Best opening vs Queen's Gambit - unique situation

I really don't understand what the OP wants. It is stated that the dad will prepare against anything new that comes up. Playing something new every time is the perfect counter to that. This way all the prep by the dad is useless.

I agree if White plays the inaccurate move 7. Bxe6?, but if White plays the accurate move 7. Bb5! then suddenly Black is in a lot of trouble. 
7. Bxe6 has been played 273 times, with a respectable 58 points for White.
7. Bb5 Bc5 8. b4 Bb6 9. a4 has been played 150 times and scores an absolutely astounding 68 points!! That’s not the score of a sound defense.

if we follow the masters database and the computer then we get

Bb5+0.3Bc5+0.48b4+0.6Bb6+0.69a4+0.5a6+0.710Bxc6++0.7bxc6+0.411Bb2+0.5Nf6+0.412a5+0.3Ba7+0.413Bxd4+0.4Nxe4+0.314Bxa 14...Qxd1 15.Rxd1 Rxa7 16.Ne5 Bd5 17.Re1 f6 18.Nd2 fxe5 19.Nxe4 Bxe4 20.Rxe4

i decided to use the chess base analysis and heres its best line

I don't know. I never heard of this Perunik guy, so I'll leave him out of things. We're dealing here with 3-digit rating people. Anything that makes them study more will likely either make them improve OR overload memory processes. Anything that drags them into the weeds will likely cause more errors to crop up. In fact, if he keeps changing up his response to d4, he'll likely leave his opponent in a confusing hall of mirrors kind of opening bind.

Against three players with rapidplay FIDE grades of 2314, 2044, and 2042, I got two draws and a win. Time limit 20 minutes and 10 seconds. My FIDE rapid is 1934. (This is in the Birmingham & District Summer League. And this has been the totality of my recent usage. No losses.)

Against the 2314 I won the exchange for a pawn but couldn't convert it. I scraped a draw with the 2044, but won very nicely against the 2042. The first two players have usually beaten me when I've face them before.)

For what it's worth, I used the rapid b5 move before it has recently become trendy.

The QGA is good enough for my level. I know that.

no the QGA is perfectly sound at all levels, ethan just does not know enough about it and thinks 3.e4 refutes it

It isn’t, I do, and no, I think it essentially refutes itself and 3. e4 is just the most fun way to do that. I don’t even play that line, I play the Showalter and Gunsberg and score ridiculously well.

The QGA for me goes in that class of openings in which they’re not optimal at top level and yet too unprincipled at lower levels and so they tend to be mostly forgotten about, with some few very dedicated followers.

3.e5 is completely equal for both sides (sure whites winrate is higher than blacks but thats how all openings are)

you could also find other defenses that are more dynamic if thats the issue

but you cannot deny the 3.e5 lines are very drawish at higher levels because the lines are very forcing and equal compared to when white players something like 4.Nc3 or even 3.e3

and as white you want to win, not draw

indeed. Sounds like your dad just has your number! And you, just scrambling around looking for something to play against him helps him. You're basically allowing him to adapt to you instead you adjusting your play to him. Play just ONE line against him for awhile, but instead of switching to another line~invest some time to analyzing some of his tendencies. Play him not the board. Obviously, you will have to look at yourself and shore up some of your own deficiencies.

I highly recommend studying Tartakower's variation. An alternative would be Lasker's. Practice n play other people. Preferably much higher rated players, even if you lose just gain experience. Of course, still play your father but strive to improve YOUR own play. Do as your dad does, watch YouTube, gain insights. Level up on your tactics, combinational play. Learn rudimentary endgame themes, basic positional strategies.

Also, it is sometimes a fact, that one does not want to beat or win against their dad. Just a thing. That could be your ultimate problem. As it is, absorb what is useful. Discard what is useless (to you) and add what is uniquely your own.

Try the Keres Defence with 1.d4 e6 2.c4 Bb4+ transposes into a Nimzo or Bogo Indian maybe even a Ragozin QGD.

Find the one that Miodrag doesn't have a video on and play it

"Countering the Queen's Gambit" GM Prusikin and the chessable course "Keep it simple for black" both recommend very similar reps based around the Queen's gambit declined. If 1.d4 d5 2c4 e6 3Nc3 Nf6 4Nf3 a6 so if for example Bf4 black can play d5xc4 and keep pawn with b5. If the classic 4Bg5 first recommends 4...h6 5Bh4 Be7 6e3 0.0 7Nf3 b6 Tartakower variation. (If your Dad refutes this publish the game because it is more than world champions and elite GM's have managed to do), and second 4Bg5 pxp because can transpose to Panov attack Caro kann he recommends 1e4 c6 in same book.

Against the catalan 1d4 d5 2c4 e6 3Nf3 Nf6 4g3 play 4...Be7 5Bg2 0.0 60.0 pxp the idea is to counter Bg2 with Bb7 or Bd7-c6, and then look for c5 break.

That is true, Upon second examination the benko isn't very solid and gives few chances to win...

Queen's Gambit Accepted is sound, Albin's Countergambit is not.
Here is a recent correspondence World Championship game (average 5 days / move, engines allowed)

Yes, these Father/Son games must be being playing had a fiendishly high standard that openings like the Queen's Gambit accepted are being refuted, lol.

In reality it's an opening that was considered sound in pre software days being played at elite level, including world championships. The main effect of software is to make just about any opening playable. In more main line areas, software and GM games have pretty much neutered the queens gambit exchange as a winning attempt for white. Formerly, it was a very popular way for white to try and bore black to death with. Matthew Sadler has a book where shows AlphaZero even prefers black's chances in exchange Queen's Gambit declined positions.