Knowledgebase: Error and Warnings
How significant are warnings about max iterations exceeded
Posted by - NA -, Last modified by Linda Lawrie on 20 April 2010 04:18 PM
Q: I get the "Max Iteration" warning often, in varying quantities. I'd like to understand better what they mean. |
1) If these are a concern, at what frequency? (e.g., "whenever it occurs more than 100? 500? 1000? times in a full year run.")
2) Roughly how much does it affect accuracy of the simulation? (A lot or a little? proportional to the number of occurrences?)
3) Any tips about how to avoid it?
A: This is a good question, but it is difficult to answer. It is something to be concerned about, but in many cases there does not seem to be a way to completely eliminate them and they aren't necessarily a cause for alarm.
1) The total count is a difficult measure to use because it varies with number of zones, number and type of air systems, and length of run period. A 1,000 might not be a problem for a big model with an annual run, but it could be way too many for a single zone design day run. The errors are more common with VAV than CV. The frequency is key though. I look at the timing of the errors. If they happen every time step during some period, then it usually means there is something wrong with HVAC. If they happen only sometimes, and those times are when things are changing quickly (like recovery from setback), then I don't worry much.
2) It depends if the system is succeeding at controlling the zone conditions. If the systems are controlling well, and the errors are intermittent, then the results are probably not affected significantly. If the systems are not controlling zone conditions, then the errors are probably very significant. Check the comfort conditions and zone air temperatures to see.
3) When the errors are significant, they usually indicate something is wrong with HVAC input that EnergyPlus isn't able to trap in some other way. Possibilities include all sorts of things that can go wrong such as: systems connected wrong (node connections usually), sized wrong (mixing hard and auto sizes), controlled wrong (check operation of set point managers by reporting node set point values).