Dear Reader,

In the 61st issue, we present two recent studies: The first explores the techno-economic optimization of solar-assisted district heating systems, while the second introduces a novel surrogate model based on Graph Neural Networks for rapid simulation of complex district heating networks. For both we added a short summary below.

1) Comparative Techno-Economic Assessment of Solar District Heating Architectures

Braimakis et al.

What was investigated?

The study examines how the relative sizing of thermal components (solar collectors and seasonal thermal storage) versus electrical components (photovoltaic systems, heat pumps, and batteries) affects environmental performance (solar fraction) and economic performance (Levelized Cost of Heating – the lifetime cost per unit of heat). Three locations were analyzed: Athens, Munich, and Copenhagen.

Key findings:

• In Athens, systems without photovoltaics and batteries can achieve up to 100% solar fraction with large collector areas and diurnal or seasonal storage. In Munich and Copenhagen, the maximum solar fraction is about 50–60%, even with large collectors and seasonal storage.
• Heat pumps are not cost-effective in Athens due to high electricity prices. In Munich and Copenhagen, small heat pumps are economically viable for lower solar fractions and lower electricity prices.
• Adding photovoltaics and batteries only slightly increases the solar fraction, which is mainly driven by collector area. Economically, large PV installations are beneficial when electricity exports to the grid are allowed. Batteries have little economic impact but are necessary for achieving very high solar fractions.
• Without grid export, purely thermal systems (solar collectors and storage) remain the most cost-effective solution.

Why it matters

Solar thermal collectors and seasonal storage remain essential for achieving high solar fractions in district heating networks. Photovoltaics can improve economic performance when grid export is possible but do little to increase the solar fraction.

2) Graph Neural Networks for Fast Simulation of Meshed District Heating Networks

Boghetti et al.

What was investigated?

This work introduces a surrogate model that dramatically accelerates hydraulic simulations in complex district heating networks. The approach uses a Graph Neural Network (GNN) to predict mass flows in internal loops and combines a learnable hydraulic component with a physics-based thermal model.

Key findings:

• The surrogate model reduces computation time by nearly two orders of magnitude compared to traditional physics-based simulations. In tests with real-world data, it achieved over 97% accuracy in predicting pressure and temperature distributions.
• The architecture is topology-independent and scalable, enabling transferability to different network layouts.
• Coupled with an efficient thermal model, the method can simulate 12 days with 60-second time steps in under four minutes—fast enough for real-time optimization and large-scale scenario analysis.

Why it matters:

This approach opens new possibilities for operational optimization and design of district heating networks, especially where conventional simulations are too slow for real-time applications.

Further Information

As always, we recommend reading the articles in full. In addition, we will as always bring as much of the research work into our Digital Twin. If you are interested, we can recommend reading our monthly feature update in our blog, summarizing important developments and new features in our heatbeat Digital Twin.

The next issue of our newsletter will be published on December 3th, 2025. Until then, you can meet us at the following events https://heatbeat.de/en/events/ :

• 18.11.2025 – 19.11.2025: AGFW - Trafotage
• 25.11.2025 – 27.11.2025: HEATEXPO