I've skimmed many posts on this subject on this board. But forgive me if I'm wrong, but I did not see mention of a few simple "tricks" to minimize the number of ADC measuring channels required to monitor power usage on our 120/240 volt circuits.
Refer to http://brultech.com/home/software/index.php/software/downloadSoft/ECM-1240_Install_Guid.pdf for a commercially available accurate system that uses these tricks - diagrams show the techniques.
1. Monitor total power flux in/out of a power panel with 2 CTs, and a single voltage reference, connected to two analogue channels total. Use 2 (200amp typical maximum) CTs, each with its own burden resistor - wire them in series. Orient the loops over the two "hot" legs coming into the main panel so that when the house is drawing power, the sum of the voltages across the burden resistor increases with consumption. If the CTs are marked or otherwise have distinguished direction put them on the supply lines opposite of each other. (Otherwise you'll have to determine which way experimentally). So now you can see that if current is drawn from either leg, it will be measured. When current is being drawn from both it simply adds. Now you say, what about the voltages on each leg? -- well just measure one (safely with a little transformer and resistor network as explained elsewhere on this site) and assume the other "leg" has the same voltage magnitude but is 180 degrees out of phase. Because the power company regulates the voltage and there's (usually) plenty of available "juice" at the step down transformer, we deem this "equal but opposite" voltage assumption plenty close enough. I suppose one could experimentally measure and check... For solar back-feed setups where the solar energy comes into the same panel - power flux will reverse and should be measured as negative when solar is generating more than the rest of the house is consuming.
2. To measured combined power draw of multiple house circuits on one channel ... you can either do the same trick as 1, adding burden resistor voltages in the "analogue domain" -- OR -- use multiple CTs wired in parallel with an single external burden resistor. For the latter we are adding currents in the "analogue" domain. This is handy if you are short of channels (indeed the Arduino only has a few) and for your purposes, perhaps it is sufficent to group together consumption of "all electric lights on the first floor" or whatever. In other words, put a separate CT on the lead to each circuit breaker you wish to monitor, but combine several CTs onto the same Arduino channel. Be sure to orient the CTs on the leads so that they add properly. To compute true power, use the same voltage measurement as in (1).
Disclaimer: I am a professional Computer Scientist but an amateur EE. So please feel free to correct me if I got this wrong... But first checkout the Brultech doc and/or experiment yourself. And then explain exactly where I got it wrong.
Re: monitor Split-phase North American 240 single phase 2 legs with netural with just a few analog channels
I'm a Chartered Electrical Engineer here in the UK, and I think most of what you've written is accurate, and you'll find if you search carefully that everything you've written has already appeared on this site in one form or another. Indeed, we've had reports from those who have done the measurements that the voltage balance is indeed very accurate - within (from memory) about 1%. You can of course alternatively measure the line-line voltage and halve it in the software, I'd suggest this might give a marginally more accurate measurement if the largest loads are connected line-line and the loads connected line-neutral are the smaller ones.
I should also point out that if currents in separate cables are in phase, then you can also thread multiple cables through the same CT and you do the addition in the magnetic domain, thus it costs you but one ct and one burden resistor.
Re: monitor Split-phase North American 240 single phase 2 legs with netural with just a few analog channels
Thanks for your affirmative feedback ;-)
Multiple loads through a single CT. Yup, that too. But a difficult or awkward retrofit on American breaker panels where the L1 and L2 (opposite phase) breakers are adjacent (see Brultech diagrams again if you're not familiar with our Yankee equipment).
L1-L2 (line1 to line2) voltage measurement - the thought crossed my mind, but not feasible when using a little 120 volt to 9 volt AC wall-wart transformer. Nor the other common/cheap American 120 volt to low voltage thermostat or "door-bell" transformers.
Sorry for any redundancy - I did read through more than a dozen old posts where the contributors proposed all manner of other techniques.