Coronavirus & Controls Part 1 - measurement

Just when you are thinking "automation can be super accurate", consider what are you measuring and how you are controlling. This chart was built on actual Coronavirus data from the State of Georgia. This chart also takes the real time data in orange and applies three different type of filters to make it look "clean".

How do you see the trended data for new cases in the State of Georgia, US, and how what would your response for hospital beds be?:

1. Blue, running 7 day average with less noise - this is on the state website (https://dph.georgia.gov/covid-19-daily-status-report). Using this curve for control: "I need more hospital beds, OK I'm good, I need less beds, OK I'm good again....". Usually the values you are trying to control are off for short periods of time if you use this type of data to control.
2. Orange, real time data with lots and lots of noisy data - see this on the state website too, but typically the data is only quoted for "gripping news flashes". Using this for control: "I need a lot more hospital beds, no a lot less, no a lot more, no less, little more....". Always trying to change something to reach a goal if you use this type of data.
3. Grey, 1st order filtered data that really looks as smooth as temperature data trend. Using this for control: "Wow, I need a lot more hospital beds, a lot more, more, more, more, OK I'm good, still good, still good now a lot less, less, less, less..." This data looks real good but probably will introduce more error for longer periods of time when you make the curve look better.
4. Purple, polynomial curve fit trend line with a tightly tuned controller. Using this for control: "Hospital Beds are right on, and on, and on. BUT, I have to wait to collect the data, and then fit a curve on the data, and when I get new medicine, the curve should fit will change?" This type of data is used for setting up a lot of automation, but you have to use experience to determine the underlying conditions stability.

Actually, I like real time data PV with derivative filtering for all my friends in low control spaces, but experience really counts when setting up measurements and controls.