Fitting a general extreme value distribution on transformed data

Question

I have a data set $X_1, ..., X_n$ and want to fit a general value distribution to this data set using R.
However, when trying to do so I get an overflow. So, my idea was to instead fit a general extreme value distribution to a transformation, let's say $f(X_1), ..., f(X_n)$, where f should be a bijective function (I tried for example log(), sqrt(), min-max-normalization).
This works and gives me the parameters $\mu_f, \sigma_f$ and $\xi_f$, but I don't know now how to get the location, scale and shape parameter of the original data from this. For the location $f^{-1}(\mu_f)$ seems to work as I get for all $f$ a pretty similar value.
Has anyone an idea how I could get those parameters?
Thank you!