Introduction
Despite implementing common troubleshooting measures, including reducing simulation timesteps, lowering radiant fractions of internal gains, increasing wall thermal mass, and reviewing HVAC configurations, the instability persisted. A support ticket was raised with IES Technical Support, who indicated that the issue was likely associated with the room’s thermal characteristics and suggested modifying the Furniture Mass Factor to increase thermal capacitance within the space.
This experience highlights the fundamental differences between the simulation engines used by IES-VE and eQUEST and explains why a model that runs successfully in one software may encounter convergence issues in another.
Blog Synopsis
- Understanding the Data Center Modeling Challenge
- Why IES-VE Encountered Thermal Instability
- Why eQUEST Simulated the Model Successfully
- Lessons Learned for High-Density Data Center Modeling
- Key Takeaways
- Conclusion
Understanding the Data Center Modeling Challenge
The data center spaces contained exceptionally high internal loads generated by IT equipment. Cooling was provided through:
- CRAC (Computer Room Air Conditioning) units
- Supplemental liquid cooling systems represented in the model as chilled beams
Although chilled beams are not a direct representation of liquid cooling technology, this simplification was considered acceptable for LEED compliance purposes where detailed CFD-level analysis is not required.
Additionally, to model a chilled beam, eQUEST does not have a direct chilled beam option. Hence, chilled beams were modeled as induction units.
An additional consideration when modeling data centers in IES-VE is the chilled beam capacity limitation. For a single thermal zone, the chilled beam capacity cannot exceed 5,000 kW. In this project, the server room had a total server load of approximately 15,056 kW, which exceeded this threshold.
To address this limitation, the data hall was divided into three equal thermal zones. Each zone was assigned approximately 431.2 kW of cooling through CRAC units, while the remaining load was met by chilled beams sized at approximately 4,750 kW per zone. However, despite splitting the space into multiple thermal zones to comply with the chilled beam capacity limit, the room instability issue persisted in IES-VE.
This indicates that the instability was not solely related to chilled beam sizing constraints but was more likely associated with the interaction of extremely high internal gains, limited thermal capacitance, and the numerical convergence requirements of the Apache simulation engine.
When attempting to run system loads and annual energy simulations in IES-VE, the software reported thermal instability within the data halls and terminated the simulation. The issue remained unresolved even after applying standard recommendations such as:
- Reducing simulation timestep
- Setting radiant fraction of internal gains to zero
- Increasing wall thickness and thermal mass
- Reviewing HVAC sizing and airflow
Why IES-VE Encountered Thermal Instability
IES-VE uses the Apache dynamic simulation engine, which solves heat balance equations at every timestep while accounting for:
- Convective heat transfer
- Radiant heat transfer
- Thermal storage in building elements
- HVAC response
- Internal gains
In data centers, heat gains exceed several hundred watts per m² and may reach levels much higher than those found in conventional commercial buildings.
When a space contains:
- Very high sensible heat gains
- Limited thermal mass
- Aggressive cooling response
- Significant mismatch between generated heat and modeled cooling
the numerical solver may become unstable. Room temperatures can fluctuate rapidly between successive calculations, causing the simulation engine to fail convergence checks.
IES Technical Support specifically suggested increasing the Furniture Mass Factor, which effectively adds thermal capacitance to the space. Additional thermal mass acts as a buffer, reducing temperature swings and helping the numerical solution converge.
At the time of writing, feedback from the IES consultancy team is still awaited regarding the underlying cause of the instability issue and the effectiveness of the proposed Furniture Mass Factor adjustment. Nevertheless, it is hoped that future updates to IES-VE will address such challenges more robustly, particularly for high-density data center applications. Additionally, the development of dedicated guidance notes, FAQs, or best-practice documentation for modeling data centers would greatly assist practitioners in diagnosing and resolving similar simulation stability issues more efficiently.
Why eQUEST Simulated the Model Successfully
The same model ran without difficulty in eQUEST because the underlying simulation methodology differs significantly.
eQUEST is based on the DOE-2 simulation engine, which:
- Uses hourly calculations rather than detailed sub-hour dynamic calculations
- Employs simplified heat balance algorithms
- Relies on weighting factors and response factors that smooth thermal fluctuations
- Is generally more numerically forgiving for spaces with extreme internal gains
As a result, short-term temperature oscillations that may trigger instability in IES-VE are often averaged out in eQUEST.
This does not necessarily mean that eQUEST is more accurate. Rather, it means that DOE-2’s calculation approach is less sensitive to transient thermal effects and therefore less likely to fail due to convergence issues.
A question may arise regarding the variation in results between IES and eQUEST. However, the difference is minimal-less than 2%-despite the HVAC system being less precisely defined in eQUEST. This is primarily due to basic calculation logic.
In a data center, over 80% of the energy consumption is driven directly by server loads and their operating hours, which are calculated consistently in both IES and eQUEST. Lighting loads also remain identical across both tools.
Since the HVAC systems consist of FCUs and CRAC units, fan energy is also broadly comparable in both models. The only noticeable variation typically appears in space cooling calculations. However, when assessed against the total building energy consumption, this difference becomes marginal, resulting in an overall deviation of less than 2%.
Because data centers are highly dynamic environments with concentrated heat sources, software using detailed heat balance calculations often requires careful calibration of thermal mass, internal gain distributions, and HVAC control assumptions.
Lessons Learned for High-Density Data Center Modeling
For high-density data centers modeled in IES-VE:
- Ensure realistic thermal mass within server spaces.
- Review radiant versus convective split of IT equipment loads.
- Verify that modeled cooling capacity matches actual heat rejection.
- Consider simplified representations of liquid cooling carefully.
- Increase Furniture Mass Factor when recommended by IES support.
- Check whether cooling systems remove the full sensible load at every timestep.
In contrast, eQUEST may successfully complete simulations because its DOE-2 engine naturally smooths short-term fluctuations and is less sensitive to extreme thermal conditions.
Key Takeaways
- Thermal Stability Matter
- High Internal Loads
- Dynamic Simulation Challenges
- Thermal Mass Importance
- DOE-2 Stability Advantage
- Data Center Complexity
Conclusion
The instability encountered in IES-VE was not a modeling error but rather a consequence of the software’s more rigorous dynamic heat balance calculations. Data centers represent one of the most demanding simulation scenarios due to their extremely high internal loads and complex cooling arrangements.
While eQUEST was able to simulate the project without difficulty because of its more simplified and numerically robust methodology, IES-VE required additional thermal stabilization measures to achieve convergence.
This experience serves as a valuable reminder that different simulation engines can produce very different modeling challenges, particularly for specialized facilities such as data centers, where the balance between heat generation and heat removal is both critical and highly sensitive to modeling assumptions.
References
How to Model Chilled Beams in eQUEST – https://energy-models.com/training/how-model-chilled-beams-eques
Need Expert Support for Data Center Energy Modeling?
Whether you’re performing LEED energy modeling, simulating high-density data centers, or troubleshooting complex HVAC systems in IES-VE or eQUEST, our engineering team can help deliver accurate, reliable, and compliant simulation models.
Contact us today to discuss your project and optimize your building energy performance.
