If I am to process time series of energy data (point estimate every half-hour interval), and fit it into the following formula for dimension reduction:

Haben's paper

I was advised to use power, not energy. Yet power is the rate at which the energy is formed. I still don't see how other than to use the values as it is. If I convert it into hourly value, it would remain the same. Must I divide by 2 to indicate that they are half-hour interval values? Then isnt it the same not applying? Since you would be normalizing with the entire year's data.

Aside from the energy/power concepts, (or maybe relevant) I actually was a bit confused how to process it to sample the time period from 6.30 am to 10.30 am for instance. If it is a point estimate, how can it be in unit of kWh?

Should perceive that the reading at 6.30 am is the electricity consumed from 6 - 6.30 am or now that I think if it is 5.30 - 6.30 am...

May I ask for clarification from anyone with solid experience in this matter? Thanks

  • \$\begingroup\$ is this a school assignment? \$\endgroup\$
    – jsotola
    Mar 16, 2018 at 2:15
  • \$\begingroup\$ no this is for my research..and for some reason i am applying clustering/coalition techniques on smart grid network without strong basics in understanding of power/energy. \$\endgroup\$
    – dia
    Mar 16, 2018 at 3:06
  • \$\begingroup\$ "Haben's paper" - is this it? centaur.reading.ac.uk/47589/1/… \$\endgroup\$ Mar 16, 2018 at 3:21
  • \$\begingroup\$ @BruceAbbott They reference just one article on the use of PCA; a powerful tool and even more so when intelligently applied (for example, to reduce to important dimensions.) The paper's summary, so far as I can tell, discounts without comment anything that may have arrived from that reference. Now I'm curious to read their first referenced paper. \$\endgroup\$
    – jonk
    Mar 16, 2018 at 3:48
  • \$\begingroup\$ Uhm.. Can anyone answer my question? I know abt PCA but im sure the author used other dimension reduction method for its advantage.. I contacted him and was advised to use Power not Energy \$\endgroup\$
    – dia
    Mar 16, 2018 at 4:07

1 Answer 1


I gather the purpose of statistical analysis here is to define the rate of peak Power demand during daily and seasonal diurnal loads, so kWh would not apply as the duration of Peak power for different times of the day or week or weekend.

These peak durations of interest will be much less than a hour yet the grid must be maintained dynamically ( impedance of net sources) to supply this peak power on demand and stay within specified voltage tolerances.

If the meter reports kWh in 30 minute intervals then you use the average energy per reading kWh x2 = P per 30 minute reading.

  • \$\begingroup\$ not sure where you read 'peak power demand'. From what I gather, the meter readings are point estimate per each 30 mins, data explained here without much details. or am I missing background that these are in fact peak power demands? \$\endgroup\$
    – dia
    Mar 16, 2018 at 5:51
  • \$\begingroup\$ At any rate I get that the correct conversion would be multiplication of 2 but still.. why would I need to do that to normalize if I apply that to all the time series data and then divide? \$\endgroup\$
    – dia
    Mar 16, 2018 at 6:07
  • \$\begingroup\$ This was a study done by Marhematicians, not the EE perspective that I gave. \$\endgroup\$ Mar 16, 2018 at 7:27
  • \$\begingroup\$ yes but the purpose is to correctly understand the nature of the data and to expand on previous researches..if there is fallacy in interpretation then I shall improve on it. So in your opinion, the meter data represent peak power during the 30-min interval? \$\endgroup\$
    – dia
    Mar 16, 2018 at 8:25
  • \$\begingroup\$ RMS power is standard but peak rms power loading even for a second is important data to a power study \$\endgroup\$ Mar 16, 2018 at 12:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.