Capturing Trends with Alarms

[LAWx]

I am trying to develop a rule that will raise an alarm when the temperature starts climbing from the daily minimum, but I don't want to catch every minor up and down. Ideally I would just take the temperature readings from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 minutes in the past and use linear regression to calculate the slope of the line created by these values. I would know when the trend had turned from decreasing or flat temperatures to increasing when the slope was greater than zero. Unfortunately I don't have access to these values in Meteobridge. The best I have are six selector values (e.g., val2, val5, val10 ..., or min2, min5, min10 ...), but using all six of those selectors means detection would lag real life by an hour. Using too few of those values means minor fluctuations will create false positives.

Any ideas on a way to reliably identify a trend and smooth out the noise using Meteobridge rules?

[admin]

When this helps you, I can add the requested selectors. Would val1 - val60 be fine?

[LAWx]

If that doesn't help me, I don't know what can. My only concern after that would be tasking Meteobridge with an alarm condition as complicated as:

Code: Select all

(((10*(([th0temp-val1=F.1:999]*1)+([th0temp-val2=F.1:999]*2)+([th0temp-val3=F.1:999]*3)+([th0temp-val4=F.1:999]*4)+([th0temp-val5=F.1:999]*5)+(([th0temp-val6=F.1:999]*6)+(([th0temp-val7=F.1:999]*7)+([th0temp-val8=F.1:999]*8)+(([th0temp-val9=F.1:999]*9)+([th0temp-val10=F.1:999]*10)))-(55*([th0temp-val1=F.1:999]+[th0temp-val2=F.1:999]+[th0temp-val3=F.1:999]+[th0temp-val4=F.1:999]+[th0temp-val5=F.1:999]+[th0temp-val6=F.1:999]+[th0temp-val7=F.1:999]+[th0temp-val8=F.1:999]+[th0temp-val9=F.1:999]+[th0temp-val10=F.1:999])))/825)<0

I guess I'll give it a go if you commit those changes. val1-val60 should be plenty.

[admin]

When you send your Meteobridge MAC to "info(at)meteobridge.com" I can add you to beta testing this new function.