing an optimization model to determine the design capacities of a proposed integrated renewable- storage energy system. The study applied machine learning techniques in developing and prototyping the prediction model, specifically k-means algorithm for clustering and support vector machines for forecasting. HOMER Grid software was used in developing the optimization model. The prediction model was initially developed on simulated data from OpenEI database and later prototyped using actual data obtained from The Building Data Genome Project, which was also used for the optimization model. The results of the study show that the prediction model had a performance better than statistical approaches previously developed in the literature and conform to building modeling standards. Improvements in model performance due to the novel integration of the clustering model were also observed in the initial model and specific cases in the prototype. Finally, the optimization results show that the proposed integrated energy system is viable and more economically attractive than stand-alone energy systems and the business-as-usual case for mixed-use buildings