florentina paraschiv, university of st. gallen, ior/cf ... · excursus price forward curves (pfcs)...
TRANSCRIPT
p.1
I n s t i t u t e f o r O p e r a t i o n s R e s e a r c h
a n d C o m p u t a t i o n a l F i n a n c e
Structural model for electricity forward prices
Florentina Paraschiv, University of St. Gallen, ior/cf
with Fred Espen Benth, University of Oslo
Wroclaw University of Technology, 28 April 2016
Outlook
Motivation – p.2
• Structural models for forward electricity prices are highly relevant: major structuralchanges in the market due to the infeed from renewable energy
• Renewable energies infeed reflected in the market expectation – impact on futures(forward) prices?
Structural changes in seasonality
Motivation – p.3
Excursus price forward curves (PFCs)
Motivation – p.4
• We observe futures prices:
– We take as input forward prices of arbitrary maturities, but only prices of alimited set of standard maturities can be observed for electricity!
– Observable prices: (weekly, monthly, quarterly, yearly)
• We derive and forecast one seasonality shape• We optimize to align the forecasted shape to the observed futures prices, resulting
the Price Forward Curve
0.00
10.00
20.00
30.00
40.00
50.00
60.00
70.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
€/M
Wh
Hour of the day
Hourly and daily price pattern
Mo-Fr January
week-end and holidays January
Mo-Fr July
week-end and holidays July
Excursus price forward curves (PFCs)
Motivation – p.5
Excursus price forward curves (PFCs)
Motivation – p.6
Literature review
Motivation – p.7
• Models for forward prices in commodity/energy:
– Specify one model for the spot price and from this derive for forwards: Lucia andSchwartz (2002); Cartea and Figueroa (2005); Benth, Kallsen, andMayer-Brandis (2007);
– Heath-Jarrow-Morton approach – price forward prices directly, by multifactormodels: Roncoroni, Guiotto (2000); Benth and Koekebakker (2008); Kiesel,Schindlmayr, and Boerger (2009);
• Critical view of Koekebakker and Ollmar (2005), Frestad (2008)
– Few common factors cannot explain the substantial amount of variation inforward prices
– Non-Gaussian noise
• Random-field models for forward prices:
– Roncoroni, Guiotto (2000);– Andresen, Koekebakker, and Westgaard (2010);
• Derivation of seasonality shapes and price forward curves for electricity:
– Fleten and Lemming (2003);– Bloechlinger (2008).
Questions to be answered
Motivation – p.8
• We will refer to a panel of daily price forward curves derived over time
• We deseasonalize and aim at a structural model for the stochastic component ofPFCs
– Examine and model the dynamics of risk premia, the distribution of noise(non-Gaussian, stochastic volatility), spatial correlations
– The analysis is spatio-temporal: cross-section analysis with respect to the timedimention and the maturity “space”
Overview of modeling procedure
Modeling assumptions – p.9
Is it realistic?
We validateassumptions
Fine Tuning
Theoretical model:
Spatio-temporal
random field of
forward prices
Empirical analysis: • Fit the model to 2’386 PFCs
• Examine statistics of:
ü Risk premia
ü Distribution of noise
ü Volatility term structure
ü Spatial correlations
Refine the model:• Volatility term structure
• Model coloured noise
• Spatial correlations
Problem statement
Modeling assumptions – p.10
• Previous models model forward prices evolving over time (time-series) along thetime at maturity T: Andresen, Koekebakker, and Westgaard (2010)
• Let Ft(T ) denote the forward price at time t ≥ 0 for delivery of a commodity at timeT ≥ t
• Random field in t:t 7→ Ft(T ), t ≥ 0 (1)
• Random field in both t and T :
(t, T ) 7→ Ft(T ), t ≥ 0, t ≤ T (2)
• Get rid of the second condition: Musiela parametrization x = T − t, x ≥ 0.
