Reasons for the drop of Swedish
wholesale electricity prices 2010-15
Lion Hirth
Project for Svensk Energi | Final report | [email protected]
Executive summary
Sweden: price drivers
• Swedish spot prices declined by
65% during 2010-15
• Declining demand and growth of
renewable generation
contributed most to the drop
• Large hydro inflows contributed
also
• Increased export helped strongly
mitigating the drop
Neon analysis.
• Power prices in the Nordic – a
hydro system with a lot of low
variable cost generation – react
particularly sensitive to volume
shocks
RES growth, reduced demand, and the wet
year 2015 decreased Swedish prices most.
Lion Hirth
2
Neon: relevant project references
Neon is a Berlin-based boutique
consulting firm for energy
economics. We combine
expertise on economic theory
with advanced modelling
capabilities and extensive
industry experience. Neon
specializes in five areas:
SystemSystem-friendly wind and solar power (IEA). Model-based study for the International Energy
Agency, Paris. Neon assessed the market and system benefits of low-wind speed wind turbines
and east- and west-oriented PV based on its power market model EMMA. 2014-16. The study is
published in Energy Economics. More
Integration costs (Agora Energiewende). Literature-based study for Agora Energiewende, Berlin.
Neon advised Agora and helped implementing workshops in Berlin and Paris. 2015. The report
has been published by Agora. More
Whole system costs (DECC). Neon advised the UK Department of Energy and Climate Change in a
project on whole system costs of wind and solar power. 2015.
Open Power System Data (BMWi). Construction of an open platform for European power system
data for the German Ministry of Economic Affairs an Energy. Neon coordinates a team of three
research institutes. 2015-17. More
1.
Market value of wind
power
2.
System costs
3.
Balancing power
Electricity market design (IEA(IEA-RETD). Project on power market design under very high shares of
variable renewables. Wholesale, balancing, and retail markets are covered in different markets,
ranging from liberalized to vertically integrated. Neon is conducting the project in cooperation
with FTI CL Energy. 2015-16. More
4.
Market design
Model development (Vattenfall). Neon supported Vattenfall in model development. 2015.
5.
Power market modeling
Wind market value in the Nordic region (Energiforsk
(Energiforsk)
Energiforsk). Model-based assessment of the market
value of wind power in the hydro-dominated power system of the Nordic region. Neon design
the study, developed the model, and wrote the report. 2016.
Power market trainings. Neon trained staff at IRENA, ERRA, Vattenfall, JRC, UFZ, Clean Air Task
Force, IG Windkraft in topics such as power markets, energy economics, and electricity policy.
More
Lion Hirth
4
Context
Motivation: the electricity price plunge
Swedish base price (yearly)
SWE vs. GER prices (monthly)
Neon analysis.
Neon analysis.
Swedish wholesale electricity prices
declined by 65% from 2010 to 2015 (dayahead base price, inflation-adjusted).
German spot prices declined, but not as
much as prices in Sweden.
Lion Hirth
6
The price structure has changed as well
German spot price structure
Change 20022002-15
The diurnal structure of German day-ahead spot price during summer months 2002-15.
Neon analysis based on data from TSOs and power exchanges.
The price structure of German prices
changed dramatically with the rise of solar.
The change of price structure between 2002 and 2015.
Neon analysis based on data from TSOs and power exchanges.
Sunny hours became relatively much
cheaper, and night hours more expensive.
