However, you may have another solution.
When noise comes from the fans, I think that the noise is narrow band.
If true, you can generate TF from TSP responses measured with the noise.
In localization with the generated TF, you should exclude noise frequencies, which can be set in the property window of localizeMUSIC.
When you also want sound source separation, there are two ways.
One way is not to use noise frequencies in separation, but you will not get any separated signals at the frequencies.
Another way is to use TF calculated from microphone positions, although the performance drops compared with the measured TF.