Вы находитесь на странице: 1из 47

Modelling Spot Prices in the Iberian

Electricity Market

Sandro Navarro

Departamento de Economía
Facultad de Ciencias Económicas y Empresariales

2018
Table of Contents Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
1 Introduction A multifactorial
model

2 Data
Calibration

Results and model


comparison
3 A multifactorial model

4 Calibration

5 Results and model comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


2 de 47
Electricity Market
Table of Contents Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
1 Introduction A multifactorial
model

2 Data
Calibration

Results and model


comparison
3 A multifactorial model

4 Calibration

5 Results and model comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


3 de 47
Electricity Market
The electricity spot market Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

The electricity market works as a mandatory pool where the pur- A multifactorial
model
chase and sale of electricity for the next day (day-ahead ) is man- Calibration
aged. Results and model
comparison

That is, all the owners of the electricity production and acquisition
units are obliged to make their oers through the market operator
in accordance with the procedure established by the latter.

The daily market is structured in a single session for each hour of


the daily programming horizon.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


4 de 47
Electricity Market
The electricity spot market Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 1: The spot electricity market is typically a day-ahead auction market that
does not allow for continuous trading.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


5 de 47
Electricity Market
Table of Contents Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
1 Introduction A multifactorial
model

2 Data
Calibration

Results and model


comparison
3 A multifactorial model

4 Calibration

5 Results and model comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


6 de 47
Electricity Market
Description of data Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 2: Serie de precios spot en el período Julio 2006 - Diciembre 2012. Mercado
Ibérico de la Electricidad.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


7 de 47
Electricity Market
Description of data Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 3: Descriptive statistics of daily spot prices for the period July 2006- December
2012. Iberian Electricity Market.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


8 de 47
Electricity Market
Stylized Features of Electricity Prices Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Seasonality, which is a product of the inelasticity of demand, form- Calibration

Results and model


ing clear patterns caused by economic and business activities. comparison

Other factors inuencing seasonal behavior are temporary changes


in supply and demand conditions and the fact that electricity is not
a storable asset.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


9 de 47
Electricity Market
Stylized Features of Electricity Prices Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 4: Precios observados y componente estacional a largo plazo estimad en el


periodo Julio 2006 - Diciembre 2012. Mercado Ibérico de la Electricidad.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


10 de 47
Electricity Market
Stylized Features of Electricity Prices Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
Stationarity and mean reversion. A multifactorial
model

The presence of jumps, because the rate of mean reversion is much Calibration

higher than during normal evolution. Results and model


comparison

High volatility, which exhibits a combination of both stochastic


and deterministic behavior.

Non-Gaussian behavior, which occurs because spikes distort the


normal distribution expected from the natural uctuations of the
data, especially around its tails

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


11 de 47
Electricity Market
A taxonomy of electricity spot price modeling Modelling Spot
Prices in the

approaches
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 5: A taxonomy of electricity spot price modeling approaches.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


12 de 47
Electricity Market
A taxonomy of electricity spot price modeling Modelling Spot
Prices in the

approaches
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
Multi-agent models generally concerned medium and long term A multifactorial

time horizons, and involved matching demand estimates to the model

Calibration
supply (Borenstein, Bushnell, and Knittel 1999; Borgosz-Koczwara,
Results and model
Weron and Wyªoma«ska 2009). comparison

Fundamental models tries to capture the basic physical and eco-


nomic relationships which are present in the production and trading
of electricity (Gonzalez, Contreras, and Bunn 2012; Karakatsani and
Bunn 2008; Kristiansen 2012; Liebl 2013).

Reduced form models, whose main intention is replicate the main


characteristics of daily electricity prices (Gjolberg and Brattested
2011; Kristiansen 2007; Ronn and Wimschulte 2009).

