studying half-hourly time series


I have to study a time serie composed of half-hourly data concerning electric load for the period of 6 years : so in my file data i have for every year all the days and for every day i have the value for the electric load consumption for every half-hour.
my study consists on finding the caractéristics of this time serie, I didn't know how to begin and what approach would I follow
then i want to ask if I should stydy all the data (for the 6years ), or it is could be sufficient to study only one years data.

I m waiting for your helps, thanks
If you can graph all of the data (x-axis is time, y-axis is load), you will probably find useful characteristics just by looking at it. I expect that you would see relatively consistent patterns for each day (peak load during certain hours each day), and possibly seasonal differences that repeat each year (higher loads in Summer for air-conditioning). Definitely use all six years of data. After you identify repeating time periods of interest by looking at the graph, you might re-organize the data to compare periods to each other with other techniques (t-test to compare the average of all Friday loads to all Monday loads).
thanks WhiteCog , i appreciate your ideas

but should i identify what type the process is : AR, MA, ARMA?

what graphs are necessary apart the line plot?

then I have another problem, my data like this : for every year I have a table in which i have a column containing the days and for every day i have the load for every half-hour , so i must recognise my data to have only too column : in the first i have the timestep (date + hour) and in the second i have the load, so how can I do this? is it easy with R software?

Unfortunately, I am not familiar with those process acronyms, and I have never used R. Hopefully an expert will comment. Your current tables may make it easy to perform t-tests comparing times of day (is load higher at 4pm than 3pm). For other comparisons, and for the master graph, I think you should reorganize the data into 2 columns (time and load). The easiest way I know would be to use a macro in Excel to repeatedly copy and "paste special" (transpose) each day's row to a single column, but there may be an easier way. Depending on your software, it may also help to add columns that label each row by a time grouping (a column with 7 values for day of the week, a column with 12 for month, others for groups such as season or "work day"). I am pretty sure that would help you make useful comparisons among time groups.

Though I may not be much more help, it might help other members respond if you say what you hope to accomplish with your study. Are you forecasting future load for any particular times? Are you identifying times to reduce or increase load? How rigorous do your results have to be?

Depending on your goals and standards, you might be fine with the tables you have, just reporting the average load for any given time of day. There are many characteristics of this data that can be found, but it may not be worth the work to get them all if you only need a few.

thank you Whitecog, very useful are your remarks and advices,

until now, I m not familar with making macros in excel! but I must try

about my goal, I have to make a detailed study for the load consumption caracteristics (this what the responsable of the training asked me) and I have to forecast futur load using neural network, I have no idea about the model to use for this forecast, so I want to identify the caracteristics of my serie in order to choose the suitable model.