Energy and Power En gi neering, 2011, 3, 537-546
doi:10.4236/epe.2011.34066 Published Online September 2011 (
Copyright © 2011 SciRes. EPE
Characterization of Peaks and Valleys of Electricity
Demand. Application to the Spanish Mainland System in
the Period 2000-2020
Fermín Moreno
Division of Inspections, Liquidations and Compensations, Comisión Nacional de Energía (CNE), Madrid, Spain
Received July 11, 201 1; revised August 14, 2011; accepted August 29, 2011
Energy planning must anticipate the development and strengthening of power grids, power plants construc-
tion times, and the provision of energy resources with the aim of increasing security of supply and its quality.
This work presents a methodology for predicting power peaks in mainland Spain’s system in the decade
2011-2020. Forecasts of total electricity demand of Spanish energy authorities set the boundary conditions.
The accuracy of the results has successfully been compared with records of demand (2000-2010) and with
various predictions published. Three patterns have been observed: 1) efficiency in the winter peak; 2) in-
creasing trend in the summer peak; 3) increasing trend in the annual valley of demand. By 2020, 58.1 GW
and 53.0 GW are expected, respectively, as winter and summer peaks in a business-as-usual scenario. If the
observed tendencies continue, former values can go down to 55.5 GW in winter and go up to 54.7 GW in
summer. The annual minimum valley of demand will raise 5.5 GW, up to 23.4 GW. These detailed predic-
tions can be very useful to identify the types of power plants needed to have an optimum structure in the
electricity industry.
Keywords: Peaks of Demand, Security of Supply, Demand Forecasting
1. Introduction
There is a broad consensus about the need of accurate
models for electric power demand forecasting (e.g., [1-3]).
Demand forecasting is necessary for the operation and
planning of electricity systems in terms of power and en-
ergy. The interpretation and use of its results are critical in
energy efficiency and sustainability issues. The role of
forecasting is essential in key decision making, such as
investments in power capacity, infrastructure development
or electric system management. Thus, depending on the
time horizon selected, demand forecasting can be classified
as: short-term (from 1h to 1 week); medium-term (from a
week to a year); long-term (from a year to ten years) and
very long-term (more than 10 years). Such a classification
corresponds to the time needed to provide different types of
reactions in implemented energy policies. A nuclear power
plant, for instance, cannot be built in the medium-term. In a
different scenario, wind power must be anticipated with 3
or 4 hours, to have some gas fired combined cycles (GFCC)
ready as back-up power. The underestimation of electricity
demand could lead to undercapacity (in power and grid),
which would result in poor quality of service including
localized brownouts, or even blackouts. On the other hand,
an overestimation could lead to overinvestment in power
plants and grid that may not be needed for several years.
Electricity demand forecasting is aimed at a triple ob-
jective due to the crucial importance of electricity in
modern economies: 1) security of supply: especially im-
portant in electricity, because of the practical impossibil-
ity to be stored as such and the long time required to
build power stations (typically varies from 5 years for
gas fired combined cycles and 10 to15 years in the case
of other thermal plants and new dams); 2) environmental
quality in the grids development and the sources of en-
ergy (includes requirements at local scale, regional scale
and global); 3) low costs, which are associated to a
competitive sector. These criteria can be understood in
different ways: if the electricity system is government
operated, the criteria used for the development of the
electricity industry are reliability, cost and acceptable
environmental impact. In a private system, the most im-
portant criterion is profitability. It is presumed that mar-
ket forces will guide the electricity industry to a close-to-
optimum state, which is something arguable, unless de-
mand can be pr edi c t e d acc u rately.
Indeed, in Spain there is a combination of both: pri-
vate investment, which is driven by public interest. Thu s,
the role of regulation is essential in electricity planning.
The development of electricity and gas transmission
networks in Spain requires government approval [4] and
the private investments are fully paid by the Spanish
electricity system [5] with a feed in tariffs (FIT) model.
Investments in generation are also private, but there are
incentives, supported by the FIT model: electricity gen-
eration by renewable energies [6] and the installation of
conventional generating capacity [7]. At this point, sev-
eral forms of regulation, specially aimed at ensuring an
adequate level of supply, are discussed, for instance, in
[8] or [9]. The regulation in Spain also sets, for example,
the tariffs access to networks ([10] and [11], in 2009) or
the official demand response programs1 (DRP). Energy
suppliers may also offer the DRP, but their existence,
and therefore its consideration, is unknown for a forecast
of electricity demand for the whole Spanish system.
