My company performs remote monitoring for a several hundred APC UPS' covering a variety of models attached to various systems we administer or maintain, primarily (but not solely) at customer - or otherwise remote - sites utilizing a 3rd party monitoring platform. In short, what we'd like to be able to do is set a threshold on our monitoring software to alert when a battery reaches a temperature that will serve to provide advance notice of a battery whose capacity is beginning to diminish irrevocably - but isn't there yet - so we can be prepared in advance to replace them, rather than only after an actual alarm has occurred and causing the customer to have to deal with it.I am led to understand that a battery's average stable temperature will increase once its total capacity begins to decrease below a certain threshold - even if average ambient temps do not change - and would like to know if there is any data available on average/observed temperatures vs. capacity over time with the additional data point of mean total failures along those time-lines? I understand, of course, that ambient temperature will have an effect on this, but whether the data includes ambient temps, or is just a controlled environment, the information would be useful to us to establish a baseline understanding of the temperatures that might affect - or be predictive of - future battery performance or failure so we can set monitor thresholds to that end.
Thank you very much,
Bump please for reply.
This sounds more like a battery tech question than a UPS question, so I guess that it can be better answered by people working with batteries every day - there may be some here that have enough experience from past issues and personally I have had my own list of failures from batteries in string-applications and as you indicated, a battery (cell) that is getting low in capacity will be prone to heat up, especially if it gets discharged to polarity reversal, which then accelerates the capacity loss. I have not seen this (yet) in my UPS, but have had to deal with it in my EV.
Note that the heating up does not occur (much) in shallow charge/discharge unless the internal resistance of the cell is significantly increased, then you may see additional heating, but my experience is that it especially presents itself during a longer discharge test, since that brings the weaker cells into deep discharge with the resulting additional heat generated.
Note that another failure mode: bad connection or cable contact also results in heat, so it makes sense to measure close to the terminals of the batteries. Hope this helps.
Temperature is not a reliable method to do what you want to do.
Yes, *some* battery failure modes cause increased dissipation, but not all. In my experience maybe 5-10% of batteries fail by developing an internal fault that cooks the battery from the inside out. The remainder simply gradually lose capacity and develop higher internal resistance until the UPS screams "enough!".
To be honest, probably the best way of monitoring what you want to monitor would be to keep tabs on the battery constant in a SmartUPS. As the UPS does it's cycles & tests, it keeps track of the battery capacity under load (as this is what it uses to provide runtime estimates). This drops in concert with the estimated battery capacity. This does not detect pending catastrophic failure however. It goes without saying there is no way to easily expose this value without delving into the old UPSLink protocol, and newer UPS hide this internally so you can't get to it anyway.
If you were to run a prediction, you'd use temperature, load & projected runtime. The combination of load & projected runtime can be reversed with a bit of maths and an estimate at peukerts constant to give you a vague measure of actual battery capacity, and battery float life is significantly reduced at elevated temperatures (see any and all SLA battery data sheets for the graphs).
I did an exercise recently where I took a stack of UPS, got the load & runtime values and reversed that back into battery capacity, then replaced the batteries and actually measured the units that came out. In most cases the calculations reflected the actual battery capacity within about 10%, and all were better than 20%. SNMP and a bit of python would do that for you.
Choose a location: