Sioshansi F.P. Smart Grid: Integrating Renewable, Distributed & Efficient Energy
Подождите немного. Документ загружается.
Figure 19.2
Evolution of the number of EVs from 2010 until 2025.
Source: Batut [2]
The question is whether the introduction of wind power and
the new nuclear plants will increase the supply fast enough to
cope with the increasing numbers of EVs. At present even
though there is surplus capacity in France and in Europe in
general [1, p. 12], occasional periods of stress occur, resulting
in price spikes. For example, on October 19, 2009, the market
price was fixed at the technical ceiling (3000 €) over a period
of four hours.
4
On that occasion there was surplus capacity in
neighboring countries, and so the high prices attracted more
sellers the following day, reducing the pressure on the system.
But the French grid operator [1, p. 14] has warned that after
2015 the existing surplus capacity in Europe may well be
squeezed between the resumption of demand (which had
dropped in Q4 2008 due to the financial crisis) and declining
generation capacity.
911
4
The results of the inquiry held by the French electricity regulator, the
CRE, were made public on November 20, 2009, and are available on its
website: www.cre fr/en/.
The objective of this chapter is to evaluate the combined
impact that the introduction of EVs and the changes in
generation fleet would have on day-ahead electricity prices
over the next 10–15 years. We are interested in statistically
predicting their impact by simulating the offers to buy and
sell electricity on the day-ahead market. By “statistical” we
mean that the histograms of the simulated prices should be
realistic; we are not attempting to predict prices on particular
dates. It is important to be able to reproduce realistic peaks
because they provide the price signal to encourage utilities to
construct peaking plants.
This chapter focuses on weekdays because electricity
consumption is much lower on weekends, and consequently
the effect of the changes will first be felt on weekdays.
Two scenarios for recharging the electric vehicles [4] are
considered:
1. recharging the batteries at night from 9 PM onwards,
2. recharging the batteries at night and drawing power from
the EVs during the day when the demand is high.
The first scenario called grid-to-vehicle (G2V) would lead to
a substantial increase in the demand at night, but would not
affect day time prices or the evening peak hour from 7 PM to 8
PM. The second scenario illustrates the vehicle-to-grid (V2G)
option where smart grids use the power in the batteries to
supply the grid during the daytime peak period.
912
After the liberalization of energy markets in 2000, the French
government gave the grid operator, RTE, the responsibility of
evaluating the risk of not having enough electricity over the
next 10–15 years. The first step in addressing this risk
consists of identifying the trends in the demands and in the
evolution of the generation fleet. Based on the latest studies
[1] and [5] and the government's environmental proposals,
Iezhova [6] identified three possible evolutions for the nuclear
fleet (High, Reference, and Low), two for wind power (High
and Low), and two for thermal plants (High and Low), that is,
a total of 12 scenarios (see Table 19.1). Figure 19.3 shows the
three scenarios for the evolution of the nuclear generation
fleet (top), the two for thermal power (middle), and the two
for wind power (lower), over the next 15 years. In this chapter
we consider the two most extreme cases denoted by HHH and
LLL. These correspond to the high option for nuclear plants,
wind power, and thermal plants, and to the low option for all
three.
Table 19.1
Summary of Twelve Scenarios: Ref = Reference Scenario, H = High, and L = Low
Scenario 1 2 3 4 5 6 7 8 9 10 11 12
Nuclear Ref H L Ref H L Ref H L Ref H L
Wind H L H L H L H L H L H L
Thermal H H H H H H L L L L L L
Source: Iezhova [6].
913
Figure 19.3
The three scenarios for the revolution of the nuclear generation fleet (top), the
two for thermal power (middle), and the two for wind power (lower). Note the
drop in the low nuclear scenario in 2021 is caused by the decommissioning of
several nuclear power plants.
Source: Iezhova [6]
The next sections review existing methods for predicting the
effect of changes, such as the introduction of EVs and of wind
power, on the electricity system as a whole and on day-ahead
prices for electricity.