Governments and/or utilities must forecast demand for
the long run (10 to 20 years), make plans to construct
facilities (grids and power plants) and begin their devel-
opment according to reliable expectations of growth or
slowdown. A fast growth rate of electricity demand will
consume the excess of base-load capacity; this will result
in higher electricity prices and lower environmental
quality of generation units, jeopardizing security of sup-
ply. Thus, an error in the demand forecast cannot be
overcome immediately and the demand cannot be met
because of lack of generating capacity or capacity of the
grid. When this happens, some authors, as [2], consider
that a reduction of the nation’s gross domestic product
(GDP) may be registered because of losses in production
of consumers in the industrial and commercial sectors;
and other aspects, as social disruption, are difficult to
estimate in terms of monetary value. Under those cir-
cumstances, other tools are necessary to prevent potential
blackouts or brownouts: DRPs can help SO to respond to
those contingencies and to manage and enhance the
overall reliability of the system ([15] or [16] give other
examples of DRP); for instance, the lack of DRP is con-
sidered as a critical reason for blackouts in California in
2000 [17]. In addition, there is an efficiency argument in
place DRP; the benefits arise from even a small amount
of demand response [18]: a small reduction in peak de-
mand can significantly reduce both energy (lower fuel
costs and higher efficiency of power plants) and capacity
costs. It is worth noting that some authors do not fully
agree with all the DRP [19], considering that stimulating
customers to refrain fro m purchasing products they wan t
seems to run counter to the normal operation of markets.
In any case, the use of electricity demand forecasting
must be also addressed to calculate the total amount of
DRP that the system needs.
This paper distinguishes three levels of demand fore-
casting: 1) macro level: national electricity consu mption;
2) intermediate level: sectorial electricity consumption;
and 3) micro level: the shaping of each load curve, na-
tional and sectorial. The wo rk focuses on the micro-level
with the aim of forecasting peaks of electricity demand
on the long term, up to the year 2020 . The macro level is
not addressed here because governmental forecasts ([4]
and [20]2) are used for electricity demand. These fore-
casts have served as boundary conditions to face the in-
termediate level in this paper, where sectorial trends of
mainland Spain’s electricity demand are estimated. The
methodology allows comparing and validating the Span-
ish energy planning forecasts.
The paper is organized as follows. Section 2 presents
the main concepts of the methodology. Section 3 shows
the main statistics and sources of data of electricity de-
mand in Spain (historic records and forecasts). Section 4
introduces the mathematical methodology and the main
boundary conditions. In Section 5, the results and their
validation are presented. Section 6 contains some ideas
about the needs of power generation facilities in the
decade 2011-2020 considering the results of Section 5.
Some concluding remarks follow in Section 7.
2. Methodology: Main Concepts
Another widely extended concept is the influence on the
electric demand of a country of various factors, as it is
said in [1,21-23]: weather cond itions, number of daylight
hours, electricity prices, day of the week, electricity us-
age habits, demographic parameters, influence of busi-
ness cycles, GDP or economic growth. Social events, as
a football match can also affect the demand of electricity:
U.K. National Grid, plc. ( ex-
plains that after England’s world cup semi-final against
West Germany in 1990 the demand soared by 2.8 GW,
close to 10% of the demand. This aspect is also high-
lighted by the Spa ni s h electricity SO [24].
Other aspects which may affect the future electricity
demand are: the introduction of energy saving measures
and continuous improvement in the consumption of elec-
trical equipment, which will contribute to a more effi-
1Tariffs with demand response mechanisms (DRM), as direct load
control (introduced by [12] and developed by [13]), are in use in Spain.
According to [14], on December 31st, 2009 were in force 142 contracts
in the mainland system, with an associated power of 2,112 MW.
2Which considers the consumption patterns in Spain to meet the objec-
tives known as 20-20-20, defined by the European Directive 2009/28/
Copyright © 2011 SciRes. EPE
cient demand scenario [25]; partially related with effi-
ciency, restructuring or renewing of networks throughout
the forecasted period would reduce electricity losses [23];
the evolution of temperature dependence patterns result-
ing from climate change [26]; the introduction or change
of DRP. For instance, the reliability of some DRP in
USA reached a load reduction ranged from 1.8% to 2.3%
over expected power [15]; or the change in the winter
peak of residential consumers with time of use rates
(TOUR or tariffs, TOUT), ranging from an increase of
0.04% to a reduction of 2.44% [27]. In Spain, TOU tar-
iffs are in place, defined at [28] and [7], and have been
considered to characterize t he sec t ori al consumption.
In the period under review, a business-as-usual (BAU)
scenario has been considered, regarding to the year 2009.
From the analysis of preliminary results for the BAU
scenario, other alternative scenarios have been consid-
ered for the winter peak and summer peak of electricity
A conventional electricity planning process forecasts
the annual increases in peak power over a chosen time
frame. Then, the methodology is applied. For instance,
the trend method to the historical data of growth in de-
mand for electricity projecting it into the future, between
10 or 15 years to accommodate construction times for
base-load power stations [2]. Electricity demand scenar-
ios can be developed applying different assumptions to a
Cobb-Douglas function, which is considered to reflect
properly the nature of demand developments [21]. Typi-
cal parameters used in that function are income and price
elasticities, increase in energy efficiency and predictions
of GDP. The Cobb-Douglas function can be applied to
each of the economy sectors individually, in order to
obtain disaggregated electricity consumption.
In this paper, the macro level of forecasting is not
faced, because the forecasts of energy planning have
been used [4,20]. The intermediate level is deduced from
those documents.
