Normalizing was great for 8 bit samplers in the 80ies.
What software think is a loud peak may not be what you think is a loud peak. Normalizing vocal tracks is probably a futile process. Avoid normalizing unless you have some long track of specific value that’s just has no power to work in a mix or to drive compressors to even react enough or something like that.
You will most likely have a peek somewhere in your vocal track and that will offset the rest. If you think you have to use this track for one reason or the other, cut the track up in phrases or sections that are equal in sound level and put them on different tracks with no FX whatsoever, then set the levels to what sounds equal and bounce the tracks to a new one. Then a compressor can shave the last unevenness off. It could work as a Plan B but Plan A would be …