It really depends on what you are trying to accomplish, what your needs are, budget, time, etc. etc. etc.
Me personally, I wouldn’t spend the effort building what you are talking about. A friend of mine loves the idea however as it is a relatively cheap way of better understanding electrical power use. Let me expand SOME of the potential benefits and challenges.
The Pros to your approach
-
A lot of the benefits of energy management can be gained from the shape of the load curve (when, how long, what does the consumption profile look like). If you are metering many loads, its not really essential that you know exactly how many RMS Watt hours they have each consumed.
-
Should you be interested in having accurate consumption values; you can just create a multi-tiered metering setup where you have one or more accurate (revenue-grade?) energy meters which have low accuracy sub-meters installed below them. (I use the utility meter as my accurate meter)
The Cons to your approach
The biggest thing is that error will be accrued throughout the measurement chain; a bit here for this reason, a bit there for that reason, and could end up being quite significant with your final values. Here are some examples of where error will come from:
-
In order to have an accurate power or energy measurement you need to have RMS voltage and current measurements. They are by no means the same as just plucking an instantaneous value off of a current sensor every once in awhile and averaging them over time. Here are some links comparing average vs. RMS 1 2 3.
-
Aside from RMS sensors… keep in mind that without knowing the voltage, you can’t calculate power. And this changes over time so imaging that it is a constant like 115VAC or 230VAC is simply wrong. There is also sensor sensitivity and accuracy. CTs (current transformers) are meant to be sized accordingly for the circuits they measure in order to provide good accuracy. Putting a 100A CT on a 16A circuit which only really has 5A maximum going through it would provide very poor accuracy as it is in the bottom 5% range of the CT… looking at your CT, it says 3% accuracy between 10-120%. As the accuracy is non-linear, measuring below 10A (10%) could have anything like 10-30% error introduced. You need small CTs to measure small loads, bigger ones for bigger loads. My whole house never draws more than 40A!!
-
Next in the chain is the input of the analog signal. This split-core CT doesn’t say what its output range is, but that would need to either match the input range, or be smaller so that you can measure the range of the sensor. It probably uses millivolts, but using a more common industrial example, if you have a 4-20mA sensor (0-100% range) plugged into a 0-25mA input, you are losing precision due to unused 4mA upfront and 5mA afterward… meaning you never use 9mA of your 25mA input channel, or 36%! This compounds the fact you may only be using a tiny range of your current sensors range.
-
Once the analog signal is received, it needs to be converted to digital using an ADC. The precision of this input will determine how many possible digital values you will have in relation to your analog values. The larger the number of bits used, more digital values, meaning smaller increments can be measured and so you’ll have a greater precision of knowing when the analog signal changes slightly. An 8-bit analog input gives you 255 digital values, and 16-bit gives you 65535. So as your analog value changes linearly and gradually, you will see steps on the digital value. The number of steps determine the level of visibility you have on the analog value. So imagine measuring 0-10A on this CT with an imaginary 8-bit analog input (imagining the signal and input were perfectly matched) - 0A=0 0.5A=1 1A=2… I’m sort of guessing, but that gives you an idea. So, not possible to measure any changes below around half an amp. Obviously a different story if the input was 16-bit.
-
Then there is the software, and there are many parts of the measurement chain in the software as well, but the biggest one will be time and frequency. The current value will probably be bouncing all over the place, so an instantaneous reading will mostlikely be useless. You’ll need to do signal processing like a low-pass filter or running average on the incoming values which you read quickly. Then you’ll need to do any planned power and energy calculations that you have in mind… problem being that you can’t do much with only a current value.
This is why I am a believer in spending money on RMS energy meters (not just power); as pretty much all of these issues go away and then you just read the meter. 
I am also of the mindset of focusing the effort on what you do with the data, rather than spending all your effort on just getting it.