The load curve shape is dependent on many factors,
such as economic development, climatic conditions, and
electricity usage habits. The prediction of the shape of
the load curve can be developed using, for instance, dis-
aggregation-aggregation, econometric techniques, or a
combination of them [3]. For instance, in [23 ] it is used a
local utility’s estimation of load curve till 2025 calculated
from the trend in the load cu rve and the expected growth
rate and considering a reduction in electric losses rate.
The micro level of forecasting is faced here using the
disaggregation-aggregation technique in order to obtain
the load curve shape of each sector of consumption se-
lected. The electricity demand in the same month of dif-
ferent years must have very similar variations from the
general rising trend because this demand is mainly con-
trolled by climatic factors [29]. If previously said is as-
sumed, then the influence of weather conditions is also
included in the load curves for the base year. Also, the
influence of, for instance, working patterns or social
events is included in those load curves of the base year in
a BAU scenario. The load curves are projected into the
future, taking into account the annual sectorial consump-
tion of electricity identified in the intermediate level for
each year. The sectorial load curves are aggregated (for
each hour of the year) and the system’s load curve is
obtained for each year of the period analyzed. Finally,
the peaks and valleys of electricity demand are iden tified
for each year.
The methodology is close to the trend method (due to
the assumptions in the macro and intermediate levels),
but the results are richer, because the full load curves of
the system and each sector of consumption are available,
which will allow a deeper study or other uses. The for-
mulation is easy and can be developed, for instance, in
an Excel workbook in a normal laptop or PC.
The selected base year (2009) contains patterns of ex-
treme temperatures, both in winter and summer, which
will allow a good extrapolation to predict possible peaks
of demand in the future. In the months of January and
December, some days were observed with temperatures
up to 7 degrees Celsius below average, as can be seen in
[30]. Up to date, in Spain, the year 2009 has been the
third warmest year in the time series of records since
1961 [31].
In the Spanish energy sector, electricity demand is
understood as the electricity available for use in the
market, prior to transmission and distribution, as it is
provided by conventional generators [32] (mainly nu-
clear, coal, GFCC and large hydro technologies). Elec-
tricity demand is technically defined as power station bus
bar demand and excludes the self-consumption of auto-
producers (electricity that has not gone through th e grid).
For each type of consumer, depending on the level of
voltage and hourly TOUT, coefficients of electricity
losses are used (contained in [10] and [11] for the year
2009), in order to transform the measure of energy at
consumption to generated energy each year.
Energy used by pumped-storage (3,73 6 GWh in 2009)
and consumption at generation facilities (7,122 GWh in
2009), that it is paid at market pr ice, is not considered as
a component of final demand forecasted in this paper.
3. Spain’s Electricity Demand: Sources of
3.1. Main Statistics
All statistics and sources of data used in the paper
Copyright © 2011 SciRes. EPE
come from: Spanish System Operator (, [14]
and [33]); National Energy Commission (CNE’s bulle-
tins; and UNESA (Spanish Association of
Electricity Industries) [3 2].
Table 1 shows the evolution of peaks of electricity
demand in the decade 2000-2009. These data will be
employed as a first validation of the results of the model.
Figure 1 shows the evolution of high voltage (HV,
voltage greater o equal than 1 kV) demand and low vol-
tage (LV) demand in Spain in the decade 2000-2009. In
[34] was pointed out that there are robust evidences that
both industrial and residential electricity demand have a
symmetric distribution for G7 countries. Spain does not
belong to G7, but the distribution of demand in Figure 1
shows that the industrial demand (HV) is parallel to the
residential plus commercial sector demand (LV). The
only exception in the period shown is the year 2009,
when industrial demand for electricity was strongly af-
fected by the economic crisis. Data from Figure 1 will
be used to project the model into the past and compare
the results with those records in Table 1.
Table 1. Annual evolution of maxima of electr icity demand.
Source: [14], [32] and [33].
Peak (MW)
Year Winter Summer
2000 31,951 29,363
2001 34,948 31,249
2002 37,274 31,927
2003 37,724 34,537
2004 38,210 36,619
2005 43,378 38,511
2006 42,153 40,275
2007 44,876 39,038
2008 42,961 40,156
2009 44,440 40,226
Figure 1. Structure of electricity supply in Spain and its
evolution taking into account voltage (HV and LV). Source:
[32], CNE (www.cne.e s) and own calculations.
3.2. Electricity Demand Forecasting in Spain
Most of the electricity demand forecasts in Spain are
mandatory developed by the SO [35]. Short-term fore-
casts are provided by the SO in real-time at its webs (in
real-time at, one
week ahead at
Some of SO’s forecasts in the medium-term [36] are
used as an update of those included at energy planning
[4], which are the long-term projections. The available
long-term predictions were made before the economic
crisis, so that should be considered outdated.
Subsequently, with the aim of meeting the obligations
undertaken by the Spanish government with the Euro-
pean Directive 2009/28/EC (objectives known as 20-20-
20), it has been developed a new planning in the field of
renewable energy [20] that includes predictions of en-
ergy demand for the period 2011-2020. In the pro- posed
scenario, net electricity demand in mainland Spain will
raise up from 252.0 GWh in the year 2009, to 330.6
GWh by 2020 (estimated from [20]) for the reference
scenario. The renewable energy planning considers all
the scenarios with growing energy consumption, both
primary and final (opposed to Germany in [25], for ex-
ample). On the other hand, the electricity grid losses will
be reduced to 8.7% of energy supplied (close to 8.5%
lower than those in 2009).