Lion Hirth
7
Three drivers of falling prices: thermal system
Demand
Demand
Demand
Variable cost
(€/MWh)
Demand
GT
GT
Previous price
OCGT
nat gas
OCGT
Nat gas
GT
Reduced price
Nuclear
CHP
Coal
Nuclear
CHP
Hydro RES
Hydro
RES
2
Reduced demand
Nuclear
CHP
Capacity (MW)
Capacity (MW)
1
Coal
Increased low-variable
cost supply
Lion Hirth
Hydro RES
OCGT
Nat gas
Coal
Capacity (MW)
3
Reduced
variable cost
8
Three drivers of falling prices: hydro system
Demand
Demand
Demand
Variable cost
(€/MWh)
Demand
GT
GT
Previous price
OCGT
nat gas
OCGT
Nat gas
GT
Reduced price
Nuclear
CHP
Coal
Nuclear
CHP
Hydro RES
Hydro
RES
2
Reduced demand
Nuclear
CHP
Annual energy (TWh)
Annual energy (TWh)
1
Coal
Increased low-variable
cost supply
Lion Hirth
Hydro RES
OCGT
Nat gas
Coal
Annual energy (TWh)
3
Reduced
variable cost
9
Potential drivers in detail
3
2
1
Reduced demand
• Declined final demand for
electricity
• Reduced export (ATC)
capacity to other
countries, particularly
outside the Nordic region
Increased low-cost supply
Reduced variable cost
• Additional thermal capacity
(mostly coal-fired plants on
the Continent)
• Declining coal price
• Year-to-year variation of hydro
inflow in the Nordics
• Improved fleet efficiency
(heat rate)
• Additional wind, solar, and
biomass capacity
• Increased natural gas
price
• Declining CO2 price
• Availability of Swedish nuclear
power
• Decommissioning of
conventional plants
• Nuclear phase-out in Germany
Lion Hirth
10
New coal plants in Germany
Plant name
BoA 2
BoA 3
Emsland
Gemeinschaftskraftwerk Irsching
GKM
Rheinhafen-Dampfkraftwerk
Moorburg B
Westfalen
Westfalen
Trianel Kohlekraftwerk Lünen
KW Walsum
Boxberg
Ulrich Hartmann(Irsching)
Kopswerk II
Knapsack Gas II
Trianel Gaskraftwerk
Trianel Gaskraftwerk
Rodundwerk II
GKL
Dow Stade
Rheinkraftwerk Iffezheim
Zellstoff Stendal GmbH
GuD Tiefstack
Emsland
Emsland
Total (incl. < 100MW)
Coal plants: new and under construction.
Fuel
lignite
lignite
gas
gas
coal
coal
coal
coal
coal
coal
coal
lignite
gas
PHS
gas
gas
gas
PHS
gas
gas
hydro
biomass
gas
gas
gas
-
Capacity
1050
1050
887
846
843
842
766
765
765
746
725
640
545
525
430
421
417
295
230
173
146
139
127
116
116
16600
New and retrofitted power plants in Germany, 2008-15.
Lion Hirth
11
Which electricity prices are we interested in?
One can analyze spot or financial markets. On average, they should be identical,
but in the past years they often deviated significantly.
Spot (day(day-ahead) markets
•
How did realized prices develop?
•
How did market fundamentals (supply, demand, costs) change?
Financial (future) markets
•
How did expectations develop?
A spot market analysis is easier to interpret, and data availability is better
(expectations are private information).
Lion Hirth
12
Research question
In short:
Why did the Swedish power price drop?
More precisely:
Which factors contributed by how much to the drop of
the Swedish electricity day-ahead base price between
2010 and 2015?
Lion Hirth
13
The EU ETS carbon price
CO2 emission certificates: an intertemporal market
• Electricity spot prices are (almost) fully explained by instantaneous factors
(residual demand and variable costs)
• Certificate markets work very different (both emission certificates and green
certificates)
• The intertemporal demand-supply balance determines the current price (= the
demand-supply balance over the entire period the ETS is in effect)
• Anticipated (expected) changes in future abatement costs or the stringency of
future emission caps determines the price today
• The intertemporal (aggregated) demand-supply balance is mostly affected by
• Aggregated business-as-usual emissions
• Aggregated cap
• Mitigation costs
Lion Hirth
15
Allowance Price
Price setting in an intertemporal ETS
Courtesy Nicolas Koch.
≈ 2050
Cumulative Emissions
Time
Lion Hirth
16
Why are EU ETS prices so low?
EUA prices are much lower than many expected (or hoped or feared). There are a
range of possible reasons (or a combination of these):
1.
Demand shock:
shock baseline emissions decreased because of the
macroeconomic recession and renewable support policies
2.
Supply shock:
shock more certificates are available than anticipated
3.
Supply shock & expectations:
expectations market participants believe (or speculate)
that the long-term cap will be less stringent than announced
4.
Demand shock & myopia:
myopia market participants do not take the long-term
stringency of the system into account
5.
Demand shock & high riskrisk-adjusted discount rate:
rate current prices are low
because long-term (high) prices are discounted at a high rate (the Hotelling
price path is very steep)
Lion Hirth
17
Methodology
Methodology
1. Replicate prices for the years 2010 and 2015
•
With a fundamental power market model
•
Using the full set of input factors of the respective year (electricity demand,
RES generation, hydro inflow, fuel prices, ...)
•
Model check: can prices be replicated?
2. Quantify impact of individual factors
•
Substitute one individual factor (e.g. electricity demand) from 2010 with
2015 values
•
Leave all other factors (e.g., RES generation, hydro inflow, fuel prices, ...)
unchanged at 2010 values
•
Replicate this procedure for each factor one-by-one
•
Estimate the impact of individual factors on price drop
Lion Hirth
19
Kallabis et al. (2015): German futures
Reasons for the German price drop
16%
11%
10%
Share in
total effect
52%
Neon illustration based on Kallabis et al. (2015).
Lion Hirth
20
Conceptual remarks on the methodology (1)
1. Sum of individual effects does not equal joint effect
• On a non-linear system like power markets, in general the sum of individual effects
does not equal the joint effect.
• Take an extreme example:
• An increase of coal prices rises the electricity price
• An increase of CO2 prices rises the electricity price
• An increase of both prices might not rise the electricity price, if all coal plants are
driven out of the money
2. Alternative benchmarks
• The two following questions are not identical
• “What would be reduction of the electricity price if all parameters are at 2010 levels,
only RES supply is increased to 2015 levels?” (2010 benchmark)
• “What would be the increase of the electricity price of all parameters are at the 2015
level, only RES supply is decreased to 2010 levels?” (2015 benchmark)
Lion Hirth
21
Conceptual remarks on the methodology (2)
3. Individual (“separate”) vs. cumulative (“added”) effect
• We test factors individually, starting always with the 2010 parameter set
• In other words, we test each effect individually, always holding all other effects at
2010 levels
• A different approach would be to add changes on top of each other
4. Cumulative (“added”) effect: order matters
• If effects are added one on the other, order of effects impacts their size
• For example:
• Start with 2010 parameters, decrease demand first, increase RES supply then
• Start with 2010 parameters, increase RES supply first, decrease demand then
• This is the reason we do not follow such an approach
Lion Hirth
22
The Electricity Market Model EMMA
Numerical partial-equilibrium model of the European interconnected power market
Objective: minimize system costs
•
•
•
•
Capital costs
Fuel and CO2 costs
Fixed and variable O&M costs
... of thermal and hydro power
plants, storage, interconnectors
Decision variables
• Hourly dispatch
• Yearly investment
• ... of plants, storage, interco’s
Constraints
•
•
•
•
•
•
Energy balance
Capacity constraints
Volume constraints of storage/hydro
Balancing reserve requirement
CHP generation
(No unit commitment, no load flow)
Resolution
• Temporal: hours
