Prediction of virtual energy storage capacity of the air-conditioner using a stochastic gradient descent based artificial neural network

Abstract

Integration of renewable energy sources (RES) poses numerous challenges in smart grids, as RES are largely dependent on natural intermittency. Hence, it is necessary to rely on smart energy storage systems (SESS) to mitigate the fluctuation of RES on the supply side. SESS can be achieved by using demand response management (DRM), i.e., by aggregating thermostatically controlled loads using state-of-art smart grid technologies. In this paper, the air conditioners (ACs) are aggregated into a virtual energy storage system (VESS) by employing an electric model of the ACs. A simple mathematical model was described to evaluate the charging and discharging pattern of the load in terms of power and energy. Based on the VESS, the temperature control strategy was designed to reduce the power consumption of the ACs when making sure the room temperature is below the permissible limit. An artificial neural network (ANN) model was designed to predict the energy capacity of ACs, using which the VESS is achieved. Besides, the stochastic gradient descent (SGD) optimization algorithm was implemented in the back-propagation of the ANN model to achieve the optimum prediction values. Four case studies were presented to demonstrate the energy-saving capacity of ACs by regulating the set point temperature of a selected site. The results illustrate the change in the power consumption pattern of the ACs with and without employing them for virtual energy storage (VES). Hence, the proposed ANN model enables the aggregators to achieve the desired VES without affecting the comfort of the occupants.