4. Boundary Conditions and Description of
the Model
Electricity demand forecasts in a BAU scenario and oth-
er specific criteria and data listed in [4] and [20] (the
energy planning documents) have been taken into ac-
count in order to define the intermediate level. For in-
stance: demographic evolution; evolution of energy con-
sumption at the industrial sector; or the possible massive
introduction of electric vehicles (it has been included in
the forecast on the basis of the available projections of
pilot projects and initiatives, evaluating their progress
throughout the pe riod under review, with the criteria and
limits of [4]).
The residential energy supply has been considered
without TOUT, although all the existing electrome-
chanical energy meters are supposed to be replaced by
2018, gradually during the analyzed period, by equip-
ments with hourly data of energy and remote controlled.
This will allow offering DRM to residential consumers,
but there is no date, and may change the patterns of
residential consumption at peaks of demand, as it is
said in [27 ].
There is no record of specific actions in a next future
(with the exception of grid losses reduction), in the
Copyright © 2011 SciRes. EPE
field of energy efficiency, which may notably reduce
the electricity consumption of devices and which may
require specific simulation. For instance, most of the
substitution of incandescent bulbs by energy saving
bulbs has just been done in Spain and it is considered
in the model.
Various sectors of consumption have been different-
tiated depending on data availability (with a criterion
similar to [24]). The sectors selected are formed by
groups of consumers in mainland Spain with similar
level of voltage and TOUT. The structure of supply is
shown in Figure 2 for the Spanish mainland market in
2009 and the forecasted in the intermediate level by
For each year of the period analyzed, the load curves
for those sectors are obtained with (1), collected at [37]
and currently used to estimate the load profiles that
H V > = 36 kV -TO U 6PHV > = 36 kV -TO U 6P
HV < 36 kV -TO U 3PLV wi t hout TO U
Figure 2. Structure of electricity supply in Spain 2009 (up)
and forecasted in 2020 (down) taking into account voltage
and periods of pricing (P) of TOUT. Source: CNE’s bulle-
tins ( and own calculations.
will be used for the liquidation of hourly energy meas-
ures at the Spanish electricity market. Equation (1) is
applied here to those consumers without TOUT (TOUT
with 1 period) and TOUT with 2 or 3 periods of pricing
(see Figure 2).
,, ,,,,
mdhjtJT p
mdhp dt mJ
dD mJ
mJ i
 
 
mdhp: Calculated hourly measure for sectorial
consumption “c”, with profile “i”, in the hour “h”, of
day “d”, month “m” corresponding to the energy of
period “p” recorded by the measurement equipment.
P: Profile for sectorial consumption “i”, month
m”, day “d” and hour “h”, which represents th e rela-
tive weight of that hour in the year. It has been ob-
tained for the year 2009 from
,, , ,
MC : Incremental energy measured for cus-
tomer “c”, between the day “t” of the month “j” and
the day “T” month “J” for the period “p”. The neces-
sary information has been obtained or estimated from
CNE’s bulletins (
Dm: number of days in month “m”.
The hourly electricity demand of the HV segment,
with voltage higher than 36 kV and TOU tariffs with 6
periods of pricing, was determined as the average load
in each period. Many large industrial consumers have
load curves with very little hourly and seasonal varia-
tion, so that its characterization in the model should be
consid ered includ ed in tha t segment.
Finally, for the year 2009, the load curve of the de-
mand in HV, with voltage less than 36 kV and TOU
tariffs with 6 periods of pricing, was determined by
subtrac ting, to the tota l demand, the calcu lated de mand
of the other sectors. Finally, it is projected with the
results o f the intermedi a te leve l. This consu mp tio n s eg -
ment represents large industrial customers with hourly
and seasonal variation of load. Additionally, this con-
sumption segment absorbs defects of allocation of
consumption of the other sectors.
Figure 3 shows those curves for the week when the
winter peak is registered in 2009. Their aggregation,
hour by hour, is the Spanish electricity system’s load
curve for that week.
5. Results and Validation
The validation of the model has been done comparing
projections of peaks of demand in the period 2000-20103
with the data of Table 1. In this way, the fitting of the
model to the reality of the Span ish electricity system can
3Some preliminary data of the year 2010 were av ailable.
Copyright © 2011 SciRes. EPE
Figure 3. Load curves of each group of consumption (Fig-
ure 2). Winter peak week in 2009 (ene = January). Source:
Own calculations.
be analyzed. The results are shown in Figure 4; they
show a similar trend with the records of peaks of demand
(Table 1), both in winter (annual) and in summer.
The biggest discrepancies in the estimated annual
maxima are observed in the periods 2002 and 2005,
with a deviation of –4.2% and –2.9% respectively, and
with a tendency to reduce the deficit with the proximity
to the base period (Figure 5).
