IEEE Power & Energy Magazine - November/December 2019 - 54

considerations were lead time, change of severity level since
previous alert, initial day of the week, valid day of the week,
time of day, severity of the event computed from a ramp rate,
the actions required, and need and ability to call back actions.
For the Irish case, the strategy for issuing an alert was
found to work best after the following two points had been
defined: 1) alerts are issued according to a simple scheme
(e.g., probability of an HSSD of 20% of capacity exceeding
25% probability for more than three consecutive forecasts)
and 2) the number of alerts is reduced by issuing updates
for changes in conditions once an initial alert has been
sent, thus preventing critical alerts from being overlooked.
This setup allows for the automatic filtering of unpredictable weather phenomena and reduces the risk for system shortages and unwanted volatility. Such probabilistic
HSSD forecasts also allow the end user to strategically
prepare for the increased risk of such events. With a suitable presentation and alert system strategy, the danger
of HSSD conditions can be determined and respective
actions taken in real time. With probabilistic forecasts
and measurements in both graphical and tabular formats,
confidence in the forecasts increases, and the operators
become more likely to take the correct actions without
wasting resources.

initial testing, EirGrid found that the amplitude of the events
is useful for these considerations. However, a temporal accumulation of the capacity that is in shutdown risk [Figure 9(b)]
was required for the control room staff to be able to consider
the total amount and length of reserve required and the associated down-ramps and up-ramps of previously down-ramped
capacity. The table in Figure 9(c) provides details for the control room staff's concrete planning.
Figure 10 shows a real-time example of two concurrent
HSSD events in which the forecast system issued warnings
when the probability for an HSSD of more than 20% of the
capacity exceeded 25%. The y-axis shows the probability
for the percentage of capacity that will experience HSSD,
and the difference between percentiles P10-P90 provides the
probability that the event will occur. Due to the variability of
forecasts in such situations, EirGrid decided that a warning
would only be issued if these thresholds were exceeded in a
minimum of three consecutive forecast cycles. Thus, outliers
are captured but not overinterpreted. In both cases, a warning was issued, and in both cases, it was appropriate.
One of the many challenges in developing the warning
system was that the alert frequency must be carefully configured to assure that alerts are taken seriously when they contain a realistic security concern for the system. The main

Case 2

35%

30%

Case 1

Case 1:
10% Probability of >84% Cutoff
30% Probability of >25% Cutoff
50% Probability of >10% Cutoff
90% Probability of ~5% Cutoff
(P90-P60)
Case 2:
10% Probability of >55% Cutoff
35% Probability of >30% Cutoff
50% Probability of >16% Cutoff
90% Probability of ~5% Cutoff
(P90-P550)

Result of Actual Cutoff Scenario:
Case 1: 25% Peak Cutoff
Case 2: 35% Peak Cutoff

figure 10. An example of a coupled extreme event that caused significant HSSD and how the HSSD warning system was
applied. In this example, the warning criteria were a minimum of 25% probability for 20% of capacity to experience HSSD in
a minimum of three consecutive forecasts. (Source: WEPROG; used with permission.)
54

ieee power & energy magazine

november/december 2019

IEEE Power & Energy Magazine - November/December 2019

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