Just recently, using Frank Bergmann's.SED-ML tools for the conversion I've added SED-ML support into PySCeS. Its pretty simple to use as well, anything generated with mod.doSimPlot() can be exported.
mod = pysces.model('chain.psc','d:\\projects\\modeldir')
mod.exportSimAsSedML() does all the work and its arguments specify that output should be produced as raw SBML/SED-ML files, a SEDML archive (the same files zipped together) and a COMBINE archive (*.omex) a recently proposed archive format that includes the same files with all sorts of metadata.
While this is the simplest use-case, to do this I've wrapped most of the SED-ML 1.1 spec into a Python class that can be used to generate arbitrary SED-ML descriptions. This actually generates Frank's SED-ML script which then either uses libSEDML, installed locally, or the web services provided here to generate the SED-ML.
sed = pysces.SED.SED(sed_id, sed_output_path)
In : sed.
In : sed.
This will all be available in the PySCeS 0.9 release (soon :-))