The trend in the difference between the estimated
peak and annual peak of demand may be justified by
the introduction of efficiency measures reducing the
energy needs in the winter maximum. Another reason
fo r th is trend may lie in the gradual disappearance of the
regulated tariffs for industrial consumers (tariffs that
ensured a maximum price of energy and were fixed by
the government), a process that ended in 2008. This ef-
fect was parallel to the contracting of supply of these
consumers with free agents, paying a higher price for
energy at peak than the price they paid for the former
regulated tariffs, forcing them to optimize their processes
and displacing consumption to valley hours, if it was
The largest discrepancy in the estimation of summer
peak corresponds to the year s 2006 and 2010, –0 .7% and
Figure 4. Annual evolution of maxima of electricity demand
in winter and in summer compared with the estimations of
the model for the same period. Source: [4], [14], [33] y [36]
and own calculations.
Figure 5. Deviation of the outputs of the model versus the
winter and summer peaks of demand observed in the pe-
riod 2000-2010. Source: Own calculations.
–1.0%, respectively ( Figure 5). The trend of th e summer
deficit is reversed to that observed with the annual peak
of demand. The reasoning may come from the growing
use of cooling equipment in the summer, which does not
seem to have yet reached the saturation point (for in-
stance, this tendency has been po inted out in [26] fo r The
Although, in general, the outputs are satisfactory, the
goodness of the results will depend on climatic patterns
in the base year (2009).
Other factors to take into account for future demand
forecasts are, for instance, effective efficiency measures,
extension of RTP or massive in troduction of TOU tariffs
in the residential sector. Future improvements on these
matters may be included in the model.
The introduction of some corrections in the model
must be considered in order to estimate correctly the
summer peak of demand. The most desirable would be
the introduction of patterns of consumption by end use
because of the rising electricity demand for cooling. That
action would correct the natural tendency of the model in
this case, the underestimation, but requires further study.
This trend in the summer peak could be reversed or sta-
bilized by the introduction of appropriate measures
and/or patterns of energy efficiency.
Figure 6 shows the results of peaks of electricity de-
mand and those included in [4] and [36], all of them re-
lated to the scenario of annual electricity demand used in
Copyright © 2011 SciRes. EPE
Figure 6. Outputs of the model compared with in [4] and
[36] forecasts related to the annual demand of electricity.
the forecasting. The tendencies observed in Figure 5 for
the annual (winter) peak and the summer peak, are in-
cluded in Figure 6, correcting the results of the model:
along the period analyzed, a progressive decline of 4%
for the winter peak and a progressive increase of 3% at
the peak of summer have been considered. By 2020, 58.1
GW and 53.0 GW are expected, respectively, as winter
and summer peaks in a BAU scenario. If the observed
tendencies continue, former values can go down to 55.5
GW in winter and go up to 54.7 GW in summer.
Another application of the methodology is the fore-
casting of minimum demand of mainland Spain’s elec-
tricity system. That would allow plann ing in advance the
need for base power required for continuous operation
most of the 8760 hours of a year. Results have been
compared with historical records of minimum electricity
demand throughout the period 2003-2010 (less data are
available for the valley of demand). For the decade
2011-2020 forecasts of minima of demand have not been
The minima of electricity demand are also related to
working patterns and temperature patterns, and also have
a strong dependence on the economic cycle (the mini-
mum of electricity demand in 2009, 17.9 GW, represents
a decline of 6.1% over that in 2008, something con-
firmed by the model’s outputs; this decline is due, almost
totally, to the reduction in industrial demand). By the
year 2020, the annual valley of demand will rise 5.5 GW
(compared to minimum of demand of the base year), up
to 23.4 GW. The relocation of consumption from the
efficiency trend at winter peak may involve an increase
of the valley; this aspect requires further study and the
redistribution of the shape of the load curve (the term
” in (1).
The comparison of annual minima of electricity de-
mand with historical data for the period 2003-2010
shows interesting results (Figure 7). The outputs fit
finely with those years close to the base year. For the rest
of the period (from 2003 to 2007), a downward trend of
overestimation is observed, which can be ascribed to an
improvement in consumption patterns of industries (per-
haps they have moved a part of their consumption to
off-peak hours, more economical because they are sup-
plied under TOU tariffs).
Another reason for that trend may also lie in the grad-
ual disappearance of the regulated tariffs for industrial
consumers as it has previously been said for the trend of
the winter peak. In any case, both explanations would be
directly associated with the tendency which has previ-
ously been called as “efficiency” when the peaks of de-
mand were desc ri bed.
A study referred in [38], looking at 87 large Spanish
customers supplied with TOU tariffs, revealed that re-
ducing production to make savings on their electricity
bill was not profitable for industrial customers. The
trends observed in the winter peak and in the annual
minimum of demand may contradict that study, and con-
firm for the Spanish electricity system the ideas ex-
pressed in [39]: customers will respond to higher prices
of electricity by purchasing more efficient appliances and
taking other efficiency measures.
6. Aplication to the Spanish Energy Sector in
the Decade 2011-2020
Along the period 2011-2020, the Spanish energy sector
will face an ambitious renewable energy deployment
[20]. As objectives for 2020, 38.0 GW of wind power,
Figure 7. Minima of electricity demand in the decade
2000-2010; results for the period 2000-2020; and deviation
for the period 2003-2010. Source: REE (
cion/simel.asp) and own calculations.
