Temperature-Dependent Parameters

[bobduif]

Have you considered introducing temperature dependency into some of the model parameters?

Based on the data available to me (though I’m not sure it’s the most accurate), I’ve observed that the measured top-oil temperature consistently lags behind the computed top-oil temperature. Interestingly, this lag doesn’t appear to be constant—it tends to vary with temperature. Specifically, the lag seems more pronounced at lower temperatures and decreases as temperatures rise.

One possible explanation for this behavior is that at higher temperatures, increased heat exchange leads to faster oil circulation. Additionally, as the oil heats up, its viscosity decreases, which reduces flow resistance and may further enhance circulation speed.

This suggests that introducing a temperature dependency for the oil time constant $\tau_o$ could improve the model’s accuracy. A similar approach might apply to the winding time constant $\tau_w$, although I currently don’t have data to support that. Alternatively, it might be possible to introduce temperature dependencies in $k_{11}$ and $k_{22}$ instead of (or in addition to) modifying $\tau_o$ and $\tau_w$ to achieve similar improvements.

Finally, I was also considering a temperature-dependent formulation for the hot-spot factor $H$ (or possibly $g_r$ or $k_{21}$), to account for the temperature-related increase in conductor resistance, modeled as:
$R = R_{ref} [1+(T-T_{ref})]$
While I don’t have data to confirm whether this would significantly improve model accuracy, conceptually it seems like a plausible direction worth exploring.