I have a set of temperature data from all the wind turbines within a wind farm. I am interested in detecting outlier values in this set and generating eventframes based on this detection.
More specifically, at a given time I want to calculate the median and the MAD (median absolute deviation) for the park (alternatively, mean and standard deviation is an acceptable solution).
Next, I want to run an analysis for each turbine that over a given period of time, e.g. one week, calculates the percentage of time in which the temperature was outside the limits given by MEDIAN-3*MAD and MEDIAN+3*MAD. Finally, I want to generate an event frame when this percentage of time is higher than a given threshold.
I'm thinking that asset-based analytics is a way to go and I have some of the steps figured out:
1) Calculate mean/median using rollup analysis: OK
2) Calculate MAD/STD: HOW??
As far as I can see, this cannot be done using rollup analysis.
An expression analysis together with the PStDev function could work, but is not desirable as the set contains more than 100 temperature values (plus has to be changed manually whenever a new element is added).
3) Set up an event frame generation analysis template using the TimeGT/TimeLT: OK
So, my problem here is the 2nd step. Besides, I'm in principle not interested in saving the intermediate mean/median and MAD/STD values, although it is acceptable.
I have considered using some workaround with SQC, but it seems that step 2 is also going to be a problem here.
Any help with this solution or suggestions for other solutions - perhaps ACE? - are most welcome.