Copyright © 2011 SciRes. EPE
more than 8.3 GW of photovoltaic and 5.0 GW of con-
centrated solar-thermal may be installed. In addition,
close to 3.5 GW of new power in combined heat and
power (CHP) are expected by 2020.
Renewable energies offer a limited contribution to se-
curity of supply from the point of view that it is impossi-
ble to guarantee they will work at a certain power a cer-
tain date. Therefore, the fo recasting in electricity systems
with a high penetration of renewable energy must in-
clude both, electricity demand and renewable generation
in order to meet properly demand and generation
It must be noted that renewable energy plants (with the
very minor exception of solar thermal units with th ermal
energy storage) are unable to contribute to stabilization
and regulation of the electric systems, because they can
not be fully governed. The back-up power for renewables
is going to be a critical point for the security of supply
and the quality of supply. From the point of view of
finding a suitable type of plant for backing-up the de-
ployment of main renewable energies (wind and solar),
GFCC seem to be the best choice, for a number of rea-
sons: 1) technical flexibility to increase power at a fast
speed if they are already working over the technical
minimum; 2) short construction times; 3) actually the
lowest specific investment cost. On the contrary, there
are some negative effects that must be cited: 1) short
number of working hours, which is a drawback for re-
covering the investment; 2) dependence on the gas price,
which is the main component of the variable cost; and 3)
the high number of start-up s (up to 100 or more in a year)
that means a lower efficiency for GFCC, greater need of
maintenance and reduction of the life time of many
By 2020, in a BAU scenario of generation of renew-
ables and CHP at peak of demand, may be needed 8-9
GW of additional power (related to the base year).
The growing expectations for the minimum of elec-
tricity demand (5.5 GW) may require installation of new
base load power plants. In a BAU scenario at valley of
demand, the new power of CHP and manageable renew-
ables (for instance, biomass) would only cover a third of
that increase. GFCC are also suitable for base load op-
eration because of the good efficiency rate, but as a
negative aspect, they have high and volatile operating
costs related with natural gas price. Coal and nuclear
power plants may also be economically competitive at
base load working. They are less able to regulate load
and they have longer periods of construction, but as an
advantage, they usually have lower operating costs. No-
wadays in Spain, coal plants work as backup power for
renewables with a low load factor (34% in 2009) and lots
of starts-ups and stops per year. Nuclear power plants
operate continuously (except for refuelling and mainte-
nance) as the base of the system, and operate without
power reduction even if the demand is too low for the
total power available of renewables plus nuclear.
Existing coal power plants in Spain may assume a part
of the minimum demand growth, as base load facilities,
without investing in new infrastructure (at least in the
medium-term), if GFCC are used as backup power of
renewables because of their better technical fit for that
Taking into account the low load factor of coal plants
and combined cycles (in 2009), the expected growth of
the minimum of electricity demand until the year 2020
and the planned deployment of renewables, it does not
seem necessary investing on base load power plants,
such as nuclear power plants. However, if in the decade
2021-2030 the forecasts of growth for the minimum of
electricity demand continue, the installation of this type
of unit, with high power per group (up to 1.5 GW each
nuclear plant) might be considered in order to reduce
operating costs of base load generation in the Spanish
electricity system.
Environmental and social aspects may also be consid-
ered in the choice of generation technologies, for base
load generation, to meet peaks of demand or as a backup
power for renewable energy.
7. Summary and Future Work
This paper presents a methodology for estimating peaks
of electricity demand. It has been validated by comparing
its outputs with historical records of demand (peak and
valley) in the period 2000-2010 and with available fore-
casts since 2011. The results of both comparisons have
been satisfactory. The maximum discrepancy in the
forecast, compared to historical records, is 4.2% (under-
estimation) in the winter peak of the year 2005. This dif-
ference may be attributable to more extreme temperature
patterns in that year than those observed in the base pe-
riod. It has also been observed a possible tendency to the
reduction of the winter maximum of demand driven by
the introduction of efficiency patterns. The observed
tendency to overestimate the valley of demand in the
period 2003-20 10, which is redu ced for those years clos e
to the base year, may confirm the relocation of demand
from peak hour to off-peak hours. These trends may lie
on two reasons: the first one, an efficient use of the TOU
tariffs by high voltage consumers; the second one, the
end of the liberalization process of electricity in Spain
for high voltage consumers, which concludes in 2008.
With regard to the summer peak of demand, the maxi-
mum discrepancy predicted by the model stands at 1.7%
in the period 2010. The explanation for this discrepancy
in the forecast may also lie in the temperature patterns. In
Copyright © 2011 SciRes. EPE
any case, there has also been observed a trend toward
increase in the summer peak of demand compared to the
values estimated, which seems related to a growing de-
mand for cool ing.
