Digital twin application for reinforcement learning based optimal scheduling and reliability management enhancement of systems

Abstract

Increasing populations and economic expansion have substantially increased the energy requirements of residential consumers. Energy storage system (ESS) and distributed generation (DGs) are key tools for tackling this problem in smart homes. This study investigates the cost of electricity for residential consumers as a result of the combination of distributed photovoltaics (PVs) and ESSs for IoT-based smart home. Moreover, this paper examines energy management advantages due to bidirectional energy flow (H2G). In order to formulate the home energy management issue, PV and ESS end-user satisfaction limitations are taken into account. This study exploits a Q value-enabled reinforcement learning (RL) method to optimize home appliance scheduling (HAS) according to end-user priority. According to simulation outcomes, the suggested scheduling for household appliances performs well, and demand response (DR) measures have been implemented. It can be seen that the cost of electricity consumption as well as the uncertainty of the system have decreased in digital twin real-based application.