Dear experts
I am trying to simulated earth faults, with saturatable inductors a get the following error: Unable to meet integration tolerances without reducing step size below smallest value allowed (3.73012e-18). The model give in model: earth_faults_satur_not_grounded_test.plecs with saturatable inductors.
If a do the same simulation without saturable inductors a get no error, the model is: earth_faults_satur_not_grounded_test.plecs
Could somebody give me a model that can do the simulation without getting a error and explain how the problem is resolft, so that i now how to fiks the problem if a happen to get it in a different simulation.
Kind Regards
Milan
Dear experts
I am trying to simulate earth faults using saturable inductors for my filter, but I encounter the following error:
“Unable to meet integration tolerances without reducing step size below the smallest value allowed (3.73012e‑18).”
This happens in the model earth_faults_satur_not_grounded_test.plecs, which uses saturable inductors.
When I run the same simulation without saturable inductors (model: earth_faults_not_satur_not_grounded_test.plecs), I do not get the error.
Could someone please provide a version of the model that runs without this error and explain how the issue was resolved? I would like to understand the cause so I can fix similar problems in future simulations.
earth_faults_satur_not_grounded_test.plecs (47.3 KB)
earth_faults_not_grounded_test.plecs (47.4 KB)
Kind regards
Milan
The issue is in how you are modeling your saturable inductors.
I_sat_ = 10e7;
flux_sat = 1.0000000001 * flux_L_g;
The shape of this saturation curve leads to the numerical instability. Essentially if the flux goes slightly above the specified value of flux_L_g then there is 100 mega amps of current!
You need to pick more reasonable values for the saturated inductance. The simple adjustment below proposes more reasonable values, but obviously the final saturation values depend on your inductors.
L_g_sat = 0.1*L_g; % Assume saturated inductance is 10% of unsaturated
I_L_g_sat = I_L_g * 1.2; % Pick some current greater than I_L_g
flux_L_g_sat = flux_L_g + L_g_sat*(I_L_g_sat - I_L_g); % Determine corresponding flux based on desired inductance
Dear Bryan
Thank you very much for your response.
The issue is that in my model the inductor is intentionally simplified to behave to make it a extreemly hard‑saturating element: after approximately 1.5 times the nominal current, the differential inductance becomes essentially zero. This extreme saturation characteristic is part of the conceptual model, so this behaviour needs to be unchanged ( or at least not a lott).
Is there a way to simulate such a hard saturation without running into numerical instability? One idea I had was to add more detail to the upstream network—for example, including the transformer that will be installed before the system, which has about 1.6 mΩ resistance and 19.7 µH leakage inductance. Would adding this transformer help, or would even more series impedance be required to make the simulation numerically stable?
In short, my question is:
How much additional resistance (and if needed inductance (not ideal)) is typically needed to stabilise a simulation with (almost) zero inductance and resistance, without having to change the extreme saturation characteristic itself?
Kind regards
Milan Van de Walle
This extreme saturation characteristic is part of the conceptual model, so this behaviour needs to be unchanged ( or at least not a lott).
You can still have extreme hard saturation, but less aggressive than your approach. The values you entered are clearly not physically realizable as at some point the limiting factor will be permeability of free space. You can still have a good conceptual understanding with a ratio of unsaturated to saturated inductance of 100 - 1000x nominal inductance instead of your current factor of 10^{18}.
Would adding this transformer help, or would even more series impedance be required to make the simulation numerically stable?
I recommend evaluating this yourself. The other thing that might be useful to is to define the saturation curve as a continuous function and use the variable inductor component. This creates less of an extreme discontinuity in your simulation, which is the approach used to model saturation in the PLECS machine models, where the tightness of the transition is defined with the form factor parameter f_T.
