September 2019 Update

New! We're moving to the new forum, eventually it will replace this blog. Please join us there.

View the Most Recent Feature Sheet by clicking HERE!

Saturday, June 15, 2019

Watering Time Predictions - More of a guideline....

Turns out that getting very accurate predictions is beyond my ability, mathwise... well, at least I don't know how to calculate the time because the soil moisture depletion is not linear as I had thought...

If you look at any of the graphs, you will see this (this is the monthly graph)


From the peaks (plant was just watered) to the lowest point (plant needs to be watered) it's a straight line, or a linear decline.  I based my calculations on this decline being steady, and linear...

You can see a sharp drop at the beginning on some... That means for the first bit the drop off is very fast, which means the calculation of average decline will be skewed...


Now, lets look at a weekly graph of the same plant...



This is the most recent watering cycle, the one our calculations would be based on. You can see the initial decline is faster, then it steadies out... It looks like a fairly significant drop... (I haven't looked at the log files yet)

There are also some flat spots, which means the value didn't change for a period of time...  This causes the prediction to waver. Let's look at an example...

Just making up some numbers here...

  • At 4am the sensor reading is 270 and the autowater trigger is 250, so 20 more to go... After looking at the average decline, it determines it will need watering in 3 hours, at 7am
  • At 4:15 the sensor reading is the same, there is still 20 more to go, and it will still take 1 hour, and need watering at 7:15
  • At 4:30 the sensor reading is the same, there is still 20 more to go, and it will still take 1 hour, and need watering at 7:30
  • At 4:45 the sensor reading is the same, there is still 20 more to go, and it will still take 1 hour, and need watering at 7:45


You can see how the predictions can waver during times when the reading doesn't change... Keep in mind, the soil is still drying out, this is just a limitation of the sensors sensitivity.


Now let's look at the Daily graph for the same plant...


Wow, it's sure flattening out at the end...  that's really causing the predicted time to slide... Here's the actual predictions from the log...



In just over 13 hours of logging, the prediction has slid 7 hours...  

The plant was actually flagged for watering at 3:37am

So, since trying to figure out the math to accommodate the non linear moisture decline is beyond me, I'll just leave it as is, and consider if more of a guideline... :)  It will make a useful tool just to give me a heads up of what's going to need watering over the next couple days... This will be handy if I want to go away for a couple days, I can top up all the plants, and then see that I've got x number of days (roughly) and if AutoWater is set up, that will take care of the first couple that need watering...

All in all, I think it's still a very cool feature.

No comments:

Post a Comment

Any comments deemed off topic or offensive will be removed