To keep it simple, ignoring the effects of critical hits, the test is:
If (Avg_Damage * (chancehit_1st_attack+chancehit_2nd_attack+...) < Avg_Damage+Power_Attack * (chancehit_1st_powerattack+chancehit_2nd_powerattack+...)) then Power_Attack(Foe)
In your case, it is clear to me that you don't want to power attack a foe with 20 or 23 AC. I suspect you'll do a little good to power attack a foe with 15 AC if you don't go overboard.
Let's suppose you drop 3-10 into power attack.
Your average damage sans power attack vs. an AC 15 foe is:
18.5*(.95+.95+.85)=50.875
Your 3 point power attack gives you:
21.5*(.95+.95+.7)=55.9
Your 5 point power attack gives you:
23.5*(.95+.85+.6)=56.4
Your 6 point power attack gives you:
24.5*(.95+.8+.55)=56.35
Now that the damage dropped, we don't really need to go further than this, but just to show you what happens:
Your 7 point power attack gives you:
25.5*(.95+.75+.5)=56.1
Your 10 point power attack gives you:
28.5*(.8+.6+.35)=49.875
I've never tried to work out a formula for max damage to calculate how much you want to power attack in a given situation, but in general I find <= 5 works well unless you know you can't miss. I'd love to see a calculation, but I suspect it would be too involved for routine use anyway.
I also suspect that once the effects of criticals are accounted for, power attacking is made even weaker because your latter attacks are more likely to critical if you don't power attack. This is especially true because your axe can already provoke a massive damage save on most rolls without power attacking. When in doubt, don't power attack.
PS: I was composing this message when Petrosian and Cheiromancer were posting and so didn't read their posts until afterwards. I suspect that the calculations they are using are either wrong or else being misused, since clearly you can increase average damage (as shown above) against targets with AC > 13 and with > +2 power attack.