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


13 de 47
Electricity Market
A taxonomy of electricity spot price modeling Modelling Spot
Prices in the

approaches
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model
Statistical models, which forecast the current price by using a
Calibration
mathematical combination of the previous prices and/or previous Results and model
or current values of exogenous factors (Kim, Yu, and Song 2002; comparison

Koopman, Ooms, and Carnero 2007).

Computacional intelligence (CI) models, which combines ele-


ments of learning, evolution and fuzziness to create approaches that
are capable of adapting to complex dynamic systems (Poole, Mack-
worth and Goebel 1998; Madani, Correia, Rosa and Filipe 2011;
Wang and Fu 2005).

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


14 de 47
Electricity Market
Objectives Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

The contribution of the present work in the modeling of electricity Data

prices is threefold. A multifactorial


model

1 First, a multifactor model is developed to incorporate the styl- Calibration

ized facts of the electricity prices and to model the pattern of Results and model
comparison
clusters of jumps through stochastic jump intensity.
2 Second, a methodological contribution is presented for the es-
timation of the jump component.
3 Finally, despite having a stochastic structure for the intensity,
explicit formulas are introduced for the forward prices and also,
since our model is of arithmetic type, these formulas are rela-
tively simple to evaluate.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


15 de 47
Electricity Market
Table of Contents Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
1 Introduction A multifactorial
model

2 Data
Calibration

Results and model


comparison
3 A multifactorial model

4 Calibration

5 Results and model comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


16 de 47
Electricity Market
A multifactorial model Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
We shall consider a model for the spot price S = (St )t∈[0.T ] of the model

form Calibration

Results and model


comparison
(d)
St = e f (t) St (1)
where:
f : [0, t] → R is a seasonal and trend component.
(d)
St are the deseasonalized prices.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


17 de 47
Electricity Market
A multifactorial model Modelling Spot
Prices in the
Iberian Electricity
Market
We propose an arithmetic two-factor model of the form Sandro Navarro

(d)
St = Xt + Yt Introduction

Data

where A multifactorial
model
1 The diusive component X is Calibration

Results and model


dXt = αX (µX − Xt )dt + σX dWt comparison

2 The jump component Y satises the SDE

dYt = −αY Yt − dt + dIt


where I es a non-homogeneous compund Poisson process with
jump distribution φ and stochastic intensity `. More precisely,
we will take ` to be the solution of
d`t = −ν`t − dt + dMt (2)
where M is an homogeneous Poisson process of constant in-
tensity λ > 0.
Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian
18 de 47
Electricity Market
A multifactorial model Modelling Spot
Prices in the
Iberian Electricity
Jump distribution φ Market

Sandro Navarro

Introduction

Data
Finally, we will consider four dierent possibilities for the jump dis- A multifactorial
tribution φ and evaluate which is more convenient for this or other model

models. These possibilities are: Calibration

Results and model


comparison

1 φ ∼ N(a, b 2 ) for some a ∈ R, b > 0.


2 φ is a mixture of two distributions with positive support, each
corresponding to the laws of positive and negative jumps, re-
spectively.

More precisely, we assume that φ has a density fφ of the form

fφ (x) = pfφ+ (x) + (1 − p)fφ− (−x)

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


19 de 47
Electricity Market
A multifactorial model Modelling Spot
Prices in the
Iberian Electricity
Jump distribution φ Market

Sandro Navarro

Introduction

Data

for some xed p ∈ (0, 1), where φ+ and φ− are two distribu- A multifactorial
model
tions with positive support and fφ+ , fφ− denote their respective Calibration
densities. Here, p represents the probability of a given jump be- Results and model

ing positive, whereas φ+ and φ− represent the laws of positive comparison

and negative jumps, respectively.

For simplicity, we will assume that both φ+ and φ− belong to


the same family of distributions, which can be either Gamma,
Log-normal or Generalized Pareto distributions. Recall that
their respective densities are given by the formulas:

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


20 de 47
Electricity Market
A multifactorial model Modelling Spot
Prices in the
Iberian Electricity
Jump distribution φ Market

Sandro Navarro

Introduction

Data
1 The Gamma distribution Γ(a, b) has density
A multifactorial
model
b a a−1 −bx Calibration
f (x) = x e 1(0,+∞) (x)
Γ(a) Results and model
comparison

for a, b > 0.

