0

What are my options to solve in scipy the ODE system:

Obviously if A is invertibile then i can solve equivalently:

using standard methods from scipy.integrate. But still remain the following questions: what about efficiency and what if A is singular and cannot be inverted?

mrkwjc
  • 1,080
  • 8
  • 17
  • Is `fun` linear? – percusse May 04 '18 at 07:54
  • In general not, but in most problems i faced it is really of the form K*y+f – mrkwjc May 04 '18 at 07:57
  • In the general case you basically have a ODE set subject to algebraic constraints. In your specific form the problem is called Descriptor systems in control literature. And eigenvalues of A and K come into play, see for example http://www3.math.tu-berlin.de/Vorlesungen/WS14/ControlTheory2/scripts/ControlDescriptorSystems.pdf – percusse May 04 '18 at 08:04
  • Otherwise also known as (systems of) differential-algebraic equations, DAE. `scipy` does not have DAE solvers, one would have to check with other packages like DSTools and SunDials. – Lutz Lehmann May 04 '18 at 08:46
  • Thanks for tips! Currently i ended up with assimulo https://trac.jmodelica.org/assimulo/browser/trunk which provides pythonic interface to sundials IDA solver. I'm going to try also petsc4py. There exist also scikits.odes, but its interface to sundials seems to be insufficient for my (large scale) needs. – mrkwjc May 05 '18 at 19:53

0 Answers0