I just came across your power monitoring project. Great work! The online graphing is ace!

I’ve been working on the same thing using a current transformer and an arduino. I think I may have got around this issue of not measuring voltage. Im getting good results by ‘synthesising’ the voltage waveform infering it from the current waveform.

I’ve put all the details up here:
http://openenergymonitor.org/emon/node/28

I haven’t done any internet connectivity stuff yet but will do soon, your work will definitely help me out!

http://jarv.org/pwrmon.shtml#accuracy

@roman: thanks for the insight into energy detective, it sounds you guys have a great product. think my next step for this project is going to see if I can incorporate voltage measurements and continue to use the kill-a-watt as a calibration tool.

I presume it would be realtively easy to add voltage measurement, with a step-down transformer plugged into any wall socket (transformer for isolation from mains – a spare AC wallwart should be fine; try to pick a socket on a circuit with low load; if you have more than one phase then just presume voltage imbalance between phases is insignificant).

Use the arduino to find the voltage phase by detecting -ve to +ve zero crossing (maybe need an RC filter to remove noise?).

You could also measure the voltage if you use the ADC to measure the step-down voltage reading (maybe need voltage divider depending on transformer output voltage). Calibrate the stepdown voltage readings against a mains voltage reading taken using a multimeter.

Reguarding the power factor :) there is magic that happens in backround but it’s nothing that magical just a matter of collecting the data and doing the propper calculations. Our system takes into the account voltage and phase factors. Though normaly a home is a single phase feed, in large appartment complexes and high-risers they sometimes use a single-derived from 3-phase power. That’s where it gets messy. Your voltage between phases is 208v not 240v so you have to account for that difference. Also as vic mentioned the most critical part of the system is the calibraton. :D Keeping a stable 200Amp load so you can calibrate is not an easy task. Also you have to remember there are two types of loads, capacitive and resistive.

-Roman

From the data I’m collecting it seems about right, but then again I don’t have a variety of different types of loads with varying power factors. From the comments so far I’m guessing then that theenergydetective has the same problems unless there is some magic happening in their setup that I don’t understand :)

:) You can not calculate anything without knowing the instantaneous voltage and power draw at same time. That’s what vic was trying to explain. The powerfactor is a phase relationship of voltage to power. For the most time you’ll actually be pretty low on the powerfactor, somwhere in 75-95% range. It’s unlikely that you’ll be at a 100% powerfactor ever, the capacitance in the powerlines cause loses and there are also a lot of other forces involved. The dynamics of power in a typical home is a lot more then what most people realize. Depending on where you live in the US. Your voltage can swing 20-30v. That’s a lot of voltage not to account for in the calculations. I also wanted to point out to make sure that you have the CT’s pointed in the right diraction. Otherwise current in one of the incoming lines your measuring will be subtracting from the other.

-Roman D

It is not possible by using only a CT, with a CT you can only measure current. If you also measure the voltage then all said above is possible.

This one is cheaper (because of the eth shield & wrt) but doesn’t allow you to use your own servers:
http://www.instructables.com/id/real_time_web_based_household_power_usage_monitor/

This one is just over the top.
http://www.kondra.com/circuit/circuit.html (not too much info here, saw on /. circa 2005)

As far as making this easier on the software side, can you make a leaky integrator in analog space and just sample the result of that? If you make the filter period long enough (longer than 1/60th of a second), it’ll do your RMS averaging for you, assuming the CT output is rectified.

Your other option is actually doing integration in software (which would be the same thing as the above filtering, except requiring cpu cycles).

I think it’s safe to assume that your line voltage will be 230 V, 3 phase within ~3% (when mine varies outside that range, the power company gets pissed at me)

@vic – my clunky way of estimating power was to empirically using a kill-a-watt come up with a scaling factor that would map the peak value measured on the ADC to watts. For the stuff I care to monitor the results were reasonable. Thanks to this post I’ve had some email exchanges with friendly people who really know what they are talking about; I might have some additional ideas to make this more accurate for different loads. I’m interested if people think it’s possible at all to measure power, compensating for power factor and whatnot, by simply using CTs on the mains. Makes me wonder how http://www.theenergydetective.com/index.html is doing it. -john

Re: Linksys router + Arduino + Ethernet shield: I have been using an Asus wl520gu wireless router with OpenWRT because it has a host USB 2.0 port, an internal serial port, it is cheap, and it is difficult to brick. I use a USB hub with a Duemilanove USB Arduino and 1GB USB memory stick plugged in (amongst other things!). The only downside of the wl520gu is that you need to use 2.4 with it because the wireless driver under 2.6 is not working well (AFAIK). Asus WL500 (not sure which – Deluxe? Premium?) have two USB ports, 32MB memory, and 8MB flash but are more expensive.

@andrew: Thanks! *very* important to know.

@vic: don’t be a tosser just because it doesn’t meet your anal standards. It is a work in progress, he calibrated it using the killawatt, and most importantly it gives him a usable indication of power usage over time.

@John – thanks for making the effort to write this up – Many of us lurkers really do appreciate it (even if we don’t say much normally!).