Modelling Power Spikes with Inhomogeneous Markov-Switching Models
(2013) FMS820 20132Mathematical Statistics
- Abstract (Swedish)
- Interest in modelling electricity prices has, despite its relatively short history, resulted
in widespread types of models that tend to be too intricate to incorporate most price
characteristics. This thesis pursues a flexible approach that comprehends stylized facts of
electricity prices while it still handles complexity in order to facilitate calibration and
forecasting. Although time-varying transitions of non-linear Markov-switching models
add a new dimension to the problem, the extension is pivotal to encompass the timing
of power spikes. Simulation studies provide a comparison between the maximum likelihood
estimator and the EM algorithm and validate the precision of the estimators. A
comprehensive study of the model... (More) - Interest in modelling electricity prices has, despite its relatively short history, resulted
in widespread types of models that tend to be too intricate to incorporate most price
characteristics. This thesis pursues a flexible approach that comprehends stylized facts of
electricity prices while it still handles complexity in order to facilitate calibration and
forecasting. Although time-varying transitions of non-linear Markov-switching models
add a new dimension to the problem, the extension is pivotal to encompass the timing
of power spikes. Simulation studies provide a comparison between the maximum likelihood
estimator and the EM algorithm and validate the precision of the estimators. A
comprehensive study of the model framework in the independent regime setting that
is applied to real data from the German and Nordic markets confirms the hypothesis
that extensive models with exogenous variables outperform time-invariant counterparts.
Improvements of electricity price dynamics and other issues involved in the process of
modelling electricity prices as well as potential future research topics are also suggested
and discussed. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/4194640
- author
- Norén, Vicke
- supervisor
- organization
- course
- FMS820 20132
- year
- 2013
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 4194640
- date added to LUP
- 2014-01-09 10:51:15
- date last changed
- 2014-01-09 10:51:15
@misc{4194640, abstract = {{Interest in modelling electricity prices has, despite its relatively short history, resulted in widespread types of models that tend to be too intricate to incorporate most price characteristics. This thesis pursues a flexible approach that comprehends stylized facts of electricity prices while it still handles complexity in order to facilitate calibration and forecasting. Although time-varying transitions of non-linear Markov-switching models add a new dimension to the problem, the extension is pivotal to encompass the timing of power spikes. Simulation studies provide a comparison between the maximum likelihood estimator and the EM algorithm and validate the precision of the estimators. A comprehensive study of the model framework in the independent regime setting that is applied to real data from the German and Nordic markets confirms the hypothesis that extensive models with exogenous variables outperform time-invariant counterparts. Improvements of electricity price dynamics and other issues involved in the process of modelling electricity prices as well as potential future research topics are also suggested and discussed.}}, author = {{Norén, Vicke}}, language = {{eng}}, note = {{Student Paper}}, title = {{Modelling Power Spikes with Inhomogeneous Markov-Switching Models}}, year = {{2013}}, }