Similarity-Based Optimization Framework for Curtailment Service Providers Through Collaborative Filtering and Generalized Dynamic Factor Model

We propose a novel demand response framework based on two similarities in building electricity consumption patterns for curtailment service providers (CSP)s to optimally manage the aggregated building energy. We find the first similarity in spatio-temporal correlations of building demand. To avoid conflicts of interests among buildings, we synthesize building demand and renewable energy scenarios based on that similarity by using the modified generalized dynamic factor model. We find the second similarity in hidden cost functions of discomfort indices for temperature and light. The cost function of a new building is extracted through the collaborative filtering based on the second similarity. Extracted cost functions allow us to align electricity and discomfort costs in the objective function and to cluster buildings to reduce computation time. Two similarities are used to improve the co-optimized day-ahead and real-time market decision process and respond to regulation signals through a two-stage stochastic optimization. Based on building energy statistics, we verify the optimality, efficiency, and comfortability of our strategy by showing that it has a lower cost, longer time period when residents feel comfortable, and lower computation time than existing strategies, particularly when a new building is aggregated.