Comparison of electrical load forecast methods validated on the hourly consumption profiles of 10 heat pumps located in Switzerland

Abstract

The integration of variable renewable energy in the electrical grid necessitates the exploitation of load flexibility. The high contribution of space heating to the residential sector, thereby, puts the balancing of heat pumps in focus. The paper compares three different approaches to predict the heat demand of 10 Swiss residential buildings on 9 winter days. We compare two different simulation based grey-box models and an elementary statistical model approach. Unexpectedly, the statistical approach almost always performs better than the others, although the anticipated advantages of the simulation-based approaches could be seen in several cases. The comparison between the simulation-based methods shows that the one requiring more detailed, higher data quality clearly outperformed the simplified version. The work led to the conclusion that in the used setup too little system information was available to fully exploit the potential of a simulation-based approach at one hour resolution.