IEEE Power & Energy Magazine - November/December 2019 - 51

november/december 2019

>8,000

7,000 to 8,000

6,000 to 7,000

5,000 to 6,000

4,000 to 5,000

3,000 to 4,000

2,000 to 3,000

1,000 to 2,000

0 to 1,000

-1,000 to 0

-1,000 to -2,000

-2,000 to -3,000

-3,000 to -4,000

-4,000 to -5,000

-5,000 to -6,000

-6,000 to -7,000

-7,000 to -8,000

<-8,000

Frequency

This is particularly important because many resources have Capturing Extreme Conditions:
lengthy lead times and must be started well in advance to A Lesson From the Electric Reliability
ensure that they are online and available to serve load. The Council of Texas
staff uses the results of these multiday reliability studies to While renewable energy forecasting has advanced, it remains
determine whether additional generation needs to be started challenging to accurately predict the output from VERs. Figahead of time to mitigate real-time forecast errors.
ure 6 is a histogram of system-wide wind generation forecast
Currently, SPP takes into account up to 10 unique NWP errors in 2018 at the Electric Reliability Council of Texas
weather models and examines their potential spread to quan- (ERCOT). Although the forecast errors are concentrated near
tify the forecast's uncertainty level. If ensemble members zero, large forecast errors still occur. Probabilistic wind foreare strongly clustered (i.e., spread is low), SPP treats the sce- casting becomes particularly valuable for the large variations
narios with higher certainty. But if the ensemble spread is in wind generation that drive these large forecast errors. When
several gigawatts, commensurate with lower overall capac- these low-probability events are predicted within the distribuity, SPP runs studies to account for higher forecast errors to tion of the probabilistic forecast, their impact on the security
ensure reliability across the system. In cases where SPP com- and reliability of the grid can be assessed, and a correspondmits additional units to mitigate the risk of forecast errors ing mitigation plan can be developed.
that do not materialize, the cost of system reliability due to
Fluctuations from VERs impose a challenge to the balincorrect interpretation of the forecast is equal to the make- ancing services that operators perform to maintain satisfacwhole payments of the generators committed.
tory grid reliability. ERCOT uses a probabilistic forecast that
In addition to multiday reliability studies, SPP assesses is updated every 15 min to alert system operators of projecprobable forecast errors in the wind forecast during 1-, 4-, tions for large wind ramp rates during the next 6 h. Figure 7
and 8-h horizons across multiple days. For this uncertainty displays forecasts from ERCOT's Large Ramp Alert System
assessment, SPP applies probable errors and then evaluates (ELRAS). ELRAS provides estimates of the probability that
available capacity during each horizon for sufficiency. The wind ramp rates on the 15-min, 60-min, and 180-min time
probable error is determined by correlating the forecasted scales will exceed a set of predefined thresholds. An early
weather with past 95th-percentile wind forecast errors. In warning for a large wind ramp event will trigger an in-depth
cases where errors exceed available capacity for a particular horiHistogram of Wind Forecast Error in 2018
zon, actions may be taken to secure
4,500
capacity beyond that horizon to
Over Forecast
Under Forecast
4,000
ensure reliable system operations.
An internal uncertainty response
3,500
team evaluates weather and other
3,000
system conditions to identify highrisk time frames during which
2,500
SPP may face greater uncertainty
2,000
in load and VER forecasts. The
1,500
resulting heightened awareness of
forecast uncertainty impacts has
1,000
helped SPP maintain reliable oper500
ations during forecast excursions.
The group continues to work with
0
forecast vendors to gain better visibility of high-uncertainty intervals.
Other operators are applying
similar analyses or considering
whether to do so. For instance, the
Portuguese TSO values the probaMW Error (HSL-Wind Forecast)
bilistic forecast's ability to quanDay-Ahead Wind Forecast Error (MW)
tify the added reserves needed due
One-Hour-Ahead Wind Forecast Error (MW)
to errors in forecasting renewable
VER generation (wind, solar, and
hydro). The probabilistic informa- figure 6. A histogram of system-wide wind generation forecast errors at ERCOT in
tion provides the confidence inter- 2018. HSL: high sustainable limit. "Over Forecast" means under production as shown
val that the operator needs to man- by negative numbers on the left. "Under Forecast" means over production and as
shown by positive numbers on the right.
age the system.
ieee power & energy magazine

IEEE Power & Energy Magazine - November/December 2019

Table of Contents for the Digital Edition of IEEE Power & Energy Magazine - November/December 2019