Representative Load Profile!
Representative Load Profile!
(OP)
I'm working on a project to develop electrical load profiles for non-metered consumers to help with the load management and peak shaving decision making.
So far, I have developed daily profiles based on available data and grouped the profiles by weekdays and weekends for each month.
The problem I have is that I need to get a representative profile each for the weekdays and weekends of each month. Averaging the data may not give a true picture of the reality.
Are there any ideas on what statistical (or other) manipulations I may employ to get a meaningful representation. I'm talking about 20 profiles and 8 profiles for weekdays and weekends per month.
Hope someone can help.
So far, I have developed daily profiles based on available data and grouped the profiles by weekdays and weekends for each month.
The problem I have is that I need to get a representative profile each for the weekdays and weekends of each month. Averaging the data may not give a true picture of the reality.
Are there any ideas on what statistical (or other) manipulations I may employ to get a meaningful representation. I'm talking about 20 profiles and 8 profiles for weekdays and weekends per month.
Hope someone can help.






RE: Representative Load Profile!
If the measurements were taken on different days, there may be a false reduction in variability because of differences in temperature on the different days. If so, you might try to weather normalize the data.
One approach may be to find the std deviation of the time of peak, the std deviation of the pu loads at different times, and the std deviation of peak loads. Then use these in a Monte Carlo simulation of multiple loads on a feeder.
Just some thoughts.
RE: Representative Load Profile!
(1) Weekdays (Monday-Friday)
(2) Weekends (Saturdays, Sundays, holidays)
You may then normalize the data for weather using the following regression equation:
L=a+bx+cy+dz
where,
L = hourly load
x = hourly temperature
y = hourly wind velocity
z = hourly cloud cover
a = y-intercept, determined by the regression program
b, c, z = coefficients determined by the regression program
Obviously, you need historical hourly data for L, x, y, z. Once you have determined a, b, c, d you may then extract the weather effect from the load data.
RE: Representative Load Profile!
Correction to my post above:
b, c, d = coefficients determined by the regression program
RE: Representative Load Profile!
http://www.ornl.gov/ORNL/BTC/Restructuring/ORNLTM200227...
http://ets.powerpool.ab.ca/asc/PFAM/Government/BIG%20In...
http://certs.lbl.gov/pdf/LoadReliability.pdf
etc. for more info
Also, visit
http://www.amazon.com/exec/obidos/search-handle-form/10...
for literature covering the electricity market
RE: Representative Load Profile!
But forgive my ignorance, can someone give me the basics of what the monte carlo simulation is all about?
Thanks again.
RE: Representative Load Profile!
In your case, you might generate load profiles for a specific number of customers to simulate the load on a feeder. You could generate many such load profiles to simulate a large population of sample loads on the feeder.
You can do a Monte Carlo simulation with a spreadsheet program like Excel either using macros or by using circular calculations and a set number of calculation repetitions.
RE: Representative Load Profile!
http://www.riskglossary.com/articles/monte_carlo_method...
etc. for more info