Ft(t + x) = Ft(T ), t ≥ 0 (3)
• Let Gt(x) be the forward price for a contract with time to maturity x ≥ 0. Note that:
Gt(x) = Ft(t + x) (4)
Graphical interpretation
Modeling assumptions – p.11
( , ) ( ),
tt T F T t T£
t( ),
tt G X
tt
t
T x T t= -
t
Influence of the “time to maturity”
Modeling assumptions – p.12
1t
2t
Change in the market
expectation ( )te mark
Change due to decreasing
time to maturity( )x
Model formulation: Heath-Jarrow-Morton (HJM)
Modeling assumptions – p.13
• The stochastic process t 7→ Gt(x), t ≥ 0 is the solution to:
dGt(x) = (∂xGt(x) + β(t, x)) dt + dWt(x) (5)
– Space of curves are endowed with a Hilbert space structure H
– ∂x differential operator with respect to time to maturity
– β spatio-temporal random field describing the market price of risk
– W Spatio-temporal random field describing the randomly evolving residuals
• Discrete structure:Gt(x) = ft(x) + st(x) , (6)
– st(x) deterministic seasonality function R2+ ∋ (t, x) 7→ st(x)R
Model formulation (cont)
Modeling assumptions – p.14
We furthermore assume that the deasonalized forward price curve, denoted by ft(x), hasthe dynamics:
dft(x) = (∂xft(x) + θft(x)) dt + dWt(x) , (7)
with θ ∈ R being a constant. With this definition, we note that
dFt(x) = dft(x) + dst(x)
= (∂xft(x) + θft(x)) dt + ∂tst(x) dt + dWt(x)
= (∂xFt(x) + (∂tst(x) − ∂xst(x)) + θ(Ft(x) − st(x))) dt + dWt(x) .
In the natural case, ∂tst(x) = ∂xst(x), and therefore we see that Ft(x) satisfy (5) withβ(t, x) := θft(x).
The market price of risk is proportional to the deseasonalized forward prices.
Model formulation (cont)
Modeling assumptions – p.15
We discretize the dynamics in Eq. (7) by an Euler discretization
dft(x) = (∂xft(x) + θft(x)) dt + dWt(x)
∂xft(x) ≈ft(x + ∆x) − ft(x)
∆x
ft+∆t(x) = (ft(x) +∆t
∆x(ft(x + ∆x) − ft(x)) + θft(x)∆t + ǫt(x) (8)
with x ∈ {x1, . . . , xN } and t = ∆t, . . . , (M − 1)∆t, where ǫt(x) := Wt+∆t(x) − Wt(x).
Zt(x) := ft+∆t(x) − ft(x) −∆t
∆x(ft(x + ∆x) − ft(x)) (9)
which impliesZt(x) = θft(x)∆t + ǫt(x) , (10)
ǫt(x) = σ(x)ǫ̃t(x) (11)
Deseasonalization approach
Modeling assumptions – p.16
• We firstly remove the long-term trend from the hourly electricity prices
• Follow Blöchlinger (2008) for the derivation of the seasonality shape for EPEX powerprices: very „data specific”; removes seasonal effects and autocorrelation!
• In a first step, we identify the seasonal structure during a year with dailyprices: factor-to-year (f2y)
• In the second step, the patterns during a day are analyzed using hourlyprices: factor-to-day (f2d)
• Forecasting models for the factors are derived, such that the resulting shape can bepredicted
• The shape is aligned to the level of futures prices
Factor to year
Modeling assumptions – p.17
f2yd =Sday(d)
∑
kǫyear(d)Sday(k) 1
K(d)
(12)
To explain the f2y, we use a multiple regression model:
f2yd = α0 +
6∑
i=1
biDdi +
12∑
i=1
ciMdi +
3∑
i=1
diCDDdi +
3∑
i=1
eiHDDdi + ε (13)
- f2yd: Factor to year, daily-base-price/yearly-base-price
- Ddi: 6 daily dummy variables (for Mo-Sat)
- Mdi: 12 monthly dummy variables (for Feb-Dec); August will be subdivided in twoparts, due to summer vacation
- CDDdi: Cooling degree days for 3 different German cities – max(T − 18.3◦C, 0)
- HDDdi: Heating degree days for 3 different German cities – max(18.3◦C − T, 0)
where CDDi/HDDi are estimated based on the temperature in Berlin, Hannover andMunich.
