A Novel Demand Response Mechanism in Smart Community with Learning-Based Neural Network Modelin

This thesis investigates how to develop a learning-based demand response mechanism for a novel smart community that can minimize the energy cost both for the smart community system operator and for the residents while meeting all requirements for all entities.;The thesis first proposes a demand response centralized energy management strategy for the grid-connected smart community with distributed energy resources. The results show that the energy management system can make proper day ahead schedule based on forecasted load and weather information, utility curtailment condition and user priority settings.;Then the thesis proposed a novel learning-based NARX (nonlinear autoregressive network with exogenous inputs) energy consumption model for the Smart Community. The mode consists of a certain size of smart homes all equipped with smart meters and typical controllable electric appliances. Different modeling methods for HVAC, electric water heater, cloth dyer, electric vehicle and energy storage system are analyzed. The performance of the NARX model shows the accuracy of applying NARX to Smart Community modeling.;After that, the thesis proposed a genetic algorithm based optimal method to solve the energy cost function. After simulation under different load curtailment conditions, the validation of the proposed method is proved as it can greatly lower the total energy cost from the grid.