Heat pump model behavior with UA against ttd in evaporator and condenser

[JanKochanski]

Hello!
So far I was using a fixed ttd (5K) in my model, now I changed this to a fixed UA and found a strange behavior of the COP for the overall model. I do not fully understand why this happens - maybe someone has some ideas about it?

The image below shows temperature lift over Q with COP coloring. It looks quite logic to me - higher temperature leads to lower COP.

When I now change to UA, I get mixed results. Below you find one example where COP is still in valid ranges, even though we see a quite clear "sweet spot" for the COP

When I vary the seawater inlet temperature (heat source for the heat pump) (same UAs), I get results like the one below. I see highly negative values (not valid) and values way above the viable range (generally about 0.8-10). I also get warnings from TESPy:

The simulation converged but the calculated result nan watt / kelvin for the fixed input parameter UA is not equal to the originally specified value: 3000000.0. Usually, this can happen, when a method internally manipulates the associated equation during iteration in order to allow progress in situations, when the equation is otherwise not well defined for the currentvalues of the variables, e.g. in case a negative root would need to be evaluated. Often, this can happen during the first iterations and then will resolve itself as convergence progresses. In this case it did not, meaning convergence was not actually achieved.

Knowing that UA does vary slightly with the mass flows - I am not sure what physical mechanisms could explain the COP "sweet spot" that appears when using fixed UA but with a fixed TTD it is rather monotonic? Some ideas what causes that nan watt / kelvin problem? I tried varying the UA manually with the same results as shown above.

I would be super happy to get some inputs on this topic!
Best
Jan

[fwitte]

Hi Jan,

there is a couple things happening here. First, please do the following: Rerun your first simulation (the one with the fixed ttd_l) and output the UA that results for that instead of the COP. Can you post that plot here?

To the error: With the fixed UA you set to the model the solver cannot find a solution, which has a non-negative pinch internally. With negative pinch the UA equation does not work and it would make the solver crash. Therefore, TESPy locally adjusts the outcome of the equation. The simulation converges but always with the adjustment in place, which means, that the result is invalid. Something to help can be sorting the simulations you do in a specific order, i.e., in a way that the change between one simulation and the next one is as small as possible. We can have a short meeting if you are interested and see how we can make that work in your model.

Best

[fwitte]

See the code for the sensitivity_analysis for sorting simulations: https://github.com/oemof/tespy/blob/dev/src/tespy/models/template.py

[JanKochanski]

Hi @fwitte,

these are the plots (scatter plots to make it a bit clearer to see) - however with different axis to make it specific for the heat exchangers.

Thanks for the link to the sensitivity_analysis! If you are open for a short meeting that would be highly appreciated!
Best, Jan

[fwitte]

You can already see in the UA plots how vastly different the numbers are based on the other boundary conditions. You can try to make use of the structure in the sensitivity_analysis method:

• Get the current state of the model with respect to the values you want to iterate your simulations over
• Use the method inside to calculate the path of minimal changes through the grid
• Execute the simulations in a loop

This will much improve numerical stability of the simulations. To keep track, if a simulation was successful, use Network.status variable also. If status is != 0, then there was an issue in the simulation. Also, maybe use the current dev branch implementation, I improved convergence behavior for UA related equations: #953

[fwitte]

For a meeting @JanKochanski: What about Wednesday?

[JanKochanski]

Hi @fwitte ,
Wednesday works great, I am available whenever it fits you best. Thanks in advance!
Best
Jan

[fwitte]

What about 14:00?

[JanKochanski]

14:00 works well, thanks and see you on Wednesday!