Electricity Consumption - Option D

You will need to calibrate the model by energy end use on a monthly (and annual) basis.
1. if you trend this data you could figure out energy use
2. how will you determine energy demand? It will need to be measured.
3. subtraction from metered data is a viable method
Yes you would need to separate out the electric end uses in order to calibrate the model. Whole building metered data is not sufficient.

But how do you know the actual energy demand for those items? Rated and actual are often very different. You will need to measure actual consumption. This can be through direct sub-metering, spot measurements, short-term trends, derived from other measured data, etc. This cannot be based on the same method used to make the prediction. You need to measure something in the actual operation of the building to confirm the consumption you predicted. If it is not the same then you make adjustments (calibrate) to the model until the values match within your tolerance range.

Under Option D you are attempting to make adjustments to your model so that it matches the utility bills (monthly and annually by end use). So what information do you need in order to do that? Totals are not enough because even if you match exactly you could just be lucky. We did an energy model a couple of years ago that was within 6 kWh based on the first year of actual operation. When we looked at the monthly data we under-predicted heating and over predicted cooling and on an annual basis we just got lucky. But as you can see just because some numbers match does not mean we had a calibrated model. The whole point of Option D is to create that calibrated model which is then used to verify savings relative to a calibrated baseline.

If you understand what you are trying to do the question of what to meter and how to do it usually falls into place.