• Spatial: bidding areas (countries)
• Technologies: eleven plant types
Input data
• Wind, solar and load data of the
same year
• Existing plant stack
Equilibrium
• Short-/mid-/long-term model
(= dispatch / capacity expansion /
greenfield)
• Equilibrium (“one year”) rather
than a transition path (“up to
2030”)
Economic assumptions
• Price-inelastic demand
• No market power
• Carbon price
Implementation
• Linear program
• GAMS / cplex
Applications
• Four peer-reviewer articles
• Various consulting projects
• Copenhagen Economics
Open source
Model extensions for this project
• Backcasting (replicating) historical prices requires high precision and more
detailed input parameters than long-term modeling
• This is even more true for hydro-dominated power systems, where small
changes in the yearly energy balance can have dramatic effects on power
prices
Amendments to EMMA (examples)
•
Market power modeling of EDF in French nuclear power dispatch
•
Low availability of Polish plants due to old equipment
•
Improved fleet efficiency over time due to new investments and retirement
of old plants
•
Net imports to model regions from neighboring countries
•
Empirically calibrated average availability
•
Demand elasticity / price spikes
Lion Hirth
24
Data
Crucial parameters 2010 vs. 2015 in the model region
Parameter
2010
2015
Data source
Electricity demand
1723 TWh
1647 TWh
IEA
of which Sweden
147 TWh
134 TWh
Monthly electricity statistic
Wind + solar generation
75 TWh
193 TWh
IEA
of which Sweden
4 TWh
16 TWh
Monthly electricity statistic
Hydroelectricity output
282 TWh
302 TWh
IEA
of which Sweden
66 TWh
76 TWh
Monthly electricity statistic
Net exports of model region
38 TWh
90 TWh
ENTSO-E
-3 TWh
18 TWh
Statistical factsheet
Net demand (demand minus wind, solar,
hydro, net imports) – of which Sweden
1404 TWh
1246 TWh
77 TWh
43 TWh
Coal price
92 $/t
8.4 €/MWh
59 $/t
6.4 €/MWh
IHS McCloskey
Natural gas price
21 €/MWh
22 €/MWh
IMF
CO2 price
16 €/t
6 €/t
EUA price
Own calculation
Northwest Europe Marker Price
German border import price
Conventional capacity includes nuclear and hydro power as well as all fossil fuel generators. Numbers are shown for the entire model region (Sweden, Norway, Germany, France,
Poland, Belgium, The Netherlands). Electricity consumption and wind/solar generation is estimated based on Nov 2015 data, because Dec data are not published yet. All prices are
nominal values (not inflation-adjusted). Dollar-denominated prices were converted into Euro using exchange rate data from the ECB. ATC values are used until the introduction of flowbased market coupling.
Lion Hirth
26
First observations: volume changes
Electricity demand from power plants with positive marginal costs (thermal
plants) declined by 158 TWh (9%).
RES growth had largest effect
•
76 TWh reduced electricity demand (176 TWh if linear trend used for
comparison)
•
118 TWh increased generation from wind and solar
•
16 TWh higher hydroelectricity generation
•
52 TWh increase in net exports from the model region (mostly SWE-FIN
and FRA to ITA/GBR/CHE)
•
Increase in RES explains largest share of volume change.
Lion Hirth
27
First observations: volume changes
Changes to net demand
Neon analysis.
Reduced consumption, expansion of renewables, and more
precipitation decreased net demand 2015 compared to 2010.
Increased net exports compensated partly.
Lion Hirth
28
First observations: price changes
Fuel prices fluctuated widely, but net change 2010-15 is pretty small. The
carbon price declined strongly during the same period.
Coal and nat. gas prices
Some fuel prices declined, while
others remained stable
•
Coal -24%
•
Natural gas + 5%
•
CO2 -63%
•
(Fuel prices in nominal terms
denominated in Euro)
•
It is pretty obvious that a 24%
decline in coal prices can, by
itself, not explain a 65% decline in
electricity prices.
Lion Hirth
Neon analysis.
