[1] Please let us not get too hung up on terminology. I understand solar does not save energy, it produces energy. But I am trying to keep the language simple, so that we can address difficult topics without too much taxing hard thinking. Let us just agree to call it savings, as it does reduce the electricity bill, saving the client money.
[2] This has always been true, regardless of if there is a solar PV system or not. For example, suppose you put in a new chiller plant in a building. Most of your savings should come during On Peak hours, which in most parts of the country is the most expensive TOU period. If you used blended rates, then you have cheated the ESCO, as the blended $/kWh is less than the On Peak $/kWh. Using TOU becomes all the more important with a solar PV system, as the solar produces kWh during certain TOU periods, not averaged out over all TOU periods.
[3] This decision is made by the non-engineering financial types who are trying to make the project financially viable. But the thing is, they really need to understand the complexity that solar is bringing to the project, which means they need to speak to the M&V person before agreeing upon the rates.
[4] Why are TOU period regressions often bad? Well, consider that if you are doing monthly, weekly, or daily data, you have HDD and CDD that are associated with the entire day, not with the TOU period. Off Peak could be at night, when ambient temperature is low, and is poorly correlated with average temperature for the day. Off peak is often even worse, as for many utilities, it applies to the entire day during weekends and to only the night times for the weekdays. That is a tenuous relationship.
[5] Briefly, let me attempt to explain this method in a footnote. If you choose tune only sum of TOU periods, then you will do a regression only on the total kWh. But how do you separate out the baseline usage back out into TOU periods? Metrix and Option C will determine the ratio of On Peak kWh to total kWh for each billing period in the tuning period. These ratios will be carried forward during the performance period. For example, if during the base year, 35% of the total usage was on peak in June, then during the performance period, the software will assume that 35% of the total usage is on peak in June.
[6] That may have been true, until we experience an end of the world event like a pandemic, where the building is closed down, but this never happens. Or does it? Perhaps it would be wiser not to make the simplification as I suggest.
[7] https://power.larc.nasa.gov/data-access-viewer/
[8] TMY data over the years demonstrates in increase in temperature.
[9] X = Y + 3 is the same as X-3 = Y