Supercomputer Maps Bologna's Building Energy Use in Minutes

Buildings account for nearly 30% of global greenhouse gas emissions, making them a critical target in the fight against climate change. In cities, where buildings consume 30–70% of total primary energy, understanding energy use patterns is essential for planning effective retrofits and allocating public funds. A new study demonstrates how high-performance computing can rapidly model energy consumption across entire urban areas, providing a scalable blueprint for cities aiming to decarbonize.

The researchers developed a pipeline that estimates the energy demand of approximately 25,000 buildings in Bologna, Italy, in less than 30 minutes. By combining open geospatial data, building archetypes, and supercomputing power, they generated detailed simulations showing how energy consumption varies across different neighborhoods and construction periods. This approach allows city planners to identify which areas would benefit most from efficiency upgrades and assess the impact of various retrofit scenarios on a citywide scale.

Ology integrates multiple data sources to create a comprehensive model of Bologna's building stock. Geometric information, including building footprints and heights, was obtained from the Bologna Open Data portal and enhanced with aerial LiDAR measurements. Non-geometric attributes, such as construction materials and insulation levels, were derived from the European TABULA database and Italian building regulations. Buildings were grouped into eight archetypes based on their year of construction, ranging from pre-1900 to post-2005, to assign standardized thermal properties. The simulations were run using EnergyPlus, a widely used building energy simulation engine, and orchestrated with the Ray distributed computing framework on the Leonardo supercomputer, utilizing 1,120 CPU cores to handle the massive computational load.

Reveal distinct energy consumption patterns across Bologna's urban fabric. Older buildings, particularly those constructed between 1946 and 1960, showed the highest energy intensity, with broader and right-skewed distributions indicating greater variability and demand. In contrast, newer buildings built after 1975 exhibited narrower and left-shifted distributions, reflecting improved energy performance due to stricter codes. The simulation output, visualized in Figure 5, maps annual normalized energy consumption across the city center, highlighting hotspots where interventions could yield significant savings. A Pareto analysis indicated that about 70% of buildings account for 80% of total energy consumption, suggesting that efficiency policies need broad targeting rather than focusing on a few outliers.

This research has practical for urban planning and climate action. By enabling rapid scenario analysis, the model helps identify priority areas for retrofits, such as neighborhoods with older building stocks, and evaluates the potential energy savings from envelope improvements and window replacements. For instance, Figure 8 shows that retrofitting mid-20th century buildings offers the best consumption improvements, while newer buildings show minimal gains. The study also uses Pareto front analysis, as seen in Figure 9, to optimize trade-offs between the number of buildings refurbished and total energy savings, guiding decision-makers toward cost-effective strategies. This tool supports evidence-based policies that can reduce emissions and combat energy poverty, especially in low-income areas.

However, the study acknowledges several limitations. The model relies on archetypes and simplified assumptions about construction characteristics and occupant behavior, which may not capture the diversity of real-world conditions. The absence of detailed, building-specific data, such as precise construction years or past refurbishments, introduces uncertainties. Additionally, the lack of a proper calibration with real energy consumption data limits the quantitative reliability of ; the researchers compared simulations with TABULA data from Turin as a proxy, noting discrepancies due to climatic and ological differences. Future improvements should integrate anonymized consumption data and temporal dynamics to enhance accuracy.