29
Replicating historical prices
(Step 1)
Factors modeled that vary from year to year
Fuel prices
Electricity generation by renewables
• Coal price
• Wind
• Natural gas price
• Solar
• CO2 price
• Biomass
Investments in thermal capacity
Hydro inflow
• Change in total capacity
Nuclear power
• Improvement of average heat rate
• Phase-out in Germany
Electricity demand
• Fluctuating availability in Sweden
• Consumption
• Net exports at model area border
Lion Hirth
31
The model is able to replicate historical prices well
Sweden
Neon analysis.
Sweden spot prices are replicated fairly well. The modeled price
drop is 33.0, reality was 34.8 €/MWh.
Lion Hirth
32
The model is able to replicate historical prices well
Germany
Norway
Neon analysis.
Neon analysis.
German prices are replicated quite well ...
... as well as Norwegian prices.
Lion Hirth
33
... as well as historical generation pattern
Real world (GER)
Model results (GER)
Neon analysis.
Neon analysis.
Modeled mix. The model overstates coal
generation somewhat, but replicates
structural shifts well.
Observed generation mix in Germany.
Lion Hirth
34
Factor decomposition
(Step 2)
The impact of individual factors: Germany
Germany
Driver
Share in price drop
Renewables growth
54%
Electricity demand
25%
Fuel and CO2 prices
23%
Hydro inflow
10%
Other factors modeled -50% (increasing)
Neon analysis. The share in price drop is the effect of the individual effect
relative to the total drop modeled.
Neon analysis.
The largest factors reducing German prices
were renewables and demand. Other
factors stabilized the price (see below).
If RES grew as they did, but everything else
remained unchanged, the price drop
would be 54% of the actual drop.
Lion Hirth
36
The impact of individual factors: Germany
Germany
sum of individual effects
joint effect
Neon analysis.
The non-linear interaction effect is the difference between the sum of individual effects and the
joint impact if all effects are modeled simultaneously. The interaction is relatively small.
Lion Hirth
37
The impact of individual factors: Sweden
Sweden
Driver
Share in price drop
Renewables growth
61%
Electricity demand
55%
Fuel and CO2 prices
6%
Hydro inflow
33%
Other factors modeled -119% (increasing)
Neon analysis. The share in price drop is the effect of the individual effect
relative to the total drop modeled.
Neon analysis.
RES growth, reduced demand, and the wet
year 2015 decreased Swedish prices most.
Compared to Germany, declining demand
and hydro inflow plays a larger role.
Lion Hirth
38
The impact of individual factors: Sweden
Sweden
Neon analysis.
Swedish price are much more sensitive to changes in fundamentals. This is the nature of a hydro
system where small changes in the yearly energy balance can lead to large shifts of prices. An
additive decomposition leads to a significant residual (non-linear interaction).
Lion Hirth
39
Factor decomposition
(Step 2):
more details
More details than above
• In the follow slides, we decompose the aggregate category “other factors
modeled”
• Nuclear availability in Sweden
• Nuclear phase-put in Germany
• Exports and imports at the border of the model region
Lion Hirth
41
The impact of individual factors: Germany
Germany
Driver
Renewables growth
Electricity demand
Coal/gas invest
CO2 price
Hydro inflow
Coal price
Nuclear availability SWE
Nat. gas price
Imports/Exports
Nuclear phase-out GER
Share in price drop
54%
25%
24%
24%
10%
8%
-1% (increasing)
-8% (increasing)
-31% (increasing)
-41% (increasing)
Neon analysis. The share in price drop is the effect of the individual effect
relative to the total drop modeled.
Neon analysis.
If the only changes was the decline in CO2
prices, the electricity price drop would
have been a quarter of the actual.
Increased exports and the nuclear phaseout stabilized prices most.
Lion Hirth
42
The impact of individual factors: Germany
Germany
Price-depressing effects
Price-stabilizing effects
Neon analysis.
Six factors reduced the electricity price, four increased it. The additive decomposition into
individual effects works quite well: the non-linear interaction term is small.
