Multi-Objective Energy Management of Smart Buildings Using Neural Prophet and Reference Point-Based Non-Dominated Sorting Genetic Algorithm

The smart building energy management system (BEMS) has become essential for prosumers to manage their power consumption efficiently and conveniently, as building energy management evolves with the integration of distributed resources. This paper combines a reference point-based non-dominated sorting genetic algorithm (NSGA-III) with a Neural Prophet-based forecasting model to address the multi-objective BEMS problem and identify an optimal solution by considering the 24 h ahead real-time energy prices. The BEMS problem is designed as a multi-objective formulation that includes minimizing operating costs and alleviating user discomfort with their corresponding weights. To implement the proposed method, a building energy system is modeled, encompassing energy storage, electric vehicles, renewable energy resources, and various building loads. Multiple test cases conducted on residential and commercial buildings demonstrate significant cost savings in both types of buildings during both summer and winter days. A comparative cost analysis shows that relative to GA and NSGA-II, the proposed model reduced the reported operating cost by approximately 6.2% and 11.9%, and the execution time by approximately 30.3% and 14.4%, respectively. A real distribution feeder model with additional building energy system components is modeled on a digital real-time simulator using a reduced model approach to create a digital real-time system. The close agreement between the offline and digital real-time simulation results supports the feasibility of implementing the proposed framework under the operating conditions.