On February 2, 2018 I posted The Internet of Things is here . . . , an article about my initial experience with IoT where I described my plan to monitor the water level in my sump and prevent a repeat of the almost catastrophic flooding of 2016.
120 days (and 151,645 measurements) later, the sump is dry and it’s time to take stock.
As the following chart shows, there weren’t really any surprises (the dotted line shows the water level in the sump and the solid line shows the cumulative number of pump activations; the unexpected mild weather in mid-February and the spring thaw are both visible in the record and the pump was triggered 39 times (first on February 21, and finally on May 1).
If that was all there was to the update, this would be a pretty boring post (even with the subtle yet dramatic colour choices in the chart). However, the data also showed how this project serendipitously saved me $10 annually, which is certainly worth some more typing.
Digression: There was also some interesting observations about the reliability of the Internet of things; there were 364 gaps in the recorded data! Fortunately most of them were very short, less than two minutes. I had originally planned to add an analysis of that issue to this post, but it is already long enough, so I may do that at another time.
As the water level in the sump rises, it raises the float of the sump pump. Once that float reaches a certain height, it flips a switch that turns on the pump and the water level (and the float) begins to fall. Once the float falls far enough, it flips the switch that turns off the pump and the water level (and the float) begins to rise again, and the cycle continues.
Like most people, I didn’t put a lot of thought into the water levels that should trigger the activation and deactivation of the pump; I thougtlessly set one at 35 centimeters (to turn on the pump) and the other at 20 centimeters (to turn off the pump). These are conservative numbers, since my sump overflows at 60 centimeters. The consequences of that choice is shown in the figure below. One can see that over a 48-hour period the pump turned on 13 times..
Two things struck me when I looked at the data more closely:
- Within two minutes of activation, the pump brings the water level down to 20 centimeters but five minutes later the water level has risen back to 32 centimeters. It seems that there is no point to bring the water level down below a certain level. I suspect that turning off the pump at 30 centimeters would work just as well.
- The pump is being triggered every hour, which implies to me that the high water level trigger is set too low. The pump is working to bring the level of an underwater stream to below its natural level. It would be much better to trigger the pump at as high a water level as one could tolerate.
As it happens, I did notice that the pump was turning on too often, and in early March, raised the trigger level to 40 centimeters, from 35. This still left me a buffer of 20 centimeters. It made a significant difference, the pump only turned on another 24 times for the rest of the season. The following figure shows the water level for April 28 and 29th.
As you can see, with a 40 centimeter trigger level the pump activated about once a day. Leaving the trigger at 35 centimeters would have been a disaster. the pump would have been turning on every ten minutes; over 100 times more often!
Based on a conservative back-of-the-envelope calculation, changing the trigger level by a mere five centimeters, eliminated well over 500 pump activations, which saved me $10 in electricity! This is an amazing ROI, considering that the monitoring equipment cost less than $10 and is likely to last for ten more years!
Serendipity is one of my favourite words! My original goal was to reduce the chances of a flooded basement. I achieved that, but I also saved beaucoup bucks.