So strightforward approach to code a macro is probably really useless (without extra constraints), because complexity of problem is 2^n (some 10^180 cases for 600 values
).
And I assume, that you need exactly zero. Not close enough to zero, like +/- 1. 
As you posted a workbook, here is proposition of randomized locating of such groups based on my recent post: http://www.excelforum.com/excel-gene...ml#post3609912
But while it is reasonably efficient with small sets (as in sheet1 of example). in large dataset - it gives up 
Funny side observation - rarely it is able to spot group two "groups" together - from maths point of view - absolutely correct, because if:
8881.92 + - 6003.60 + - 2878.32 = zero
and - 196,546.73 + 211,632.56 + - 10,229.00 + - 4856.83 = zero
then
8881.92 + - 6003.60 + - 2878.32 + - 196,546.73 + 211,632.56 + - 10,229.00 + - 4856.83 = zero too.
such result is visible in first sheet.
Color coding of data is not cleared, so after runnimg a macro color interiors remain. (change in code clearcontents to delete if you wish).
Bookmarks