New curve generation methodology

I am changing the way the curve is generated starting in Jan 2004 to better
replicate seasonal fundamentals. There are convincing arguments as to why
the summer/winter spreads should tighten over time. However, in the previous
methodology they blew out. For instance summer/winter in Cal 3 was .232
while Cal 10 was .256.
I have added a seasonality dampening function that both contracts the
summer/winter spread and applies a premium to the electric load demand months
of July and August over time.

The formula for the curve remains the same except for a premium lookup for
the month as well as for the year. These premiums are as follows:

Jan -.008
Feb -.004
Mar -.001
Apr .002
May .003
Jun .004
Jul .004
Aug .004
Sep .003
Oct .002
Nov -.003
Dec -.006

These premiums start in Jan 2004
On Wednesday Jan 2003 settled 2.959, the 3/4 spread was marked at .0375, the
4/5 spread was marked at .0475.
In the old methodology
Jan 2003 = 2.959
Jan 2004 = 2.959 + .0375 = 2.9965
Jan 2005 = 2.9965 + .0475 = 3.044

In the new methodology
Jan 2003 = 2.959
Jan 2004 = 2.959 + .0375 - .008 =2.9885
Jan 2005 = 2.9885 + .0475 -.008 = 3.028

The only change in the formula is from:
Month x = Month (x- 1 year) + lookup on year on year table
to
Month x = Month (x- 1 year) + lookup on year on year table + lookup on month
premium table

The seasonality premiums will change over time and I will let you know when I
change them