After having a look at the variations you guys gave it really seems that micawber is right, and 17...Qd7 can be countered with the idea Bb4 (either at move 18, or after Qb3) thus preventing that the queen can go to d6 after Nc5. And when the queen captures on c7 the position seems favorable for white. Still black might have some improvements in this line or maybe after 17...Qe8.

But I've thought for a while about the general plan in this position. The previous analysis was always directed at challanging white in the center with the extra piece.
What do you think of the plan to attack on the kingside? After all this is a KID, and maybe white has devoted to many sources to his attack on the queenside/center to defend his own king?!

So again: 
1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 e5 7. O-O Nc6 8. d5 Ne7 9. Ne1 Nd7 10. Nd3 f5 11. Bd2 Nf6 12. f3 f4 13. c5 g5 14. Rc1 Ng6 15. Nb5 a6. 16. cxd6 axb5 17. dxc7 Qe8 18. Qb3 

and now black plays the topical 18... g4. White follows with Nc5, and as I already said I had this position as white recently, and won after 19...gxf3? 20. Bxb5 +-

But during the game I was more concerned about another move, namely 19...Nh4!? . This frees the g6-square for the queen, and prepares a sharp attack. 

I won't spit out masses of computer analysis, but going through some lines Fritz often changed his evaluation from winning for white to winning for black.
The thougthless approach 20. d6+ Kh8 21. Bxb5 Qg6 for example seems to be quite dangerous for white.

So what do you say? Is this plan a reasonable antidote, or must black fight back on the queenside?

Micawber, I try to find variations played in grandmaster games, and don't you think that if your 18.Bb4 was that brilliant, one or two strong players would have played that before ? Any computer will find something like 18.Bb4 Rf7 19.Qb3 and then 19...Qe8 (or even the risky 19...b6), which is assessed "=+" (or "=") almost in every variation. 
No matter to discuss such computer moves, it's better to consider real improvements by strong players. Otherwise everybody's gonna post endlessly variations of his own chess programm...

   Finally, I found an old chess Informant giving an interesting analysis by Soloviov (Biryiukov-Soloviov, St Petersburg 1999 ; 77/535). Black must accept the sacrifice and here is the line after 15.Nb5 a6 ! :
 
16. cd axb5 17.dxc7 Qd7 (! ; a splendid move, allowing Ne8) 18.Qb3 Ne8 19.Nc5!? Qd6 20.Ne6 (20.Fb4?! Qb6 21.Kh1 Rf7 22.Rfd1 Bf8 =+) Bxe6! (and not 20...Qb6+ ??, since after 21.Kh1 Bxe6 22.dxe6 Cxc7 23.e7+ Rf7, white has the decisive 24.Bc4!!, which was actually played in the game Biryukov-Soloviov) 21.dxe6 Nxc7 22.Bb4 Qb6+ 23.Bc5 Qa5 24.Bxf8 (trying to avoid the draw after 24.Bb4 Qb6+, but...) Kxf8 25.Rfd1 Qb6+ 26.Kh1 Nxe6 27.Qb4+ Kg8 and white doesn't have anything at all. 

The position is even, but I guess black is doing ok. So I think white must find something else in the classical system. The variations after 10.Be3 seem sharper.

So I've recently stumbled over this interesting line:
 
1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 e5 7. O-O Nc6 8. d5 Ne7 9. Ne1 Nd7 10. Nd3 f5 11. Bd2 Nf6 12. f3 f4 13. c5 g5 14. Rc1 Ng6

---------------------------
In my view black has already some problems here. An interesting line to try is
11.....fxe 12.Nxe4,Nf5 with the folowing strategic objective in mind:
Black is going to stabelize the queenside with b6/e5, aims to post a knight on d4, and can
exchange white's good knight after Nf6.

An example:
11.... fxe 12.Nxe4,Nf5 13.Bg4,b6 14.Rc1, a5 Kocic-Nevednichy, Rumania 1997 (0-1)
---------------------------
1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 e5 7. O-O Nc6 8. d5 Ne7 9. Ne1 Nd7 10. Nd3 f5 11. Bd2 Nf6 12. f3 f4 13. c5 g5 14. Rc1 Ng6

15.Nb5  ,a6
16.cxd6 
16......, axb5 This is quite dangerous but it is doubtfull if 16...cxd 17.Nb5!,a6 18.Nc7 is any better
                   Also note that this position can also be reached after 15.cxd,cxd 16.Nb5,a6!
                   While 15.cxd,Qxd6 16.Nb5,Qb6 17.Nf2!,Rf7 18.Na3! also looks quite good for white:
                   18....Qxb2? 19.Nb5 and the queen is trapped.
17.dxc7,
17....... Qe8  Other queenmoves don't solve black's problems either
                  17....Qd6 18.Bb4, Qd7 19.Qb3,Rf7 20.Nc5!, Qxc7 21.Ne6,Qb6 22.Bc5,Qa5 23.Nxg5
                  looks pretty uncomfortable for black (Novikov-Lillo, 2002)

"Gallagher stops here, and says that white's compensation is insufficient. I'm not so sure if he's right on this one, and I'd welcome some advice on how to play with black."
 
I hesitate to contradict an expert such as Gallagher, but to me it seems that white has sufficient compensation in his pawns, his active pieces and his attack on the queenside. I couldn't find many practical examples though.

18.Qb3         This seems white's optimal choice.
18......,  b6    More cautious seems 18....Qf7 but even here black is in trouble:
                   19.Bb4,Re8 20.Nc5 (now white threatens Bxb5), b6? 21.Ne6!,
                   ---Bxe6? 22.dxe and black can't recapture because white threatens Bc4
                   ---Rxe6  22.dxe,Bxe6 23.Qc3 with Rfd1 to follow
19.Rc6
19......, Rf7    Again 19...Qf7 looks not entirely convincing
                   19...Qf7 20.Bb4,Re8 21.Nf2 and blacks queenside faces anihilation. 
                   White has more than sufficient compensation.
20.Nf2             Uncovering an attack on b5
20....     Bf8
21.Bxb5,Bd7

22.Rfc1!         Now c8Q  is threatend. For instance:
                      22.....g4, 23.c8Q,Bxc8 24.Rxf6!, Qe8 25.Rxf7,Kxf7 26.Nxg4 +-

22......., Bc5     22...Bxc6 is not an option:
                      22....Bxc6 23.Bxc6,Qc8 24.Bxa8 +- (24...Qxa8?? 25.c8Q)

23.Bb4,           23.Re6 looks interesting as well. It enables white to protect c7 with d6
23......,  Ne7    23....Bxb4 24.Qxb4,Bxc6 25.Rxc6 was maybe best. 
                     But again white should have sufficient compensation as he is going to play d6

Now white initiated a forced sequence:
24.Rxf6!,Rxf6
25.Bxc5,bxc5
26.d6+ , Kg7
27.Bxd7, Qxd7

So far we have been following the game Biriukov-Solovjev, St.Petersburg 2001.

And here white missed a chance to decide the game with
28.Qb7!! 

So I think that black faces an uphill task defending this position.
I'd rather be white anyway.