District cooling loop with pump heat addition and solving for rejected heat

[smunns]

Hello,

I am working on a district cooling loop application. Below is a much simpler toy model to illustrate the problem I am trying to solve.

Code:
dc_secondary_test.txt

Diagram:

Results:
output.txt

I would like to add two capabilities:

1. Show the effect of a temperature rise across the circulation pump based on a stated hydraulic efficiency.
2. Instead of explicitly setting the hx_rej heat flow to -2000.0 W to oppose the heating supply of 1000 + 1000 W, solve for the heat flow required to balance the heat being added to the loop. This will not be known explicitly in advance, because of the circulation pump temperature rise. It doesn't look like I can do this directly with a heat_consumer in its current form, because my temperature setpoint is from the pump exit, not from the heat exchanger exit.

What would be the best way to accomplish these two capabilities, which seem fairly modest?

Also, notice that in the results, the Circ Pump Pressure t_outlet_k is 313.15 instead of 303.15 like I expect. This happens as a result of the first Newton-Raphson iteration, but I can't track it easily beyond this point to see why it is happening. Is this a bug?

Thanks,
Scott

[dlohmeier]

Hi Scott,

thanks for this really interesting case study. I must admit that cooling networks have not been within our scope yet, so my suggestions should all be a little experimental and have not been tested.
First of all, I wonder how you manage the flow within your system. I assume that in your real system, there are many heat sources in parallel distributed over a long distance between a flow and a return (I am not sure if the terminology is the same between heating and cooling networks, please correct me if I'm wrong). So in order not to end up with 99% of the flow just flushing through the heat exchanger closest to the pump, you will need a high resistance, which we model with a flow controller that must know the mass flow in advance. You might have a look at the heat_consumer component, which is really convenient for modelling the behavior of a combined throttle valve with heat exchanger.
If you want to configure a temperature rise across the circulation pump, does that mean that the outlet temperature will rise effectively? The circulation pump currently doesn't have that capability, I just think that it could be reached by using another heat exchanger that feeds in the pump's heat power. This can be reached by using a controller. The same applies to the heat rejection, you could use a controller to determine the required heat flow and use it as heat exchanger input. However, if you only want to know how much heat needs to be extracted to get back to the pump outlet temperature, you can probably just use the pump and calculate it from inlet / outlet temperatures. This all might seem counter-intuitive to use a heat exchanger for feeding in pump power loss while leaving out the heat rejection and just let the circulation pump extract the required heat, but the component models currently implementen in pandapipes do not display the exact capabilities you are describing (as far as I know). But of course, you could always use controllers to determine the exact power of rejection and pump and just add it to the system in a loop around the pipeflow.
Regarding the outlet temperature of the circulation pump, this is indeed a bug and has been reported in #669 . Unfortunately, a clean fix is not quickly done, so you cannot rely on this value at the moment.
I hope that helps you get along with your problem. Don't hesitate to ask other questions or give feedback.

Best regards,
Daniel

[smunns]

Thank you for the detailed response! I guessed the situation might be as you described, where I would need to use controllers to get the functionality I need.

My "district cooling loop" is actually the secondary loop of a data center, so pipe diameters and lengths are quite a bit smaller than what you are usually accustomed to. Long pipe lengths would be on the order of 100 meters. Knowing the amount of flow maldistribution between different heat exchanger branches is important, as lower flow puts some servers/CPUs/GPUs at higher temperatures, and therefore higher risk of failure. We use balancing valves in each branch to minimize the flow maldistribution. We can also use a "reverse return" piping arrangement to reduce the flow maldistribution. It is not so costly since the distances are small.

I will probably utilize controllers to achieve the required results for now. Now that I am finally getting more comfortable with the details of the calculation mechanisms, I might be able to make modifications to support some of these features directly in pandapipes, if you would be willing to consider them.