Model validation

  • 7 January 2020
  • 5 replies

Userlevel 2
Badge +4

Dear All,

In many publications related to power system expansion planning, model validation is done by applying the developed model to a small system such as Garver 6 Bus and describing the result without detailed explanation to justify the model is valid or not.

Is there any specific methods to validate the optimization model?

Thank you very much.



Best answer by rahmat 16 January 2020, 02:47

View original

5 replies

Userlevel 2
Badge +4

@mohansx , many thanks...

Userlevel 5
Badge +5

@rahmat , not needing to validate a model because it is convergent will not hold true all the time. Your model is convergent yes, but that does not always mean that the model accurately represents your physical process. That is why we need the validation step. 


From a quick search, seems like the Garver 6 bus system is a commonly used system to validate power system models and the output of your model must match the numbers presented in the literature. If your model is a modified version for which there is no existing literature, then manual calculations on a small dataset is a good validation step. 

Userlevel 2
Badge +4

Hi @mohansx  and @Andrew ,

Even with a small system like Garver 6 bus, it needs very hard work to do a manual calculation.

Some suggested that to do a validation it can be done by comparing the result with real-world data. However, the model I developed is the modification from the existing model with the addition of several parameters and variables. Therefore, this comparison can not be applied.

Some also suggested there is no need to validate the model since the model is convergent.

So for the next step, I’ll develop a very small system, say 2 or 3 bus system, that can be easily calculated by manual (or excel) calculation as suggested by Mohansx.


Hi @rahmat 

Was Mohan’s answer sufficient in answering your question?


Userlevel 5
Badge +5

@rahmat , using a small data set with known results (like the Garver 6 Bus system in your case) or results can be calculated/verified manually is a common way of validating MIP models. The rationale is that the results of your mip model must match the known results. If there is a mismatch and it is not expected, usually there is something wrong with the formulation. 


Didn't find what you were looking for? Try searching on our documentation pages:

AIMMS Developer & PRO | AIMMS How-To | AIMMS SC Navigator