Home Power Monitoring Hack

dvogt writes "You think your power bill is bad? I built a power monitoring system to monitor every circuit in my house with three second resolution for over a year. And while I had to rewire all my electrical to do it, I can now reconcile my electricity bill down to the penny... Of course when my wife figured out most of the bill was because of my computer gear I had to build her a dome, so reader beware!" From the article: "About a year ago I developed a web based power monitoring application for data centers. The application was designed to monitor thousands of individual branch circuits using current transducers at the breaker panels. Among other things, the data logging requirements were to provide one year of min/max/mean measurement data with one minute resolution per circuit. Since I had all the hardware for testing, I figured what better way to test things than to install it in my own home."

Comments

Another way to do it: read the meter

[seanadams.com]

If you want to monitor the whole house (as opposed to individual circuits) you can do it for less than $20 in parts plus a Linux machine. I made a system to do this a couple years ago - unfortunately I never hooked it up again after I moved, but it worked just fine.

If you're lucky enough to have the kind of electric meter with a blinking LED on it, you could do this much more simply. Also if I had to do this again I would ditch the op-amp circuit and feed the signal from the photo-resistor straight into the sound card and then do the filtering in software (if the photo-resistor is exposed to sunlight it can be a little tricky to tune using this circuit - software could be smarter).

1 minute resolution is not enough

[zippthorne]

In fact he must sample at greater than 120hz* to get meaningful results. He has neglected the possibility that voltage and current can and will be out of phase for each of the loads in his house. Without determining the phase difference, there is no way to accurately deterimne the average power over any interval.

There are quite a few meters that measure RMS voltage and RMS current, (though most of the cheap ones actually measure peak values and multiply by .707), there are fewer still that accurately resolve power factor

This is a common mistake to make for first year EE students and "over-unity" power converter proponants.

As I understand it, the Kill-A-Watt, http://www.professionalequipment.com/xq/ASP/Produc tID.3375/id.5/subID.57/qx/default.htm makes a pretty good approximation. In fact, it even does the integration for you. You could pepper every outlet with these things or just move them around as needed.

*I know you need 2f according to nyquist to resolve the frequency, but I'm not sure what you need to gather the phase information**

** There are other ways to obtain the phase information involving bridge circuits and such, It does not appear that the boards in question provide that information.

Re:1 minute resolution is not enough

[dan42]

I believe the 1 minute resolution was for the reported data and not the actual analog sampling rate. The 1 minute data may (should) be the RMS of the all the samples collected during that minute.
His device seems to be a prototype version of the Veris H663. The released version of the device apparently samples at 1280Hz and reports data every second.
The "phase information" you mention is good to have, but only if the power is at just one frequency. The fact is that computers (as with most devices running of a switching power supply) draw current as a short pulse during the peaks of the input voltage waveform (with no phase angle between voltage and current at the fundamental frequency). Often the 60Hz component (or whatever the fundamental frequency) makes up only half the RMS current. And since these harmonics are only dominant in the current and not the voltage waveform, the real power consumed by a computer will typically be ~50% x RMScurrent x RMSvoltage (even though the phase angle is zero).