Two questions: First regards derating. It seems it is really important to get these parameters right. For example, if you have a lot of snow or dirt build-up, do you somehow try to guesstimate an anticipated average value over the course of the year? Are there more sophisticated tools for more accurately calculating a derating factor?
Also, you mentioned the NEC requires to size an off-grid for largest load. Wouldn't you want to size for the largest load of all combined devices that typically would be on at one time?
Question about estimating production:
When you are dealing with so many variables, including weather, which is famously unpredictable for meteorologists, you cannot be exact.
Even if you look at the way shading is calculated. Most software does not take tree growth and bypass diodes into consideration. They just determine how much direct sunlight they are getting in a year. We do get production from sunlight that is not direct. Software is also not taking climate change and droughts into consideration, however for the last 5 years, the drought in California has definitely made some more PV kWhs.
Snow is also very difficult to enter into the equation. Most of the time people think that snow is going to hamper production more than it does. Usually the snow will blow off, slip off or melt off within a day of the snowfall. In some cases however, it can freeze and stick. You can have a location where show never sticks for 5 years and then have conditions that make it stick for a week.
The same can go for soiling. You can have rain every month for 5 years and then a year where the array is not washed by the rain for 6 months, which would make soiling deratings hard to predict.
The best we can do is take educated guesses, check our guesses against what is actually happening and re-adjust our predictions about the future.
Here is a SolarPro article about production modeling where they compare different types of software and get different answers:
It was interesting that PVWatts, the free software came out in the middle for production.
What it comes down to is the person entering the estimates for derating. They will probably overestimate a few factors and underestimate a few other factors, making it closer to even in the end.
You can also see eager startups overestimating production to sell the job.
The financed jobs will have specific requirements for predicting output, so the banks know they are getting a good investment.
Question about inverter sizing for stand-alone systems.
690.10(A) Requires that the stand-alone (off-grid) inverter be sized to be able to handle at least the largest load. Many installers in the off-grid field, where inspectors are seldom seen do not adhere to this requirement. They actually will have some large loads, such as an arc welder that the inverter cannot power and they will turn on a generator when they are arc welding.
You do have a good point that we should ideally design a system that will be able to handle the loads that we can expect to be on at one time. This is not a safety issue that the NEC covers, but a good idea if you want to vacuum, watch TV and make coffee at the same time. If you have a customer that is always having to reset their inverter, it will not make you popular.
The people writing the NEC decided that people living off-grid usually know that they cannot live like grid-connected people and have to be aware of how many things they can use at once.