Automated generation of models for demand side flexibility using machine learning

Abstract

Flexibility in consumption and production provided by distributed energy resources (DERs) is a key to the integration of renewable energy sources into the energy system. However, even for identical DERs, the flexibility can vary widely, based on local constraints and circumstances. Therefore, handcrafting models can be labor-intensive and automating the generation of models could help increasing the volume of controllable flexibility in smart grids. Depending on the underlying mechanism for controlling demand side flexibility, there are various ways how an automation can be achieved. In this paper, we discuss fundamental concepts relevant to the automated generation of models for demand side flexibility, give an overview of different approaches, and point out fundamental differences. The main focus lies on model generation by means of machine learning techniques.