|
|
re: photovoltaic sizing
4 jul 2000
muz (with his upside-down brain address) wrote:
>|...say the risk is 10% that one off-the-shelf panel is 5% low.
>|it's less likely that two panels will both be 5% low, and so on...
>they just love wallys like you at las vegas.
and i'd like to go there sometime, with their inexpensive tours--$300
for 4 days from phila, hotel and airfare included. i just got back from
atlantic city. the 2-hour bus ride was $20.50, round-trip. i walked into
the claridge casino, collected their $20 cash gift, signed up for their
free comp card, which included a free drink, watched some people playing
roulette, walked out, and spent the day on the free beach.
roulette is interesting. the ball can land in 38 places on the wheel.
a bet on one number pays 35:1, so a $10 bet returns 10x35/38 = $9.21
on average, much better than slot machines. betting on either of two
numbers pays 17:1, for an average return of 10x17x2/38 = $8.95. why
would anyone do that? betting on 3 pays 11:1, ie $8.68; 4 at 8:1 pays
$8.42, and returns progress downwards to $5 average for 18 numbers.
i used to work with an engineer who modified a (1980?) cb radio so it
fit in the small of the back under a suit jacket. one of the men on
his team wore it into a casino with a wire to a switch in his pocket.
he tapped the switch with his finger whenever the roulette ball passed
a certain point on the wheel, which sent a pulse to a computer outside
the casino which measured the times between pulses and predicted and
displayed the most likely spot for the ball to end up, based on the
increasing pulse intervals. the outdoor man spoke the number into the
radio, and the person inside bet that number at the last possible
minute before bets on that spin were prohibited...
my engineer friend ended up with a $30k van for his efforts, and the
team leader bought about 12 houses with his share of the money before
he retired. nowadays, casinos have radio scanners and metal detectors
and better ways of finding out who is winning at what. there's also a
local law. if a casino catches you with this kind of equipment, they
can throw you in jail for 90 days. (is that constitutional?) then again,
computers are so much smaller these days, and lots of people use palms
to keep track of appointments...
>every flip of a coin has the same chance of coming up heads. ie 50/50
yes.
>if a panel has a 10% chance of being 5% low so do the next one on the shelf
yes.
>in these circumstances the previous choice has no effect on the current
>choice.
yes, but suppose we rephrase the question to "if each panel has a 10%
chance of being 5% low, what is the likelihood that the combined output
of n panels will be 5% low?" with bernouilli coin flips, that's 0.1^n,
a 1% chance for 2 panels, 0.1% for 3, and so on. or we can ask "what's
the 10% lower confidence limit for the sum of the outputs of n panels
with independent identically-distributed outputs?" that comes from the
mean times the number of panels, with a standard deviation decreasing
as the square root of the number.
of course, given the choice, it's better to cheat and measure them all
and pick the best.
nick
|
|
