Power_analyze-increase accuracy (state-space)

I'm using power_analyse to extract state space data from a *.mdl file (a SymPowerSystems-model). Then I transform the matrices to Clarke domain, aka alpha-beta-zero coordinates (with a manually written routine). Since the matrix entries are inaccurate, the matrix after transform contains many small numbers at the places where zeros should stay. The situation is complicated, because the model contains some uH-stray inductances and some nF-capacitors, which represent fast dynamics. Because of them some "small numbers" are not errors, esp. in case when the circiut is unsymmetrical. If I could increase the precision beyond the standard double, then the errors would lay beyond e-14 (m.b. e-40 or even e-140), which makes them clearly detectable.

I'd be also glad to receive some advices, how this problem can be solved by other means. M.b. some recursive algorithms to make the errors disappear, while the true data stays how it is...

There is no option to derive everything manually and then use VPA-ariphmetic to directly calculate the ABCD-matrices from parameters, since the order of the system is high.

Answer 1

[Joel Van Sickel]

If you need greater than double precision tolerances you will need to solver this problem with custom data types and math libraries that support enhanced precision.

[Ilya]

Uff, I don't remember what was exactly my question 8 years ago, but thanks for replying.

May be B = vpa(A) and working with B is fine indeed.