2 The Log-normal distribution LN(a, b 2 ) has density

1 (logx−a)2
f (x) = √ e− 2b 2 1(0,+∞) (x)
x 2πb 2
for a∈R and b > 0.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


21 de 47
Electricity Market
A multifactorial model Modelling Spot
Prices in the
Iberian Electricity
Jump distribution φ Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration
3 The Generalized Pareto distribution GP(a, b, c) has density Results and model
comparison

1
 x − c −1− 1a
1[c,+∞) (x)

f (x) = 1 +a
b b
for a, b, c > 0.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


22 de 47
Electricity Market
Table of Contents Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
1 Introduction A multifactorial
model

2 Data
Calibration

Results and model


comparison
3 A multifactorial model

4 Calibration

5 Results and model comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


23 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model
The procedure includes: Calibration

Results and model


1 Deseasonalization of data. comparison

2 Identifying the factors in the deseasonalized data.


3 Estimating the parameters in the diusive component.
4 Estimating the parameters in the jump component Y .

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


24 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Step 1: Deseasonalization Sandro Navarro

Introduction
The rst step in the algorithm is to remove the seasonality in the
Data
data, obtaining an estimated trend function fˆ and the deseasonal- A multifactorial
ized data series given by model

Calibration
(d) ˆ
Ŝt := e −f (t) St . Results and model
comparison

f (t) = s(t) + L(t) (3)


where s is a weekly periodic component and L is a long-term sea-
sonal component. Here L takes the form
2πt
 
L(t) = α1 sin + α2 + α3 + α4 EWMA(t; 0.975),
365
where EWMA stands for exponentially weighted moving average,
which is dened inductively by the relation

EWMA(t; λ) = (1 − λ)St + λEWMA(t − 1; λ).

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


25 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 6: Deseasonalized prices

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


26 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

Step 2: Identifying the factors in the deseasonal- A multifactorial

ized data model

Calibration

Results and model


The next step is to identify and separate the diusive and jump comparison

components from the deseasonalized data.

We will apply a method know as the Hard-Thresholding ltering


algorithm, but rst we need to obtain a preliminary estimate of
the reversion rates αX and αY which the algorithm will need as an
input.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


27 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Step 2.1: Estimating the reversion rates αX and Introduction

αY Data

A multifactorial
model
We will obtain these by studying the autocorrelation function (ACF) Calibration
associated to the deseasonalized process S (d) . Since X and Y are Results and model
independent Ornstein-Ühlenbeck processes, the (stationary) auto- comparison

correlation function of S (d) = X + Y is

ρ(h) = ωe −αX h + (1 − ω)e −αY h (4)

where
Var(X )
ω= .
Var(X ) + Var(Y )
We compare the empirical autocorrelation (ACF) with the theoreti-
cal autocorrelation function for dierent factors using the L2 norm.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


28 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Table 1: The tted ACF with a sum of two exponentials. Iberian Electricity Market Calibration
between July 2006 and December 2012. Results and model
comparison

Figure 7: Empirical ACF (black) vs Theoretical ACF (red)

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


29 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Step 2.2: The Hard-Tresholding algorithm Data

A multifactorial
model
With α̂Y and α̂Y we proceed to separate the diusive and jump Calibration
components in the deseasonalized spot prices. The idea is to de- Results and model
(d)
compose the deseasonalized data as Ŝt = X̂t + g (t) where the comparison

function g represents our estimation of the jump process Y and X̂


that of X .
(d)
This algorithm proposes to determine the parameters in Ŝt above
through an iterative scheme of K steps, which can be summarized
as follows:
We take as our initial data series X (0) := Ŝ d .

