Let’s talk today about big data in relay protection realm.
For those not in the know microprocessor device of relay protection Sepam, hold on to configuration in the file settings with extensions according their type.
Of course, vendor provide own software - SFT2841, in order to work with that file settings.
It's classic that software has such pattern flaws:
1) possible to work just with single file
2) Unload protection settings possible only to pdf and with unfit forms
3) It is impossible to analyze numerously files of settings
As well as many other things.
Long story short it is usual software designed for electrical engineer’s manual labor, indeed get interesting challenges :)
For example, we have 300 Sepam devices.
This devices S81, S82, B83, M81 and T82 types.
Supposedly too we have the task to make analysis ALL protection settings, which it has on that quantity devices.
It is not very good that settings for the same protections everywhere different.
We have to take broad picture - what exactly settings same type protection set on different devices.
The best thing to do that if will be all settings to place in Excel table, where handy filters give possibilities in several clicks to sort necessary parameters.
I didn’t notice intentionally variation with SQL, because of Excel has enough over the top possibilities for such data quantity.
Let’s estimate data.
For Sepam devices their 5 types it is possible 25 unic types of protections.
It means for example type Sepam T82 has 19 types of protections, for Sepam M81 - 21 available type of protections.
For every type of Sepam device it has as similar protections as device specific protections. We take only unic 25 protections.
By analyzing (read below about) we determined 14 used protections from 25 possible.
Every protection setting it could to take as 1 value (current, time, voltage etc.) as up to 18 different types of values, for example type of protection curve.
Not a few.
But if we try to calculate all possible values for all parameters?
For 14 types of protection, if we take all steps, we have to take all steps, looks like 8801 possible values.
8801 we multiply on 300 devices and as result 2 640 300 possible values, which we have to not only estimate but to do something with that.
If it's taken average novella weight in 70 000 words, but 1 value is 1 word, that we have to read approximately 37 novels.
It's a lot to take in for one reader, isn't?
If it would be interesting that 37 novels, but there just dry figures...
In this case it could be better compare old settings with new one, but this additional same quantity come upon our comparation.
But this is no big deal if we incorporate programming skills and automated facilities on Python.
However, let’s stop afraid of fears and turn to implementation such automated approach.
Setting files of Sepam devices it has one interesting specific thing, it keeps own data in text mode, where protection settings looks like usual words, and possible values - numbers.
SFT2841 encode these parameters by the means, which I could find apply myself on R&D work.
Own to this Sepam particularities we can extract from device all necessary parameters.
Certainly, having all Sepam setting files.
The script iterates all setting files and extract parameters for 14 possible types of protections to Excel table.
Having data about protection settings in table, using by filters, even visually very easy determine any aberration of normality - mistakes.
Far less that it possible to use many other automated facilities for analysis.
As well as with help automated facilities it could easy compare two patterns of configurations - old and new one, and mark those settings, which were changed.