Artificial Neural Network-Aided Energy Management Scheme for Unlocking Demand Response

European Horizon 2020 raises three important targets-calling for 20% reduction green-house emission, 20% increase of renewable resources in energy sectors, 20% energy efficiency upgrade. Resilience to demand response (DR) scheme is commonly regarded as one of the key technologies to realize these targets across the world where 70% of the total DR potential is estimated from residential and tertiary buildings. The advantage of model predictive control (MPC) lies on its potential in shifting controlled demands towards strong generation or low-tariff periods which in turn encourage the change of energy consumption behavior of end-users under economic incentives. This research develops a smart energy management system based on MPC scheme by using dynamic program to formulate the optimization problems in the residential buildings. This paper presents a smart energy management system for unlocking demand response in the residential sector. The approach comprises the predictive estimation of one-hour energy demand and PV generation (supply) for rescheduling the 24-h ahead demand profiles by shifting potential flexible loads. Real-time electrical demand is met by combining power supplies from PV, grid and batteries while minimizing consumer's cost of energy. The results show that the peak-to-average ratio is reduced by 22.9% with the cost saving of 34.6% for the selected day.