I recommend that you first consider using a different Type of cooling coil that does internal flow control calculations (such as Type508c).
PID controllers are useful in real machines, where things may change every millisecond. In simulations with reasonable timesteps of 1 minute or longer, PI controllers, or even just proportional controllers should be enough.
There is also an iterative controller (Type 22), which will slow the model down, but is capable of controlling as you would expect a real PID to behave.
I have also had some luck in TRNSYS using simple feed-forward type controls, where you predict the output using a simple equation then adjust that output with a proportional controller. For example:
control signal = predicted signal f(input 1, input 2) + adjustment f(actual result - desired result)
Ben Heymer. PE