Zero is a terrible starting guess for a Controller
It will make the controller take a LONG time to get moving
I have seen many models with controllers that have 0 for the initial guess in the L2START or L3START specification. This is almost always a bad idea.
I’ll show you why with the IGCC2.ebs example model. I did a test with the circled controller (Controller_GasifierTemp).
The ultimate mass flow found by this Controller is 125.8 kg/s. The model, as-opened, has an L2START of 1 kg/s. This is actually an OK guess, even though it’s not close to 125 kg/s.
I tried a few starting guess. The 0 kg/s starting value is quite far from the solution, which took 528 iterations to reach.
But even a starting guess of 0.1 kg — still FAR away from the solution — ran MUCH faster — 293 iterations.
Why?
When giving 0 kg/s as a starting guess, Ebsilon assumes 1e-6 kg/s.
Controllers can change the guess no more than 15% on each iteration. So it will take a LONG time for the mass flow to become physically significant.
The first 200 or so iterations are essentially wasted, as shown in this controller plot:
What are we looking at?
This is a “Controller Plot”. On these plots,
The black line is ALWAYS the controller’s “x” variable (the thing it’s changing).
The red line is ALWAYS the “y” variable (the thing it’s trying set to a target).
The green line is ALWAYS the target variable
This is a Comp 39, where the target is constant, specified value. With a Comp 12, the green target line can be a “live” value read from a green logic line, and can change as the model iterates.
Convergence happens when the black line is flat, and the red line meets the green.
I highlighted the first 200 guesses. This shows a LOT of wasted iterations in meaningless, low kg/s values.
How to get these plots?
Select any controller, then hit Ctrl-K. Or from the main menu: Extras… Diagrams… Convergence diagrams… Controller Convergence.
Or make your life a lot easier and put a button on your toolbar to get there in 1 click.
On the next screen, select “Controller-Output”, then the Controller you’re interested in. To be honest, I never look at “Controller-Convergence.
Here’s that same plot, but after using a 0.1 kg/s starting guess. The black line gets up to meaningful values MUCH sooner, and the red line meets the target green line much sooner.
Conclusion
Don’t start with 0 kg/s. You’re just wasting time.