The search space (or optimizer values) are for the DC rated capacity of the array, and the cost table is also for the DC rated capacity, regardless of what bus it is on. If you have a 10 kW array in HOMER (search space value is 10), that array will produce 10 kW DC in standard conditions (Nominal operating temp and 1 kW/m^2 irradiance). The actual output on the bus will be less, and the loss calculation is the same whether its on the AC or DC bus (although the numbers you enter would probably be different). Losses are:

- Derating factor: If the derating factor is 90%, the 10 kW is reduced to 9 kW (1 kW lost)

- Temperature effects: For the canadian solar, the NOCT is 45 C, and the temperature effect coefficient is -0.41%. So if the cell is at 45 C, there is no change here. If the cell is at 55 C, then you lose another 4.1% or 0.41 kW. This is (55C - 45C) * 0.41%.

- Inverter (AC bus) or MPPT (DC bus). The name changes depending on what bus the PV is on, but the inputs and the calculations are identical. You can enter a single efficiency or an efficiency table, and HOMER will calculate the losses and stack them on top of whatever was calculated from the derating factor and temp effects. The output is also capped at the capacity of this dedicated converter (inverter or MPPT), which you specify in the search space box in the Inverter / MPPT tab at the bottom of the PV menu.

The efficiency is effectively multiplied together for these. If you have a 95% efficient inverter and using the examples above, you could calculate the overall efficiency between the DC output and the bus:

Efficiency to bus = Derating * (1 - Temp effects) * Inverter efficiency

Efficiency to bus = 0.9 * (1 - 0.041) * 0.95 = 0.82 or 82%

Notice that you could simply specify a derating factor of 82% and have the same output for this time step (but of course, the temperature effects change timestep to timestep based on the ambient temperature and the solar irradiance! and an efficiency table in the inverter would also not fit a single derating factor.). So of course, you can't skip the temperature effects, but in a lot of cases, you might be better off including the inverter losses in the Derating Factor input rather than actually checking the box "Explicitly model Inverter" and specifying an inverter in detail. In this case you would leave the "Explicitly model inverter" box unchecked, and include the cost of the inverter, mppt, and any other power electronics in the cost table (also based on DC capacity).

Since you need to pick a size(s) for the inverter in the search space, it makes optimizing the total PV array (DC) capacity hard. Modeling the inverter explicitly is mainly useful when you've already run an optimization without it and know about what size array you want. Then you can pick a few sizes of inverters and see if it makes sense to undersize the inverter, and by how much (DC to AC ratio). You can also enter an efficiency table, which could have a small effect on the ideal array size, economics, etc.