The idea of a 25 Big Bet buy-in in a limit game - and a buy-in of 100 Big Blinds in a no-limit game - is not actually based on the size of the blinds or the bets. The blinds/bets are serving as a rough proxy of the size of the game - that is, how big the pots run. You really want to scale your buy-in to the average pot size - especially the flop pots, I think.
In no-limit, 100 Big Blinds is the rough starting point for "big stack" poker theory to apply, but if the game is loose and nothing ever gets to the flop without seeing a bet that's ten or more Big Blinds, then you're not really playing that game, you're playing a bigger game, and the 100 Big Blinds is probably not enough. If pots are typically 40 or 50 BB before the flop comes (and we've seen such games, right?) then 100 BB is definitely not a big enough stack for big-stack theory to really apply. You're playing a bigger game, just with reduced blinds.
Which is all a long-winded way of saying that it's not the blinds, it's the pot size - your buy-in is based on the pot size.
But if I had to propose a buy-in based on the limits in a limit game with three bet sizes, I'd use this logic:
In a 15/30 limit game, 25 big bets ($750) is a typical "full" buy-in.
In a 22.50/45 limit game, double the size, $1125 would be the typical 25 big bets.
A 15/30/45 limit game will be bigger than a 15/30 limit game... but not as big as a 22.50/45 game. So somewhere between the two.
Opening bets are still 15, but the closing bets will have a little extra oomph - so I think the game size will be closer to 15/30 than to 22.50/45. If you split the difference between the two, you're allowing a safety margin by buying in for $937.50... which is 31.25 times the $30, or 20.8 times the 45.
That makes intuitive sense to me:
For 15/30, you need 25 x 30.
For 15/30/45, you need more than 25 x30... so 31 x 30 makes a kind of sense.
Alternately, for 15/30/45, you don't need as much as 25 x 45... so 20 x 45 makes a kind of sense.
Either way, that's $900+, and $1000 is convenient.