Abstract
Increasing the share of renewable energy sources in the electrical grid requires new balancing means. Therefore, activating flexibility at the demand side with Demand Response (DR) programs has gained significant attention in recent years. In this regard, Thermostatically Controlled Loads (TCLs) are interesting appliances, due to their inherent storage capacity. In this paper, double Q-learning is applied to an Electric Water Heater (EWH) in two different DR settings. The first assumes the EWH's energy consumption is billed according to a Time-of-Use (ToU) price. The second scenario assumes households have a power consumption limit. In both cases, the control policy learned with double Q-learning is compared to a thermostat controller. With ToU pricing the EWH consumes 35% less energy during peak hours. When the EWH is charged according to a capacity tariff Q-learning reduces power consumption limit surpasses by 31%.

This publication has 7 references indexed in Scilit: