Rotating compact disc   Electric power forecasting - prediction methods, thesis in PDF-format, download page      

Location

HOME  >  ELECTRIC POWER ENGINEERING  >  POWER FORECASTING

 
HOME
ORACLE 7/8/9 TIPS & TRICKS
NATURE & OUTDOOR LIFE
WORLD INFO
ELECTRIC POWER ENGINEERING
POWER FORECASTING
GALLERY PAGES
ABOUT ME
ABOUT THIS SITE
CONTACT INFORMATION
INQUIRIES

Roger Felix - Lund - Sweden
Launch pop-up site map

Latest update:

2002-02-20

Rate this page

Download PDF-document containing my control theory examine thesis on Electric Power Forecasting

Download PDF-file By clicking here
The document is compatible with Acrobat reader version 3.0.

This PDF  document describes methods of electric power load prediction using a seasonal ARMAX method and periodic (cyclic) models, giving an prediction error of 0.7% for an one-hour prediction horizon, and a prediction error of about 1.5% for the 24-hour  prediction horizon. The graph below shows the prediction error vs. prediction horizon for stable periods (power load influenced only by weather and time of week).

 

In swedish, please Läs den här sidan på svenska.
Read this page in frensh. Read this page in german. Read this page in italian. Read this page in spanish. Read this page in portuguese.
[HOME] [INQUIRIES] [CONTACT INFORMATION]
This site is HTML 4.01 and CSS2 compliant.

 

 

MatLab m-files:

lspred.m

pem1pred.m

pem2pred.m

 

More on the power load process

The lower prediction error is accomplished by using multiple models representing the power load process on different times.  The process dynamics varies during the day, as does the process static gain from temperature [centigrade Celsius] to power load [MW]:


The shape of the average daily power load curve has a seasonal dependence. This is shown by the following graph, having one curve plotted for each month:

The hourly correlation from temperature to power load is varying during the 24 hours of the day. On the average there is a delay of 2 - 3 hours from temperature to power load. From the right figure (showing correlation as  a function of delay and time of day)  it is evident that the correlation has the largest magnitude in the early morning hours.


 

 

In swedish, please Läs den här sidan på svenska.
Read this page in frensh. Read this page in german. Read this page in italian. Read this page in spanish. Read this page in portuguese.
[HOME] [INQUIRIES] [CONTACT INFORMATION]
This site is HTML 4.01 and CSS2 compliant.