[algorithm] Forcasting Energy on a free market

[lukasz p.]

Hi.
I'm using fann library to predict energy on a free market. I got 4 input and
1 output.
When I learning my network I use such data:

Input data are:
- 1st energy source and real energy at this wire [from 0 to 4,8 MWh]
- 2nd energy source and real energy at this wire [from 0 to 4,8 MWh]
- day of a week [ 0 - sunday ... 6 - saturday]
- recognize if the day we learn is holiday
Output data:
- real energy from 2 energy sources

How Could I predict my energy, which data I need? I use arithmetical average
but I'm not sure is it a good idea.

[unknown]

Hello Lukas,

I was involved in such a project for short term load forecasting in
the Swiss markets. We used an Autoregressive Neural Net to solve this
problem. Our inputs were: Yesterdays load (15 min intervals), Weekday,
Daily Temp (High and Low), and Cloud Cover (in 1/8's).

When inputing weekday, you need to classify work days from weekends
and also working days which follow a holiday. For example a Monday
followed by a Tuesday which is a holiday has a different trend than a
Monday when Tuesday is not a holiday. Similarly a Thursday is very
different when Friday is a holiday since a lot of companies bridge
such days.

Hope this helps

Good Luck

WT



-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/