Building Energy Performance Gap: Different Approaches for Mitigation

Abstract

The persistent Energy Performance Gap (EPG) in buildings poses a major obstacle to achieving predicted energy savings and sustainability targets. This study provides a comprehensive review of EPG root causes and mitigation strategies across the building life cycle, encompassing the design, construction, and operation phases. Technical and soft mitigation measures were synthesized, including smart monitoring, model calibration, machine learning, and stakeholder engagement.
Empirical investigations were conducted on a full-scale laboratory building, with EPG evaluations performed using four building energy modeling (BEM) tools and a TMY climate file for Lund. Results revealed discrepancies in simulated performance of up to 57% from 2020 to 2024, with anomalous increases in actual energy use observed in 2024. Subsequent analyses focused on thermal test cells, highly controlled environments monitored since 2017. Two BEM tools (IDA ICE and ClimateStudio) were used to develop calibrated models based on real-time 2022 data. Post-occupancy evaluations, actual weather data, and reverse-engineered envelope parameters were integrated into the models, reducing EPG from over 100% in baseline simulations to below 23%. A digital shadow framework was implemented to incorporate dynamic heater and thermostat behaviors using machine learning and reverse-engineered envelope properties, further narrowing the EPG to 4%. Verification with another year’s (2021) real-time data confirmed improved accuracy, with prediction errors reduced from 40%–67% to 11%–19%.
The study proposes a systematic, data-driven framework for EPG diagnosis and mitigation, demonstrating that post-occupancy evaluation and empirical modeling can significantly improve predictive reliability. The findings advocate for integrating real-time data streams, dynamic operation, and probabilistic prediction methods into future simulation tools. These insights support the advancement of digital twins and automated calibration workflows, contributing to more energy-efficient and resilient building operations.

Popular Abstract

Many buildings use more energy than expected, raising costs, emissions, and harm to the planet. Our research explores the energy performance gap and how smarter, data-driven models can help close it.
Imagine buying your dream home with a promised low energy bill, but once you move in, it doubles. Why? Buildings often consume far more energy than predicted. When buying or renting, you’re usually told how energy-efficient a building is, but actual usage can be much higher, sometimes even four times more. This difference is known as the Energy Performance Gap (EPG). For homeowners, this can mean hundreds of thousands of extra kronor over time, turning a dream home into a financial burden. It also has serious long-term effects on the planet.
In the European Union, buildings account for 33% of total energy use and about one-third of energy-related greenhouse gas emissions. Every kilowatt-hour of unpredicted energy use increases costs, accelerates fossil fuel consumption, and raises carbon emissions, making climate change harder to combat.
Our degree project explored this hidden but significant issue by reviewing 20 years of research to identify the root causes of the EPG. Key factors include inaccurate design model predictions, unrealistic assumptions about occupant behavior, and outdated weather data. We also examined solutions such as improving modeling tools and using real-time monitoring to reflect actual energy use.
To test these ideas, we studied a real building: the EBD Laboratory at Lund University. Using simulation software and energy data from 2020 to 2024, we found that predicted energy use was off by as much as 58%. The original models failed to consider real-world factors like building material performances, actual weather conditions, and mismatches in heating, cooling, and ventilation (HVAC) systems.
Here’s the exciting part: when models were updated using post-occupancy evaluations (real data on building use), detailed weather records, HVAC settings, and real measurements, the energy gap dropped to just 4%. The key? Moving beyond fixed estimates to smarter, data-driven models that better reflect how buildings truly operate.
One promising discovery was the use of real-time data analytics and machine learning to predict energy demand. These tools can spot patterns in how energy is used, but they work best when there is a lot of information and a good understanding of the building’s materials. Despite these challenges, the potential is huge. With smart sensors and clever software, buildings can be connected to digital versions of themselves. These virtual models could help architects, engineers, and owners to make better design choices and operate more efficiently, reducing costs and environmental impact.
This approach has the potential to transform how we understand and manage building energy use, and it could play a vital role in achieving global climate goals.

• author: Parvaz, Md Omor Faruk Monsi and Jakkula, Jitendra Pavan
• supervisor: Pieter de Wilde; Akram Abdul Hamid
• organization: Division of Energy and Building Design
• course: AEBM01 20251
• year: 2025
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Building energy modeling, Energy performance gap, Data-driven approach, Machine learning, Energy model calibration.
• language: English
• id: 9195444
• date added to LUP: 2025-06-12 13:56:24
• date last changed: 2025-06-12 13:56:24