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


30 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

Having constructed the series X (i) for 0 ≤ i < K , we dene A multifactorial


model
inductively the series
Calibration

(i+1) (i) Results and model


Xt := Xt − g̃ (t, αi+1 , τ̃i+1 ), comparison

where g̃ (t, τ, α) := αe −αˆY (t−τ ) 1[τ,T ] (t) and the pair (αi+1 , τ̃i+1 )
solves the least squares problem

2
( T )
X  (i)
min Xt − g̃ (t, α, τ ) : α ∈ R, τ ∈ {1, . . . , T } .
t=1

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


31 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

The algorithm concludes dening the series Data

A multifactorial
K model
X
X̂ := X (K ) and g (·) = g̃ (·, αi , τ̃i ), Calibration

Results and model


i=1 comparison

(K )
where K is chosen so that in the return series (Rt )t=1,...,T
of X (K ) dened as
(K ) (K ) (K )
Rt := Xt − e −α̂X Xt−1

there are no evidences at level 0,05 to reject its Gaussianity


when performing a Jarque-Bera test.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


32 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 8: Detected log jumps and log diusive component. Iberian Electricity
Market between July 2006 and December 2012.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


33 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Step 3: Estimating the parameters in the diusive Introduction

component Data

A multifactorial
model
We t the equations Calibration

Results and model


comparison
X̂t+1 = β0 + β1 X̂t + σεt

for t = 0, . . . , T − 1 by the maximum likehood method and then


obtain the estimators

log β̂12
s
βˆ0
α̂X := − log β̂1 µ̂X := σ̂X := σ̂ .
1 − β̂1 β̂12 − 1

We obtained the values α̂X = 0.0858, µ̂X = 1.0149 and σ̂X =


0.0664.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


34 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Step 4: Estimating the parameters in the jump


Data

component Y
A multifactorial
model

Calibration

Results and model


We shall base our estimation procedure in this case on a simulated comparison

moment method originally proposed in Carrasco et al. (2002).

As information one has that Y satises the system of n equations

E(hi (Yt ; θ0 )) = 0 (5)

for certain known functions h1 , . . . , hn called moment conditions.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


35 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction
The eciency of the method will depend on the amount n of equa- Data
tions at our disposal, although it will always remain below the one A multifactorial

achieved via the maximum likelihood (ML) method. model

Calibration

Results and model


Carrasco and Florens proposed a method that achieves an asymp- comparison

totic eciency matching using a continuum of equations based on


the characteristic function of the process

hτ (Yt ; θ) := e iτ Yt − Eθ (e iτ Yt ),

so that E(hτ (Yt ; θ)) is exactly the dierence between the true char-
acteristic function and the theoretical one implied by the model with
parameter vector θ.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


36 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

The estimation method can be summarized as follows: Sandro Navarro

1 Observe that Y satises the equation Introduction

Data

E(hi (Yt ; θ0 )) = 0 A multifactorial


model

Calibration
for all τ ∈ R, where θ0 are the true parameters and
Results and model
comparison
iτ Yt iτ Yt
ht (τ, Y ; θ) = e − Eθ (e ).

2 We approximate E(ht (τ, Y ; θ0 )) by its sample version

T J
1 X iτ Yt 1 X iτ Ytθ
ht (τ, Y ; θ) = e − e
T t=1 J t=1

where (Y θ )t=1,..,J is a simulated series with parameters θ.


3 We take our estimator θ̂0 := arg minθ∈Θ ||ht (•, Y ; θ)||, where
|| • || is a suitable norm in the space of functions (in τ ).

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


37 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
We use an algorithm developed by Grynkiv (2012) which selects a model

nite number of moment conditions from the continuum of available Calibration

ones in such a way that the estimator obtained using only this nite Results and model
comparison
set of conditions is essentially ecient.

