ANN-Based Scenario Generation Approach for Energy Management of Smart Buildings

Abstract

The growing trend of integrating renewable energy generation (e.g., PV) and electric vehicle charging solutions in buildings in addition to increasing electrification of demand (e.g., heating) requires an efficient Building Energy Management System (BEMS). BEMS is needed to manage the operation of controllable distributed energy resources (DER) in buildings according to the priorities of the building owner, such as maximizing own renewable generation utilization or minimizing the total energy costs or maximizing customer satisfaction and comfort-related needs. Related to the management of DER units, the uncertainty of renewable energy resources is a big challenge. To tackle the challenge resulting from uncertain renewable generation artificial intelligence can be employed to increase the prediction accuracy of intermittent weather-dependent generation. In this regard, an artificial neural network (ANN) model can be used to generate multiple scenarios, for example, for PV generation and solar radiation forecasting based on historical data. Moreover, a two-stage stochastic model can be deployed to model the operation of smart buildings in the day-ahead and real-time stages. Using the scenarios generated by the ANN-based approach and the two-stage stochastic model the optimal operation of a smart building can be conducted to schedule the operating point of different assets such as PV system, battery energy storage, electric vehicle, and space heater. The developed model assists smart buildings in reducing their energy cost and encourages them to deploy more renewable energy resources as well as electric vehicles, and to participate actively in the demand response program.