Regression model for the temperature
Modeling assumptions – p.18
• For temperature, we propose a forecasting model based on fourier series:
Tt = a0 +
3∑
i=1
b1,i cos(i2π
365Y Tt) +
3∑
i=1
b2,i sin(i2π
365Y Tt) + εt (14)
where Tp is the average daily temperature and YT the observation time within oneyear
• Once the coefficients in the above model are estimated, the temperature can beeasily predicted since the only exogenous factor YT is deterministic!
• Forecasts for CDD and HDD are also straghtforward
Factor to day
Modeling assumptions – p.19
• The f2d, in contrast, indicates the weight of the price of a particular hour comparedto the daily base price.
f2dt =Shour(t)
∑
kǫday(t)Shour(k) 1
24
(15)
• with Shour(t) being the hourly spot price at the hour t.
• We know that there are considerable differences both in the daily profiles ofworkdays, Saturdays and Sundays, but also between daily profiles during winter andsummer season
• We classify the days by weekdays and seasons and choose the classification schemepresented in Table 1
Profile classes for each day
Modeling assumptions – p.20
Table 1: The table indicates the assignment of each day to one out of the twenty profile classes.The daily pattern is held constant for the workdays Monday to Friday within a month, and forSaturday and Sunday, respectively, within three months.
J F M A M J J A S O N D
Week day 1 2 3 4 5 6 7 8 9 10 11 12Sat 13 13 14 14 14 15 15 15 16 16 16 13Sun 17 17 18 18 18 19 19 19 20 20 20 17
Profile classes for each day
Modeling assumptions – p.21
• The regression model for each class is built quite similarly to the one for the yearlyseasonality. For each profile class c = {1, . . . , 20} defined in table 1, a model of thefollowing type is formulated:
f2dt = aco +
23∑
i=1
bci Ht,i + εt for all tǫc. (16)
where Hi = {0, . . . , 23} represents dummy variables for the hours of one day
• The seasonality shape st can be calculated by st = f2yt · f2dt.
• st is the forecast of the relative hourly weights and it is additionallymultiplied by the yearly average prices, in order to align the shape at theprices level
• This yields the seasonality shape st which is finally used to deseasonalizethe electricity prices
Deseasonalization result
Modeling assumptions – p.22
The deseasonalized series is assumed to contain only the stochastic component ofelectricity prices, such as the volatility and randomly occurring jumps and peaks
0 20 40 60 80 100 120 140 160-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF before deseasonalization
0 20 40 60 80 100 120 140 160-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF after deseasonalization
Figure 1: Autocorrelation function before and after deseasonalization
Derivation of HPFC Fleten & Lemming (2003)
Modeling assumptions – p.23
• Let ft be the price of the forward contract with delivery at time t, where time ismeasured in hours, and let F (T1, T2) be the price of forward contract with delivery inthe interval [T1, T2]. Since only bid/ask prices can be observed, we have:
F (T1, T2)bid ≤1
∑T2
t=T1
exp(−rt/a)
T2∑
t=T1
exp(−rt/a)ft ≤ F (T1, T2)ask (17)
where r is the continuously compounded rate for discounting per annum and a is thenumber of hours per year.
min
[
T∑
t=1
(ft − st)2
]
+ λ
T −1∑
t=2
(ft−1 − 2ft + ft+1)2 (18)
• In the original model of Fleten & Lemming (2003), applied for daily steps, asmoothing factor prevents large jumps in the forward curve. However, in the case ofHPFCs, Blöchlinger (2008) (p. 154), concludes that the higher the relative weight ofthe smoothing term, the more the hourly structure disappears.
