How to Troubleshoot Complex Building Models and HVAC Systems

One of the most common questions we receive on our Support Desk goes something like this: “my annual summary simulation results are not what I was expecting, what’s the problem?” The techniques listed below are routinely used by the DesignBuilder Support Desk engineers and other experienced energy modellers to answer this sort of question.

1. Check hourly or sub-hourly results for some key parts of the building and systems. For example, by checking the temperatures downstream of a cooling coil you might find that it is undersized and causing inadequate cooling. Or the results might reveal that a schedule is set up incorrectly and heating is not operating during the expected periods. It is usually sufficient to run these diagnostic simulations for short periods such as typical winter, summer and “shoulder season” weeks.
2. When running Detailed HVAC, a quick, easy way to generate diagnostic system temperature, flow rate and humidity data is to select the appropriate options under the “Miscellaneous” header on the “Outputs” tab of the Simulation Calculation options dialog. These results can be viewed after the simulation using the DesignBuilder ResultsViewer.
3. Sometimes simulation models are so large and complex that it can be difficult and time-consuming to track problems down. In this case, simplifying your model can help you to focus on the problem areas. You might do this by creating a version with most of the building blocks removed to allow you to focus on a few key zones. Sometimes this is a straightforward process, but some large models have a lot of interconnected zones and HVAC components and in this case, you may need to apply caution when trimming your model. For example, if there is an issue with AHU coil sizing, the sizing process will use data from all of the zones served by the AHU, and by removing most of the zones it serves you may not see the problem in the cut-down version of the model. Similar care needs to be applied with Calculated natural ventilation models in which the flow of air through the building is affected by the airflow paths between all of the zones, and removing large parts of the model can significantly affect the airflow patterns in the remaining zones.
4. It should go without saying that if you do decide to simplify your model to help diagnose issues, you must take a backup of the full model first!
5. Alternatively, instead of actually removing blocks and zones from the model, it can be more convenient to use the “Include zone in thermal calculations” option on the Activity tab to exclude zones from simulations. If you do use this simplification method you should still take a backup of the full model first, especially when using Detailed HVAC, because excluding zones from the simulation in this way also drops the zones from the HVAC zone group and re-including them later won't automatically restore the zone-HVAC zone linkage.
6. Also, if your diagnosis needs you to access any of the less common outputs not directly supported by DesignBuilder, you can select from the thousands of different results provided by EnergyPlus by adding your own outputs using IDF text. The program help explains how to do this and includes a simple tutorial.
7. Sometimes in HVAC models, it can help to revert any hard set input data to "autosize" on all HVAC components to see if that solves the problem. If it does, then you can gradually change back to hard-set values to find the problem component.
8. Finally, it is always worth looking at the EnergyPlus error file (eplusout.err) to check for possible convergence, sizing or other issues, even if you are already happy with your results.

Having checked key building and system detailed results and the error file and having made any changes necessary to ensure that the model is behaving as expected, it is time to run the annual simulation. You should now have every reason to be confident in the summary results!

The take-home point is that detailed model checking should be routinely carried out to ensure that you understand how your model is behaving before even looking at summary results. The more complex the systems involved, the greater the scope for errors and the more important it is to check your model carefully before using the results.