By 2020, 58.1 GW and 53.0 GW are expected, re-
specttively, as winter and summer peaks in a busi-
ness-as-usual scenario. If the observed tendencies con-
tinue, former values can go down to 55.5 GW in winter
and go up to 54.7 GW in summer. If both trends are con-
firmed and the trend in the summer peak is not corrected,
by the year 2022, the summer peak may be close to the
winter one and both seasons will beco me critical periods
for electricity demand. The valley of annual demand may
increase 5.5 GW (related to the base year) up to 23.4 GW
in a BAU scenario.
The trends highlighted in this paper, deserve a further
study to be addressed in future works.
Detailed predictions of this analysis can be very useful
to identify the types of power plants needed to have an
optimum structure in the electricity industry. As a prac-
tical application of this study on mainland Spain’s elec-
tricity system, it can be said that in th e decad e 2011-2020
no new power plants would be needed for base load op-
eration because the existing coal power plants may as-
sume that role, although this policy would not comply
with the objective of reducing CO2 emissions, but it
would be the cheapest one. It goes without saying that
life extension of nuclear power plants has already been
assumed as part of the principles of the electricity Indus-
try in some countries; others, as Germany, are discussing
the shutdown of nuclear power by 2022, as a cones-
quence of the nuclear incident at Fukusima, after the
tsunami on the 5th of March of 2011. However, due to
the limited contribu tion of renewables to security of sup -
ply, 8 - 9 GW of additional power will be needed in
Spain to meet peak demand with a safety margin similar
to that registered in the base year. Environmental and
social aspects may also be considered in the choice of
generation technologies, for base load generation, to
meet peaks of demand or as a backup power for renew-
able energy.
8. Acknowledgements
Guidance of Prof. J.M. Martínez-Val, from ETSII-UPM,
is highly recognized. The help from CNE’s staff is also
gratefully acknowledged.
9. References
[1] C. García-Ascanio and C. Maté, “Electric Power Demand
Forecasting Using Interval Time Series: A Comparison
between VAR and iMLP,” Energy Policy, Vol. 38, No. 2,
2010, pp. 715-725.
[2] W. Soontornrangson, D. G. Evans, R. J. Fullerc and D. F.
Stewart, “Scenario Planning for Electricity Supply,” En-
ergy Policy, Vol. 31, No. 15, 2003, pp. 1647-1659.
[3] M. Mehra and A. Bharadwaj, “Demand Forecasting for
Electricity. Tata Energy Research Institute, New Delhi,
India,” Proceedings of the National Conference on Regu-
lation in Infrastructure Services: Progress and Way For-
ward, New Delhi, 14-15 November 2000.
[4] Spanish Industry Office, “Planning of electricity and gas
2008-2016. Development of transmission networks”, 2008.
[5] Spanish Industry Office, “Royal Decree 325/2008 and
Royal Decree 326/2008, Februa ry 29th, 2008”. Spain’s Of-
ficial Journal, BOE No. 55, March 4th, 2008, pp. 13079-
[6] Spanish Industry Office, “Royal Decree 661/2007, May
25th, 2007”. Spain’s Official Journal, BOE No. 126, May,
26th, 2007, pp. 22846-22886.
[7] Spanish Industry Office, “Order ITC/2794/2007, Sep-
tember 27th, 2007”. Spain’s Official Journal, BOE No.
234, September 29th, 2007, pp. 39690-39698.
[8] F. Castro-Rodríguez, P. L. Marín and G. Siotis, “Capacity
Choices in Liberalized Electricity Markets,” Energy Pol-
icy, Vol. 37, No. 7, 2009, pp. 2574-2581.
[9] L. de Vries and P. Heijnen, “The Impact of Electricity
Market Design upon Investment under Uncertainty: The
Effectiveness of Capacity Mechanisms,” Utilities Policy,
Vol. 16, No 3, 2008, 215-227.
[10] Spanish Industry Office, “Order ITC/3801/ 2008, De-
cember 26th, 2008”. Spain’s Official Journal, BOE No.
315, December 31st, 2008, pp. 52672-52685.
[11] Spanish Industry Office, “Order ITC/1723/2009, June
26th, 2009”. Spain’s Official Journal, BOE No. 156, June
29th, 2009, pp. 53812-53821.
[12] Spanish Industry Office, “Royal Decree 1634/2006, De-
cember 29th, 2006”. Spain’s Official Journal, BOE No.
312, December 30th, 2006, pp. 46656-46679.
[13] Spanish Industry Office, “Order ITC/2370/2007, July
26th, 2009”. Spain’s Official Journal, BOE No. 185, Au-
gust 3rd, 2007, pp. 33442-33451.
Copyright © 2011 SciRes. EPE
Copyright © 2011 SciRes. EPE
[14] Spanish Electricity System Operator (Red Eléctrica de
España, S.A.), “Spanish electricity system 09”, 2010.
[15] N. Hopper, C. Goldman, R. Bharvirkar and D. Engel,
“The Summer of 2006: A Milestone in the Ongoing
Maturation of Demand Response,” The Electricity Jour-
nal, Vol. 20, No. 5, 2007, pp. 62-75.
