Almost all of the comments are valuable.
I think that a consensus is building (probably better: has been built) that the standard base system for science use is the numpy/scipy/matplotlib stack. But there are packages that don't build on that stack. I'm afraid you'll have to do some digging to see which packages will work for you.
There are many many many packages that build on the numpy/scipy/matplotlib stack. There are also many packages for more specialized tasks, such as dealing with large data sets, or inhomogeneous data sets. And packages for specific scientific fields, astronomy for example.
So you see it's hard to give a straightforward answer.
But one very important package that is extremely useful for adding visualization to a simulation is VPython ("3D Programming for Ordinary Mortals"). I would strongly encourage you to take a serious look at it.
There are also several "batteries included" meta-packages that greatly simplify the installation of python for scientists. One is Anaconda, another is PythonXY, another is Enthought Canopy. And, again, there are others.
I think you will have to take a look at what's out there, choose things that looks sensible to you, and start working with them. My apologies to the many many packages that I haven't mentioned.