Bo, thanks. I will look more into applying a FFT. I see we have some articles on it - http://code.jsoftware.com/wiki/Guides/Fourier_Transform - I just need to study how to apply it. Thanks for the tip
On Fri, Aug 26, 2016 at 2:37 AM, 'Bo Jacoby' via Programming < [email protected]> wrote: > Hi! > When you have got positive numbers only, then first take the logarithm, > because whatever calculation you make you never want negative values. Next > multiply by 100 because the accuracy of the measurements is around 1 > percent. Next round to integer values because the fractional part is > uninteresting. > ([:<.[:100&*^.) > > Then make a discrete fourier transform to get a power spectrum. Low > frequency peaks tell the base load. High frequency peaks tell how often the > air conditioner turns on. > Good luck! > Bo. > > > > Den 7:20 fredag den 26. august 2016 skrev dhkelly <[email protected]>: > > > > I would suggest that you could get away with samples taken at longer > intervals. You won't lose useful information as none of the major > appliances including the air conditioner should draw a startup peak > power for over 6 seconds and what you really need is the increase in > load over a longer time. In addition, the air conditioner is a motor > load which has a lower power factor than something dominated by a heater > element. > > A change between two readings at the interval you are using or an > interval of the order of 60 seconds (or longer) will essentially give > the same information -something changed in the interval- whatever it > is. Determining whatever is a bit more iffy. If you get large jumps at > 2AM- blame the air conditioner. If at 8AM it could be stove and/or water > heater (breakfast and shower) indicating that you got up then. > The trouble with a whole house monitor, is that a measure of changes in > power alone, may not really give you all the data to distinguish > sources. Another is that a high sampling rate may give variations that > are really not of interest. I think that a step would be to correlate > the data with your direct observations of what went on or off - when( > i.e at the time you turned on a major load such as a stove or dryer- > both of which will cycle their power levels "on/off" in the same way > that an air-conditioner does. > (e.g. an oven will turn on until the desired temperature is reached, > then turn off , repeating this cycle to maintain a a given temperature > (+/- a bit) just as an air conditioner or dryer does). > > Don Kelly > > > > Does the meter also measure power factor or "vars"? > > > > On 8/25/2016 8:05 PM, Joe Bogner wrote: > > I posted an article that looks at some data I captured using a whole-home > > energy monitor. I also posted the data. I put up a challenge for anyone > > who wants to take a stab at it. > > > > http://code.jsoftware.com/wiki/User:Joe_Bogner/EnergyUsageAnalysis > > > > How often does the air conditioner or clothes dryer turn on? > > How long does it normally run, when did it run the longest? > > Is it possible to determine when I go to bed or wake up? > > > > I'd be interested in any approaches to answering any of these questions > or > > other analysis on the data > > > > For example, the air conditioner draws a significant amount of energy. I > > just turned it on and my usage went from 1460-6000 watts. > > > > Looking at the change in prior reading may be a good start > > > > plot (}. reading - (_1 |.!.0 reading)) > > > > I have an electric dryer and electric oven. > > > > I hope the data and article is interesting/useful > > ---------------------------------------------------------------------- > > For information about J forums see http://www.jsoftware.com/forums.htm > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > > > > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