Input mix for electricity production Germany
Empirical results – p.24
2009 2010 2011 2012 2013 2014Coal 42.6 41.5 42.8 44 45.2 43.2Nuclear 22.6 22.2 17.6 15.8 15.4 15.8Natural Gas 13.6 14.1 14 12.1 10.5 9.5Oil 1.7 1.4 1.2 1.2 1 1Renewable energies from which 15.9 16.6 20.2 22.8 23.9 25.9Wind 6.5 6 8 8.1 8.4 8.9Hydro power 3.2 3.3 2.9 3.5 3.2 3.3Biomass 4.4 4.7 5.3 6.3 6.7 7.0Photovoltaic 1.1 1.8 3.2 4.2 4.7 5.7Waste-to-energy 0.7 0.7 0.8 0.8 0.8 1
Other 3.6 4.2 4.2 4.1 4 4.3
Table 2: Electricity production in Germany by source (%), as shown in Paraschiv, Bunnand Westgaard (2016).
Data
Empirical results – p.25
• We employed a unique data set of 2’386 daily price forward curvesFt(x1), . . . , Ft(xN ) generated each day between 01/01/2009 and 15/07/2015 based onthe latest information from the observed futures prices for the German electricityPhelix price index.
• We firstly deaseasonalize the prices:
Ft(x) = ft(x) + st(x) , (19)
• Estimate the risk premia (θ), the volatility term structure (σ(x)) and analyse thenoise ǫ̃t(x)
Zt(x) := ft+∆t(x) − ft(x) −∆t
∆x(ft(x + ∆x) − ft(x)) (20)
which impliesZt(x) = θft(x)∆t + ǫt(x) , (21)
ǫt(x) = σ(x)ǫ̃t(x) (22)
Increase in renewables: increase in the PFC’s volatility
Empirical results – p.26
-20
-15
-10
-5
0
5
10
15
20
02.2010
03.2010
04.2010
05.2010
06.2010
07.2010
08.2010
09.2010
10.2010
11.2010
12.2010
01.2011
02.2011
03.2011
04.2011
05.2011
06.2011
07.2011
08.2011
09.2011
10.2011
11.2011
12.2011
01.2012
EU
R/M
Wh
-20
-15
-10
-5
0
5
10
15
20
02.2011
03.2011
04.2011
05.2011
06.2011
07.2011
08.2011
09.2011
10.2011
11.2011
12.2011
01.2012
02.2012
03.2012
04.2012
05.2012
06.2012
07.2012
08.2012
09.2012
10.2012
11.2012
12.2012
01.2013
EU
R/M
Wh
Figure 2: Stochastic component of PFCs generated at 01/17/2010 (upper graph) and 01/07/2011second.
Increase in renewables: increase in the PFC’s volatility
Empirical results – p.27
-20
-15
-10
-5
0
5
10
15
20
02.2013
03.2013
04.2013
05.2013
06.2013
07.2013
08.2013
09.2013
10.2013
11.2013
12.2013
01.2014
02.2014
03.2014
04.2014
05.2014
06.2014
07.2014
08.2014
09.2014
10.2014
11.2014
12.2014
01.2015
EU
R/M
Wh
-20
-15
-10
-5
0
5
10
15
20
02.2014
03.2014
04.2014
05.2014
06.2014
07.2014
08.2014
09.2014
10.2014
11.2014
12.2014
01.2015
02.2015
03.2015
04.2015
05.2015
06.2015
07.2015
08.2015
09.2015
10.2015
11.2015
12.2015
01.2016
EU
R/M
Wh
Figure 3: Stochastic component of PFCs generated at 01/17/2013 (upper graph) and 01/07/2014second.
