PowerForecastErrors - consider envelope
Request Summary
Currently, the generated normally distributed errors are of the same magnitude regardless of the actual yield level of a power plant.
However, in reality, relative error levels are substantially larger when a power plant is operating at lower yields.
Thus, apply an "enveloping function" to stretch (compress) errors are low (high) yield levels of a renewable energy power plant.
Enveloping function could be of type a * exp (-x * b) + c
, where
-
x
is the actual yield (0 <= x <= 1), -
a
is a scaling factor -
b
is a decay factor -
c
is an offset
Since errors might be asymetric, two enveloping functions (one for negative errors and one for positive errors) might be required.
Use case
Provide more realistic power forecast errors for variable renewable energy plants.
Images / Sketches
Some relative yield forecast error factors derived with fitted error envelope exponential functions (red).