Effect of EVs on the Power System as a Whole
Kempton and Tomic [7] were among the first to study the
effect of V2G power transfers. They concluded that it would
not be a viable source of baseload power but that it could be
economically profitable in three markets: spinning reserves,
regulation of the voltage, and peak production. Denholm and
Short [8] and [9] and Tomic and Kempton [10] further
explored the possibilities of the V2G concept. Scott et al. [11]
computed the life-cycle cost of a plug-in hybrid electric
vehicle and compared it to that of a fuel-efficient
conventional vehicle, for two cases: (1) if the host utility and
the grid only have to make minor adjustments to
accommodate the plug-in hybrid EVs and (2) if additional
generation capacity is required. The results of their
914
comparison are favorable in the first case. In addition to
reducing greenhouse gas emissions and dependency on
foreign oil, it improves the economics of the electricity
industry, which has spare generation and transmission
capacity at night at present. They note that unless additional
generation is built before the surplus generation capacity is
fully used, the system will experience problems providing the
required power in the future.
Rousselle [12] carried out a detailed study of the impact of
EVs on the electric system in France sponsored by the French
grid operator, RTE. She developed a model for simulating the
recharge times and the state of charge in the battery, in order
to estimate the total load curve. After analyzing the impact of
different load curves on the electric balancing system, she
proposed several solutions to lower these impacts. The total
load curves that she found were reported by Batut [2].
Rousselle did not study the impact of EVs on electricity
prices.
Predicting the Impact of Changes in Supply and
Demand on Electricity Prices
The approaches that are used for predicting the impact of
changes in supply and demand for power on day-ahead prices
depend on (1) the size and complexity of the system, (2) the
way in which prices are determined, and (3) the information
that is available to the public.
Chapter 18 by Hindsberger et al. presents a case study on
New Zealand, which has a small, self-contained electrical
system, whereas in this chapter we consider a much larger
915
system with interconnections to six neighboring countries. In
New Zealand most electricity is generated in the South Island
whereas the demand is concentrated in the North Island, so
transmission constraints arise. Electricity prices are
determined by co-optimizing energy and instantaneous
reserves every five minutes after taking into account the
different transmission costs and other technical constraints.
Hindsberger et al. used the Plexos modeling tool that mimics
the algorithm used in New Zealand. In their case only 170
nodal points were required to model the electrical system.
Because of the size of the system in France
5
and its
interconnections with other countries, it would not be possible
to implement an overall optimization of this type. Other
approaches have to be used.
5
The generation fleet as of May 2011 consisted of 58 nuclear plants, 20
coal-fired plants, 8 gas-fired plants, 18 peaking plants, 39 lake
hydropower stations, and 14 run-of-river generators. Source:
http://clients rtefrance.com/lang/an/clients_distributeurs/vie/prod/
parc_reference.jsp.
Studies on EVs rarely consider their impact on electricity
prices. One exception is Hadley and Tsvetkova [13] who
determined the marginal generation type in 12 regions in the
United States, and hence the impact on wholesale prices. That
is, they were effectively using the full merit order. Similarly,
most studies on the impact of wind power on day-ahead
prices [14], [15] and [16] assume that:
• the demand is inelastic
• the supply curve is equal to the full merit order.
Saenz de Miera et al. [14] evaluated the impact on day-ahead
prices in Spain, whereas Weigt [16] and Sensfuss et al. [15]
916
both studied the impact in Germany. Table 19.2 summarizes
the average drop in the electric price found by these three
studies over different periods from 2005 until mid-2008.
Despite some differences, there is an overall agreement
between the values obtained.
Table 19.2
Reduction in the Day-Ahead Electricity Price due to the Introduction of Wind
Power
Authors Country 2005 2006 2007 2008
Saenz de Miera et
al. [14]
Spain
7 €/
MWh
5 €/
MWh
12 €/MWh (1st
half)
Weigt [16] Germany
10 €/
MWh
17 €/MWh
19 €/MWh (1st
half)
Sensfuss et al. [15] Germany
7.8 €/
MWh
Source: Saenz de Miera et al. [14], Weigt [16] and Sensfuss et al. [15].
Pfluger et al. [17] studied the impact of solar power
6
from
Africa on day-ahead prices in Italy using the same
agent-based simulation approach as Sensfuss et al. [15]. One
particularly interesting part of their study is a comparison
between simulated prices and the observed prices. Their
simulation model delivers realistic predictions except during
peak hours and in the early morning. As the authors noted,
“the price in peak hours turns out to be problematic as the
very high prices in these hours cannot be explained
fundamentally by variable costs used in PowerAce.”