Since we do not know a priori the distribution of jumps in our model,


we calibrate our model using four dierent possible jump distribu-
tions and then select the one which is most appropriate following
the outcome of the goodness-of-t test.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


38 de 47
Electricity Market
Procedure Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Table 2: Estimating parameters of jump component

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


39 de 47
Electricity Market
Table of Contents Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data
1 Introduction A multifactorial
model

2 Data
Calibration

Results and model


comparison
3 A multifactorial model

4 Calibration

5 Results and model comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


40 de 47
Electricity Market
Testing model Modelling Spot
Prices in the
Iberian Electricity
Benchmark models Market

Sandro Navarro

Introduction

Data

A multifactorial
model
Cartea-Figueroa (CF): Calibration

Results and model


dXt = −αX Xt dt + σX dBt + dJt (6) comparison

Hambly-Howison-Kluge (HHK):
(d)
St = exp(Xt + Yt ), where
dXt = −αX Xt dt + σX dBt (7)
dYt = −αY Yt dt + dJt (8)

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


41 de 47
Electricity Market
Testing model Modelling Spot
Prices in the
Iberian Electricity
Benchmark models Market

Sandro Navarro

Mayer-Schmid-Weber (MSW): Introduction

Diusive component is similar to HHK model, but σX (t) is a Data

GARCH (1,1) process: A multifactorial


model

Calibration
σX2 (t) = α0 + α1 ε2t + β1 σX2 (t − 1) (9) Results and model
comparison

Meyer-Brandis and Tankov (MT):


(d)
St = Xt + Yt

is proposed, where X and Y are respectively given by the SDEs



 dXt = αX (µX − Xt )dt + σX dWt
(10)
dYt = −αY Yt − dt + dIt .

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


42 de 47
Electricity Market
Analysis of path properties Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Figure 9: Simulated spot prices under dierent model specications.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


43 de 47
Electricity Market
Quantile analysis Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Table 3: Goodness-of-t statistics for M and benchmark models tted to the


deseasonalized prices in the Iberian Electricity Market between July 2006 and
December 2012

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


44 de 47
Electricity Market
Moment analysis Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Introduction

Data

A multifactorial
model

Calibration

Results and model


comparison

Table 4: Moment comparison

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


45 de 47
Electricity Market
Conclusions Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

About the analysis of path properties, we observe that all models Introduction

Data
perform reasonably well, being MT and M those with the trajectories
A multifactorial
which most resemble that of the original series. model

Calibration

About the quantile analysis, we see that for our model the quantile Results and model
comparison
adjustment is worse for the lower quantiles, but this also occurs for
the benchmark model.This suggests that the lowest levels of the
spot price might correspond not only to negative jumps, but also
to periods of strong volatility, a feature that cannot be captured by
constant-volatility models.

In the moment analysis, compared to the other models, we see that


our model M achieves the best results in this regard, followed closely
by MT.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


46 de 47
Electricity Market
References Modelling Spot
Prices in the
Iberian Electricity
Market

Sandro Navarro

Cartea, A., Figueroa, M., 2005. Pricing in electricity markets: Introduction

a mean reverting jump diusion model with seasonality. Applied Data

Mathematical Finance, 12(4),313335. A multifactorial


model

Calibration
Hambly, B., Howison, S., Kluge, T., 2009. Modelling spikes and Results and model
pricing swing options in electricity markets. Quantitative Finance, comparison

9(8), 937949.

Mayer, K., Schmid, T., Weber, F., 2015. Modeling electricity spot
prices: combining mean reversion, spikes, and stochastic volatility.
The European Journal of Finance, 21(4), 292315.

Meyer-Brandis, T., Tankov, P., 2008. Multi-factor jump-diusion


models of electricity prices. International Journal of Theoretical and
Applied Finance, 11(05), 503528.

Sandro Navarro (UDEP) Modelling Spot Prices in the Iberian


47 de 47
Electricity Market

Вам также может понравиться