Reducing peak load of renewable energy at district level with predictive twins

Abstract

One of the major challenges in the energy transition of the built environment is how to integrate energy-producing neighbourhoods into the existing energy infrastructure. The aim is to avoid the peak load of local renewable energy by consuming it at district level as much as possible. The energy transition puts an increasing burden on the local energy grid, because of the increasing electrical load on the demand side by heat pumps and electrical vehicle charging, but also because of the increasing intermittent supply of energy through solar power and wind turbines. With a predictive twin, a digital representation at building and neighbourhood level, supply and demand can be better balanced, so more solar power is used locally and peak load is avoided. The predictive twin “SirinE”, developed by TNO, is a hybrid scalable model consisting of both a physical model for the building and installations, and an AI (Artificial Intelligence) model to describe the user behaviour. In the Horizon 2020 project syn.ikia, we are deploying this predictive twin in a model predictive controller (MPC) to use a temporary excess capacity of on-site solar energy as efficiently as possible. In this paper we present the model structure and the first simulation results for the energy prediction needed for the MPC.