7
In fact,
the simulated prices during peak periods on weekdays rarely
exceeded 80 €/MWh, whereas the actual peak prices ranged
between 110 €/MWh and 130 €/MWh.
6
The idea was to set up solar panels in the Sahara and transport the power
generated by direct current cable to the southern tip of Italy.
917
7
PowerAce is the name of their agent-based simulation model.
Two reasons can be advanced to explain the discrepancy: first
the authors averaged sets of 50 simulations, which effectively
smoothe out any peaks and troughs. More importantly, as
Sensfuss et al. [15] noted, most electricity is traded via
bilateral contracts (i.e., OTC) rather than through organized
markets. For example, in France about 10% of the power
consumed is traded via the day-ahead market. In fact,
producers and major consumers do not wait until the day
before delivery to start the process of buying and selling
electricity. They spread their transactions over time by using
futures markets and financial derivatives. For example,
buyers often purchase part of their requirements months
ahead of time through the futures and forward markets, and
then as the delivery date approaches, they make adjustments
by purchasing more or selling off unwanted power through
the day-ahead market. Last-minute adjustments are made
through the intra-day market. So the key question is whether
or not the aggregate offers to sell power on the day-ahead
market are a scaled-down version of the full merit order.
There are two ways of tackling this question: one is by
computing the merit order and comparing it to the aggregate
offers. This requires a detailed knowledge of the costs of all
the generators and being able to evaluate the opportunity
costs for hydropower stations and the nuclear plants,
8
which
is difficult to do. An alternative way of testing this hypothesis
is by studying the evolution of the aggregate curves over short
periods (e.g., a few consecutive days) when the same power
plants should be operational and hence when the merit order
should be constant. If the offers reflect the full merit order,
then they would remain stable over these periods [20].
918
Armstrong et al. (2011b) showed that the offers to sell power
at a given time of day change radically from one day to
another, particularly when the system is stretched. This means
that the curves are not a scaled-down version of the full merit
order. This is why a method for simulating the offers to buy
and sell power on the electricity exchange, and hence the
prices, was developed.
8
As the dates for routine maintenance and for loading fuel rods are fixed
well in advanced, nuclear plants have to decide how much to produce at
any given time in much the same way as hydro plants.
The primary aim of this chapter is to present a new method
for simulating the aggregate offer to buy and sell power on
the day-ahead market without these restrictive assumptions.
The proposed method uses three types of data that are
available hourly 365 days per year: (1) the aggregated offers
to buy and sell electricity in France, (2) the total consumption,
and (3) the actual production for each class of power plant. As
these data are available at exactly the same times, whatever
factors (weather, etc.) affected one set also influenced the
others. While part of the data has been available since the
liberalization of the electricity market in 2001, some of it has
only become public more recently.
As the information on the production per class only became
available to the public from November 2006 onward and as
this study started early in 2010, there were only two full years
that could be used as the reference: October 2007 to
September 2008, or October 2008 to September 2009. During
the fourth quarter of 2008, the electricity consumption in
France dropped sharply because of the financial crisis. So that
year is atypical. This is why we used the year October 2007 to
919
September 2008 as the reference year. In the future it will be
possible to base the study on several reference years, which
will be more representative.
This chapter is structured as follows. The next section
presents our method for simulating the aggregate offers to
buy and sell electricity on the day-ahead market. It takes
account of the evolution in the structure of the generation
fleet on the supply side, and two factors on the demand side:
the expected increase in consumption and the quantities
required by the EVs. The results of the simulations are
presented in the following section. Our conclusions and
perspectives for future work are given in the last section.
Approach
Having demonstrated that the aggregate offers to sell
electricity are not a scaled-down version of the full merit
order, and that the demand on the bourse is not inelastic, we
need a method for simulating realistic curves of the aggregate
offers to buy and sell electricity in order to simulate
day-ahead prices at future dates. In fact we need matched
pairs of curves corresponding to the same market and weather
conditions. The key point is to link the overall supply and
demand to the aggregated offers in the auction market. Figure
19.4 summarizes our approach.
920