[16] H.-P. Chao, “Price-Responsive Demand Management for
a Smart Grid World,” The Electricity Journal, Vol. 23,
No. 1, 2010, pp. 7-20. doi:10.1016/j.tej.2009.12.007
[17] H. Fraser, “The Importance of an Active Demand Side in
the Electricity Industry,” The Electricit y Journal, Vol. 14,
No. 9, 2001, pp. 52-73.
[18] A. Sheffrin, H. Yoshimura, D. LaPlante and B. Neenan,
“Harnessing the Power of Demand,” The Electricity
Journal, Vol. 21, No. 2, 2008, pp. 39-50.
[19] C. Rochlin, “The Alchemy of Demand Response: Turn-
ing Demand into Supply,” The Electricity Journal, Vol.
22, No. 9, 2009, pp. 10-25.
[20] Spanish Institute for Energy Diversification and Saving
(IDAE) “National Action Plan for Renewable Energy in
Spain (PANER) 2011-2020”, 2010.
[21] C. von Hirschhausen and M. Andres, “Long-Term Elec-
tricity Demand in China—From Quantitative to Quali-
tative Growth?” Energy Policy, Vol. 28, No. 4, 2000, pp.
231-241. doi:10.1016/S0301-4215(00)00014-8
[22] R. Mamlook, O. Badran and E. Abdulhadi, “A Fuzzy In-
Ference Model for Short-Term Load Forecasting,” En-
ergy Policy, Vol. 37, No. 4, 2009, pp. 1239-1248.
[23] J.-S. Huang and J.-H. Wu, “A Portfolio Risk Analysis on
Electricity Supply Planning,” Energy Policy, Vol. 36, No.
2, 2008, pp. 627-641. doi:10.1016/j.enpol.2007.10.004
[24] Spanish Electricity System Operator (Red Eléctrica de
España, S.A.), “INDEL Project. Atlas of Spanish elec-
tricity demand, 1998.
[25] D. Keles, D. Möst and W. Fichtner, “The Development of
the German Energy Market until 2030-A Critical Survey
of Selected Scenarios,” Energy Policy, Vol. 39, No. 2,
2011, pp. 812-825. doi:10.1016/j.enpol.2010.10.055
[26] M. Hekkenberg, R. M. J. Benders, H. C. Moll and A. J.
M. S. Uiterkamp, “Indications for a Changing Electricity
Demand Pattern: The Temperature Dependence of Elec-
tricity Demand in the Netherlands,” Energy Policy, Vol.
37, No. 4, 2009, pp. 1542-1551.
[27] D. C. Mountain and E. L. Lawson, “Some Initial Evi-
dence of Canadian Responsiveness to Time-of-Use Elec-
tricity Rates: Detailed Daily and Monthly Analysis,” Re-
source and Energy Economics, Vol. 17, No. 2, 1995, pp.
189-212. doi:10.1016/0928-7655(94)00025-F
[28] Spanish Industry Office, “Royal Decree 1164/2001, Oc-
tober 26th, 2001”. Spain’s Official Journal, BOE No. 268,
November 8th, 2001, pp. 40618-40629.
[29] E. González-Romera, M. A. Jaramillo-Maron and D.
Carmona-Fernández, “Forecasting of the Electric Energy
Demand Trend and Monthly Fluctuation with Neural Net-
Work,” Computers and Industrial Engineering, Vol. 52,
No. 3, 2007, pp. 336-343. doi:10.1016/j.cie.2006.12.010
[30] Spanish Gas System Operator (ENAGAS), “The Spanish
gas sytem. 2009 report”, 2010.
[31] National Meteorological Agency, “2009, the Third Warm-
est Year”, 2010.
[32] Spanish Association of Electricity Industries (UNESA),
“Electricity Report—Activity Report—Statistical Report
[33] Spanish Electricity System Operator (Red Eléctrica de
España, S.A.), “Spanish Electricity System. Advance of
2010 Report”, 2010.
[34] P. Narayan and S. Popp, “Can the Electricity Market be
Characterised by Asymmetric Behaviour?” Energy Policy,
Vol. 37, No. 11, 2009, pp. 4364-4372.
[35] Spanish Secretary of State for Energy, Industrial Devel-
opment and Small and Medium Enterprises, “Resolution,
March 17th, 2004”. Spain’s Official Journal, BOE No. 92,
April 16th, 2004, pp. 15890-15916.
[36] Spanish National Energy Comission, “Informe Marco
Sobre la Demanda de Energía Eléctrica y Gas Natural, y su
Cobertura”, years 2005, 2009 and 2011.
[37] Spanish General Directorate of Energy Policy and Mines,
“Resolution, December 19th, 2008”. Spain’s Official
Journal, BOE No. 314, December 30th, 2009, pp. 52435-
[38] J. Torriti, M. G. Hassan and M. Leach, “Demand Re-
sponse Experience in Europe: Policies, Programmes and
Implementation,” Energy, Vol. 35, No. 4, 2009, pp.
1575-1583. doi:10.1016/
[39] K. Spees and L. B. Lave, “Demand Response and Elec-
tricity Market Efficiency,” The Electricity Journal, Vol.
20, No. 3, 2007, pp. 69-85. doi:10.1016/j.tej.2007.01.006