IEEE Power & Energy Magazine - November/December 2019 - 55

Use of Probabilistic Forecasts by the TSO:
A Spanish Case
Red Eléctrica (REE), the Spanish TSO, is responsible for
dealing with the imbalances between generation and consumption in the electrical system; thus, it has generation
reserves that provide frequency regulation. The physical
value of these necessary reserves is determined by variables
that have a stochastic behavior. The following items stand
out in the Spanish system:
✔ deviation between market schedules and the wind
forecast
✔ forecast errors in other variables (demand, solar, cogeneration, and so on)
✔ unscheduled outages of the generation units.
The large amount of wind generation (currently 23,070 MW)
installed in the Spanish peninsular system has a significant
impact on system balancing. Short-term (1-2-h) fluctuations
in wind energy on primary, secondary, and tertiary reserves
remain less important because demand and sudden unscheduled outages of thermal units more significantly affect such

reserves. However, the longer-term (8-12-h) fluctuations have
a primary impact on system reserves, especially on spinning
reserves. The calculation of these reserves is the main role of
probabilistic forecasts at REE. Both probabilistic and deterministic methods are used.
The probabilistic method assumes uncertainty as an independent probabilistic variable and calculates the combined
probability of loss of reserve as the sum of a convolution
of pdfs calculated with the type 1 method (see Figure 1) and
the additional reserve. The type 1 probabilistic method takes
into account demand forecast errors, wind forecast errors,
and unexpected outages of thermal units. The additional
reserve allows for possible uncertainties regarding deviations between forecast and market schedules and the management of international interconnections.
Peak-hour demand forecast errors for different horizons are fit to Gaussian distributions [Figure 11(a)], while
the wind forecast errors [Figure 11(b)] are fit to Weibull
distributions. For thermal unit outages, it is necessary
to calculate the probability that a unit trips k times in a

Probability Density
Functions of the Demand Forecast Errors

Probability

0.012
0.01
0.008
0.006
0.004

(MW)

Probability Density Functions
of the Unavailability of Generation Units

0

0

00
5,

0

00

00

4,

0

3,

0

Convolution of Previous
Probability Density Functions

-8
,0
0
-6 0
,0
0
-4 0
,0
00
-2
,0
00

00
0
-1
0,

50
0
1,
00
0
1,
50
0
2,
00
0
2,
50
0
3,
00
0
3,
50
0
4,
00
0
4,
50
0
5,
00
0
(MW)

0
2,
00
0
4,
00
0
6,
00
0
8,
00
0
10
,0
00

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-0.1

Probability

0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0

0

00

(MW)
(b)

(a)

Probability

2,

-5

1,

00

,0
0
-4 0
,0
00
-3
,0
0
-2 0
,0
00
-1
,0
00

0

0
00

0

5,

0

00
4,

0

00
3,

0

00
2,

00

0

0.002
0

1,

00
,0

00

-1

00

,0
-2

00

,0

,0

-3

-4

00

Probability

0.016
0.014
0.012
2
1
0.01
0.008
8
0.006
06
0
6
0.0
004
0
0.004
0.0
0.
0
..0
00
002
0.002
0

,0
-5

Probability Density
Functions of Wind Power Forecast Errors

(MW)

(c)

(d)
h-1

h-4

h-10

h-20

h-36

figure 11. An example from the REE system of the convolution of pdfs for demand, wind, and outages: (a) demand
forecast error pdfs, (b) wind forecast error pdfs, (c) pdfs of unscheduled outages of thermal units, and (d) convolution of
demand, wind, and outage pdfs.
november/december 2019

ieee power & energy magazine

55
IEEE Power & Energy Magazine - November/December 2019

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