Prediction is very difficult, especially if it's about the future.Niels Bohr
The Mojave Desert plant, built with the aid of a $1.6 billion federal loan guarantee, kicked off commercial operation at the tail end of December 2013, and for the eight-month period from January through August, its three units generated 254,263 megawatt-hours of electricity, according to U.S. Energy Information Administration data. That’s roughly one-quarter of the annual 1 million-plus megawatt-hours that had been anticipated. Output did pick up in the typically sunny months of May, June, July and August, as one might expect, with 189,156 MWh generated in that four-month period. But even that higher production rate would translate to annual electricity output of less than 600,000 MWh, at least 40 percent below target.
BB. wrote: »
I will give this quote:
Long term solar variations are easily in the 10% to 20% year over year averages (one Standard Deviation--guessing on my part)... And if you live near the coast (marine layer)--Your variations can be 50% or more.
Anyone that setups a financial model that fails if the output is 10% less than predicted (year over year)--Then the assumptions are not financially supportable.https://www.greentechmedia.com/artic...ity-production
This plant, apparently, had a whole bunch of sales reps, engineers, and MBAs lying to each other to get funding. So far, their "best guess" modeling has been off by a factor of ~1/2-1/4x (60% to 25% of "predicted" energy production so far).
Anyone that is selling their modeling software based on accuracy of predictions--They are just telling you (and your MBAs) what they want to hear. And when the project fails to meet predictions--They will tell you it was "bad luck".
I don't know if you are looking at this from an engineering or a financial point of view (predicting the future)--I would suggest that you get a daily data set that is (typically) 30-50 years long and run some standard statistical software (spread sheet) calculations. Get some Standard Deviations numbers (and average, median, etc.) numbers and feed that back through the financial model.
You (or somebody) needs to understand the stability of the financial model (let alone the stability of the engineering solution)--If you get a 50% predicted output over one day, one month, one year--What does that do to the financing--If, for example, 50% of predicted output over one month is enough to put your payments (and/or production guarantees) in default--Then solar power is not where you want to put your money.
Any decent modeling package can give you 10% accuracy with a fixed solar irradiation model/set of assumption (i.e., based on sun position, shading, etc.)... However, making decisions that assume 10% or 1% accuracy (~1% accuracy is typical lab grade ground based measurement accuracy--from a quick Google search) from the modeling software when overlaid with a high possibility of 50% variation (one chance in every few years) in actual weather related solar insolation over a monthly period--We are kidding/lying to ourselves.
BB. wrote: »
But what is "good"? How accurate can the modeling be when the "noise" in daily life is many times the "accuracy" of the model.