Experimental validation of the demand response potential of residential heating systems

To effectively exploit the potential of demand response, knowledge regarding the availability of flexibility is crucial. In this paper, the flexibility of residential heating systems is assessed using the results of the Dutch smart grid pilot PowerMatching City. Within this pilot a multi-objective multi-agent system is used to exploit the flexibility of residential heat pumps and micro-CHPs. To validate the practical flexibility, a data-driven approach is proposed, in which the measured load is used to evaluate the effect of the smart-grid control signal on load changes. As this effect is subjected to other variables as well, the load is considered as a function of time, weather circumstances and the control signal. Two types of load forecasting techniques are applied to model the response of the load to these variables: multiple regression and Artificial Neural Networks (ANNs). Both forecasters obtain comparable results regarding the available flexibility of the devices. However, the results indicate that ANNs are slightly better at capturing the non-linearity of flexibility.