Re: controller cycling

[Jeff Thornton <thornton@xxxxxxxxxxxx>]

At 09:47 AM 11/12/2003 +0000, you wrote:

> trying to avoid cycling in the solar collector loop, I wrote a TRNSYS 
> component implementing the collector one node capacitande model, which is 
> solved by finite diferences.
>
> Response of the collector dynamical model is damped, but cycling persists, 
> even highly decreasing the time step. This behaviour, apparently, has not 
> physical meaning and I think is due to numerical effects.
>
> Can somebody tell me more about this or related questions?
>
> Is there any paper describing (more detailed than in the TRNSYS user 
> manual) the numerical algorithms behind TRNSYS?
You have come across a common but often overlooked issue when simulating 
fluid loops in TRNSYS.  Imagine you have the case of an auxiliary heater 
adding heat to a flow stream at a constant rate where the fluid outlet from 
the device is connected to the fluid inlet to the device.  In real life, 
this system will eventually run away and the inlet temperature will soar 
until some high-limit condition is reached or the device fails.  In TRNSYS, 
the fail condition is reached on the very first timestep as the inlet 
temperature keeps rising and therefore the outlet temperature keeps rising 
in a vicious circle.  Your situation is exactly the same - except you have 
some capacitance in your collector to dampen things slightly.  Even though 
you are starting with the same initial temperature each time in the 
collector model (well you should be anyway) , the inlet temperature to the 
collector keeps rising and therefore the outlet temperature keeps rising 
and you're stuck back in the vicious loop again.

Luckily there is a simple solution.  Simply insert a pipe in your system 
that is able to hold at least one timestep worth of fluid.  That will 
completely solve the problem.  In fact, we have modified the Type 31 model 
and added a mode that calculates the required length given the fluid flow 
rate and pipe dimensions (the loss coefficient is then modified 
accordingly).  That model will be part of the new TESS libraries available 
with version 16 of TRNSYS.  Otherwise any device that holds fluid for at 
least a timestep can be used (storage tank etc..)

We have caught many user-simulations without the required pipe length and 
the collector system will iterate until the inlet temperature dictates that 
the losses from the collector are equal to the collector gains.  The system 
solves and the user never knows that he has encountered a run-away solution....

The TRNSYS manual does a nice job of explaining the solution method; 
describing the successive substitution method as it is implemented in TRNSYS.

Jeff



*********************************************************************************
Thermal Energy System Specialists (TESS)
Jeff Thornton                                   2916 Marketplace Drive
Principal                                       Suite 104
Phone:  (608) 274-2577                  Madison WI 53719
Fax:  (608) 278-1475                            USA
E-mail:   thornton@tess-inc.com
Web Page:       www.tess-inc.com

"Providing software solutions for today's energy engineering projects"
*********************************************************************************