The CSEM Digital energy solutions team has developed a unique data driven solution for quantifying demand response of these assets and for activating it in response to requests. Our solution is non-intrusive and scalable. It quantifies the uncertainty of the electrical consumption of the assets to mitigate dispatch risks and to comply with the requirements of the grid.

In this talk, we will present the main technical aspects of our solution. We will explain our approach of modelling both thermal and electrical assets as virtual batteries, whose baseline behaviour is learned using sample-efficient and uncertainty capable machine learning algorithms. For thermal systems based on heat pumps, we will explain how we use risk measures to quantify the asset flexibility in a robust manner. This quantification mitigates the unpredictability of the response to energy consumption requests and the robustness it affords is necessary to enter the secondary and tertiary control markets. We will discuss some of the challenges of extending this approach to the management of car charging. Finally, we will present the implementation of this solution through interoperable microservices and results of functional testing.