[TRNSYS-users] Powell's method solver

[Benne, Kyle]

Hello TRNSYS group,

Thank you Werner for helping with my previous question.  I am developing a thermosyphon water heater model such as the type45, however it is desired to have a model that will accommodate reverse thermosyphoning such as what often happens at night.  I am using TESS Type604 pipe that can handle reverse flow and TESS Type534, a stratified storage tank.  The flow rate is calculated by a thermosyphon promoter that computes the flow rate based on the temperature of the elements in the system.  This type previously did not handle reverse flow, therefore I modified it to output a flag if the flow was reversed.  A small piece of code interprets this flag and directs the flow to the correct connections of the system.  For example If the flow is in the assumed positive direction, the inlet of the pipe designated inlet A is input with the magnitude of the flow calculated by the thermosyphon promoter.  The opposite end of the pipe is given a flow rate of 0.  This strategy successfully determines forward and reverse flow and directs the inputs accordingly, but fails to converge when the flow rate is switching between forward and reverse direction.  I have read in the TRNSYS documentation that the successive substitution method often has difficulty with discontinuities.  I then decided that Powell's solution technique might be more capable for this problem, however I am having difficulty getting it to work.  When the Powell solution technique is chosen the program errors saying the input file needed by the Type534 storage tank cannot be read.  This is odd considering it reads the file fine using the substitution method.  I believe the Powell method is more particular about what initial values I choose for the inputs, and may require other things, but I am not certain about what those might be.  If there is someone who managed to read this entire message and can offer some pointers I would appreciate the advice.

	-Kyle