Risk premia
Empirical results – p.28
• Short-term: it oscillates around zero and has higher volatility (similar in Pietz(2009), Paraschiv et al. (2015))
• Long-term: is becomes negative and has more constant volatility (Burger et al.(2007)): In the long-run power generators accept lower futures prices, as they needto make sure that their investment costs are covered.
0 100 200 300 400 500 600 700−0.04
−0.03
−0.02
−0.01
0
0.01
0.02
0.03
0.04
Point on the forward curve (2 year length, daily resolution)
Magnitude o
f th
e r
isk p
rem
ia
Term structure volatility
Empirical results – p.29
• We observe Samuelson effect: overall higher volatility for shorter time to maturity• Volatility bumps (front month; second/third quarters) explained by increased volume
of trades• Jigsaw pattern: weekend effect; volatility smaller in the weekend versus working days
0 100 200 300 400 500 600 700 8000.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
vo
latilit
y (
EU
R)
Maturity points
Explaining volatility bumps
Empirical results – p.30
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
18,000
01/2
009
04/2
009
07/2
009
10/2
009
01/2
010
04/2
010
07/2
010
10/2
010
01/2
011
04/2
011
07/2
011
10/2
011
01/2
012
04/2
012
07/2
012
10/2
012
01/2
013
04/2
013
07/2
013
10/2
013
01/2
014
04/2
014
07/2
014
10/2
014
To
tal
vo
lum
e o
f tr
ades
Front Month 2nd Month 3rd Month 5th Month
Figure 5: The sum of traded contracts for the monthly futures, evidence from EPEX, owncalculations (source of data ems.eex.com).
Explaining volatility bumps
Empirical results – p.31
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
01/2
009
04/2
009
07/2
009
10/2
009
01/2
010
04/2
010
07/2
010
10/2
010
01/2
011
04/2
011
07/2
011
10/2
011
01/2
012
04/2
012
07/2
012
10/2
012
01/2
013
04/2
013
07/2
013
10/2
013
01/2
014
04/2
014
07/2
014
10/2
014
To
tal
vo
lum
e o
f tr
ades
Front Quarter 2nd Quarterly Future (QF) 3rd QF 4th QF 5th QF
Figure 6: The sum of traded contracts for the quarterly futures, evidence from EPEX,own calculations (source of data ems.eex.com).
Statistical properties of the noise
Empirical results – p.32
• We examined the statistical properties of the noise time-series ǫ̃t(x)
ǫt(x) = σ(x)ǫ̃t(x) (23)
• We found: Overall we conclude that the model residuals are coloured noise, withheavy tails (leptokurtic distribution) and with a tendency for conditionalvolatility.
ǫ̃t(xk) Stationarity Autocorrelation ǫ̃t(xk) Autocorrelation ǫ̃t(xk)2 ARCH/GARCHh h1 h1 h2
Q0 0 1 1 1Q1 0 0 0 0Q2 0 1 1 1Q3 0 0 1 1Q4 0 1 0 0Q5 0 1 1 1Q6 0 1 1 1Q7 0 1 1 1
Table 3: The time series are selected by quarterly increments (90 days) along the maturity points on one noisecurve. Hypotheses tests results, case study 1: ∆x = 1day. In column stationarity, if h = 0 we fail to reject thenull that series are stationary. For autocorrelation h1 = 0 indicates that there is not enough evidence to suggestthat noise time series are autocorrelated. In the last column h2 = 1 indicates that there are significant ARCHeffects in the noise time-series.
Autocorrelation structure of noise time series
Empirical results – p.33
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q0
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q1
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q2
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q3
Figure 7: Autocorrelation function in the level of the noise time series ǫ̃t(xk), by takingk ∈ {1, 90, 180, 270}, case study 1: ∆x = 1day.
Autocorrelation structure of noise time series (squared)
Empirical results – p.34
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q0
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q1
0 20 40 60 80 100-0.5
0
0.5
1
Lag
Sam
ple
Auto
corr
ela
tion
ACF Q2
0 20 40 60 80 100-0.5
0
0.5
1
LagS
am
ple
Auto
corr
ela
tion
ACF Q3
Figure 8: Autocorrelation function in the squared time series of the noise ǫ̃t(xk)2, bytaking k ∈ {1, 90, 180, 270}, case study 1: ∆x = 1day.