Lion Hirth
43
The impact of individual factors: Sweden
Sweden
Driver
Renewables growth
Electricity demand
Hydro inflow
Coal/gas invest
CO2 price
Coal price
Nuclear availability SWE
Nat. gas price
Nuclear phase-out GER
Imports/Exports
Share in price drop
61%
55%
33%
14%
13%
0%
-5% (increasing)
-7% (increasing)
-12% (increasing)
-105% (increasing)
Neon analysis. The share in price drop is the effect of the individual effect
relative to the total drop modeled.
Neon analysis.
RES growth, reduced demand, and the wet
year 2015 decreased Swedish prices most.
Lion Hirth
44
The impact of individual factors: Sweden
Sweden
Price-depressing effects
Price-stabilizing effects
Neon analysis.
Swedish price are much more sensitive to changes in fundamentals. This is the nature of a hydro
system where small changes in the yearly energy balance can lead to large shifts of prices. An
additive decomposition leads to a significant residual (non-linear interaction).
Lion Hirth
45
The impact of RES: two perspectives (“plus” & “minus”)
Germany
Sweden
10.1
20.1
11.5
20.1
Neon analysis.
If nothing changed since 2010 except
renewables, prices would have dropped by 10.2
€/MWh. If 2015 would materialize in all aspects,
but renewables remain at 2015 levels, prices
would increase by 11.5 €/MWh.
In Sweden as in Germany, the two perspectives
lead to very similar estimates. This increases the
confidence in robustness of the analysis.
Lion Hirth
46
The impact of Swedish renewables
Sweden
• Above, we reported the impact
of renewables growth in the
entire model region
• This is the joint price impact of
increasing renewables in all
countries on Swedish electricity
prices
• Alternative, we could ask:
• What is the impact of Swedish
RES growth on Swedish prices?
•
If only Swedish RES grew, Swedish prices
would have dropped by 11 €/MWh.
Swedish RES alone represents
60% of the joint RES impact
Lion Hirth
47
Conclusions
Most impacts are transitory – but might take a while
• A cost shock (e.g. a change in fuel or CO2 prices)
can have a lasting impact, if most (or all) pricesetting technologies are affected
• A volume shock (e.g. decrease of demand or
increase of RES supply) decreases the wholesale
electricity price
• This triggers market exit, increasing prices again
• The long-term equilibrium price remains (nearly)
unchanged
•
Crucial question: how long is “long-term”?
•
In power systems with long-living assets and
little demand growth, this can be decades
Lion Hirth
“In the long term, we are all
dead” – John Maynard Keynes
49
Renewables in the Nordic region: long vs. short term
• In a different study, we have reported that the Nordic region, thanks to the
large amount of highly flexible hydro power, is well suited to integrate large
amounts of variable renewables, such as wind power
• The market value of wind power (average spot revenue) drops less in the
Nordics than on the Continent
• This study reports that the average (base) electricity price in Sweden was
much more depressed by renewables expansion than the German price
This leads us to the following interpretation
• The flexibility of hydroelectricity allows easy integration of large-scale wind
power
• The sunk nature of hydro and nuclear assets makes the transition towards
large-scale wind deployment less smooth
Lion Hirth
50
Summary and conclusions
• Wholesale power prices throughout Europe have declined substantially
• Several factors depressed, several increased the price; the former dominated
• The Nordic system, a hydro-dominated system with large volumes of
generation with low variable costs, is much more sensitive: changes in
fundamentals have a much larger price effect
Germany: important price drivers
Sweden: important price drivers
• Downward: RES growth was largest
driver; demand, new investments
and the CO2 price were about half
in size
• Downward: RES growth and demand
decline about the same size; followed
by hydro inflow
• Upward: increase exports (very large
effect)
• Upward: nuclear phase-out,
followed by increased exports
Lion Hirth
51
Reasons for the drop of Swedish
wholesale electricity prices 2010-15
[email protected]