Stochastic modelling and statistical inference for electricity prices, wind energy production and wind speed
File(s)
Author(s)
Rowińska, Paulina A.
Type
Thesis
Abstract
Although wind energy helps us slow down the increase of global temperatures,
its weather-dependence and unpredictability make it risky to invest in. In this
thesis we apply statistical and mathematical tools to enable energy providers
to accurately plan such investments.
In the first part we want to understand the impact of wind energy on electricity
prices. We extend an existing multifactor model of electricity spot prices by
including stochastic volatility as well as the information about wind energy
production. Empirical studies indicate that these additions improve the
model fit. We also model wind-related variables directly, using Brownian
semistationary processes with generalised hyperbolic marginals. Finally, we
introduce a joint model of prices and wind energy production suitable for
quantifying the risk faced by energy distributors.
The second goal is to produce accurate short-term wind speed forecasts based
on historical data instead of computationally expensive physical models. We
achieve this by splitting the wind speed into two horizontal components
and modelling them with Brownian semistationary processes with a novel
triple-scale kernel. We develop efficient estimation and forecasting procedures.
Empirical studies show that such modelling choices result in good forecasting
performance.
its weather-dependence and unpredictability make it risky to invest in. In this
thesis we apply statistical and mathematical tools to enable energy providers
to accurately plan such investments.
In the first part we want to understand the impact of wind energy on electricity
prices. We extend an existing multifactor model of electricity spot prices by
including stochastic volatility as well as the information about wind energy
production. Empirical studies indicate that these additions improve the
model fit. We also model wind-related variables directly, using Brownian
semistationary processes with generalised hyperbolic marginals. Finally, we
introduce a joint model of prices and wind energy production suitable for
quantifying the risk faced by energy distributors.
The second goal is to produce accurate short-term wind speed forecasts based
on historical data instead of computationally expensive physical models. We
achieve this by splitting the wind speed into two horizontal components
and modelling them with Brownian semistationary processes with a novel
triple-scale kernel. We develop efficient estimation and forecasting procedures.
Empirical studies show that such modelling choices result in good forecasting
performance.
Version
Open Access
Date Issued
2019-12
Date Awarded
2020-04
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
Veraart, Almut
Sponsor
Engineering and Physical Sciences Research Council (EPSRC)
EDF Energy (Firm)
Grant Number
EP/L016613/1
Publisher Department
Mathematics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)