Leptokurtic distribution
Empirical results – p.35
-20 -15 -10 -5 0 5 10 15 20 250
0.05
0.1
0.15
0.2
0.25
epsilon t (k=7)
Normal density
Kernel (empirical) density
-20 -15 -10 -5 0 5 10 15 200
0.05
0.1
0.15
0.2
0.25
epsilon t (k=10)
Normal density
Kernel (empirical) density
Normal Inverse Gaussian (NIG) distribution for coloured noise
Empirical results – p.36
-4 -3 -2 -1 0 1 2 3 40
0.2
0.4
0.6
0.8
epsilon t(1)
Normal density
Kernel (empirical) density
NIG with Moment Estim.
NIG with ML
-15 -10 -5 0 5 100
0.5
1
1.5
2
epsilon t(90)
Normal density
Kernel (empirical) density
NIG with Moment Estim.
NIG with ML
-10 -5 0 5 100
0.5
1
1.5
2
epsilon t(180)
Normal density
Kernel (empirical) density
NIG with Moment Estim.
NIG with ML
-15 -10 -5 0 5 10 150
0.5
1
1.5
2
epsilon t(270)pdf
Normal density
Kernel (empirical) density
NIG with Moment Estim.
NIG with ML
Spatial dependence structure
Empirical results – p.37
Figure 9: Correlation matrix with respect to different maturity points along one curve.
Conclusion and future work
Empirical results – p.38
• We developed a spatio-temporal dynamical arbitrage free model for electricityforward prices based on the Heath-Jarrow-Morton (HJM) approach under Musielaparametrization
• We examined a unique data set of price forward curves derived each day in themarket between 2009–2015
• We examined the spatio-temporal structure of our data set
– Risk premia: higher volatility short-term, oscillating around zero; constantvolatility on the long-term, turning into negative
– Term structure volatility: Samuelson effect, volatility bumps explained byincreased volume of trades
– Coloured (leptokurtic) noise with evidence of conditional volatility– Spatial correlations structure: decaying fast for short-term maturities;
constant (white noise) afterwards with a bump around 1 year
• Last step (future work): test for a realistic Lévy stochastic process in Hilbert spacefor the noise time series, with NIG marginals and spatial covariance operator.
References
Empirical results – p.39
• Andresen, A., S. Koekebakker, and S. Westgaard (2010): Modeling electricityforward prices using the multivariate normal inverse Gaussian distribution, TheJournal of Energy Markets, 3(3), 3–25.
• Benth, F.E., and S. Koekebakker (2008): Stochastic modeling of financial electricitycontracts, Energy Economics, 30(3), pp. 1116–1157.
• Cartea, A., and M. Figueroa (2005): Pricing in electricity markets: a mean revertingjump diffusion model with seasonality, Applied Mathematical Finance 12, pp.313–335.
• Fleten, S.-E., and J. Lemming (2003): Constructing forward price curves inelectricity markets, Energy Economics, 25, pp. 409–424.
• Kiesel, R., G. Schindlmayr, and R. Boerger (2009): A two-factor model for theelectricity forward market, Quantitative Finance 9, 279–287.
• Lucia, J.J., and E. S. Schwartz (2002): Electricity prices and power derivatives:evidence from the Nordic power exchange, Review of Derivatives Research 5, pp.5–50.
• Paraschiv, F., S.-E. Fleten, and M. Schuerle (2015): A spot-forward model forelectricity prices with regime shifts, Energy Economics, 47, pp. 142-153.
• Roncoroni, A., and P. Guiotto (2001): Theory and Calibration of HJM with ShapeFactors, Mathematical Finance, Springer Finance, 2001, 407–426.