IEEE Power & Energy Magazine - November/December 2021 - 81

a more sophisticated balancing of costs and risks from different
reserve levels. Another example might be constructing
market offers for renewable output based on a preset risk
level, for instance, as is commonly reflected in the value-atrisk
limits to which traders are subject.
Figures 3 and 4 give an example of how probabilistic solar
predictions for a particular time and place, the simplest type
of forecast, are useful for informing the selection of a reserve
requirement based on a preset probability of exceedance. The
1,200
1,000
800
400
600
200
figures show how forecasts of the degree of uncertainty in solar
power could be used to develop weather-conditioned requirements
for the California Independent System Operator (CAISO)
real-time flexible ramp product. In particular, they show how
the width of the 50th percentile of global horizontal irradiance
(GHI) derived from 2-h-ahead probabilistic forecasts, such as
those in Figure 3, could inform ramp requirements.
The CAISO's flexible ramp product procures spare capacity
in upward and downward directions to accommodate not
7 April 2020 (Cloudy)
20 April 2020 (Sunny)
(Blank) <5% and >95%
5-25%
25-50%
Median
50-75%
75-95%
figure 3. A site-specific, 2-h-ahead probabilistic global horizontal irradiance forecast [7 and 20 April 2020, site location:
Los Angeles, California], generated by the Watt-Sun probabilistic forecasting system. (Source: IBM; used with permission.)
100%
Up-Ramp Requirement
Up-Ramp Requirement
90th Percentile
75th Percentile
50th Percentile
200 300 400 500 600 700
Region of
Region of
Low Solar Uncertainty
(a)
High Solar Uncertainty
Forecast Width 25th-75th Percentile Interval,
Global Horizon Irradiance (W/m2)
200 300 400 500 600 700
Region of
Region of
Low Solar Uncertainty
(b)
figure 4. (a) An example of the classification of days into two types (low and high solar uncertainty, based on 50% prediction
intervals for the forecast global horizontal irradiance) and dependence of upward real-time ramp uncertainty (from
11 a.m. to 2 p.m., May 2019, the CAISO system). The ramp requirement is based on the 95th percentile within each classification.
(b) An example of the quantile regression relationship between GHI uncertainty (50% prediction interval from the
forecast) and upward CAISO real-time ramp error. Three estimated quantiles are shown for ramp error (90, 75, and 50%).
(Source: Yijiao Wang; used with permission.)
november/december 2021
ieee power & energy magazine
81
High Solar Uncertainty
Forecast Width 25th-75th Percentile Interval,
Global Horizon Irradiance (W/m2)
100%
Upward Forecast Error (Normalized)
Global Horizontal Irradiance (W/m2)
Upward Forecast Error (Normalized)
6:00
6:45
7:30
8:15
9:00
9:45
10:30
11:15
12:00
12:45
13:30
14:15
15:00
15:45
16:30
17:15
18:00
18:45
19:30
6:00
6:45
7:30
8:15
9:00
9:45
10:30
11:15
12:00
12:45
13:30
14:15
15:00
15:45
16:30
17:15
18:00
18:45
19:30
IEEE Power & Energy Magazine - November/December 2021

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2021
