Operations Management in Short Term Power Markets

Publikation: Bog/antologi/afhandling/rapportPh.d.-afhandling

  • Ditte Mølgård Heide-Jørgensen
Electricity market models have often been modelled as deterministic or at most
two-stage stochastic models with an hourly time resolution. This thesis looks into
possible ways of extending such models and formulating new models to handle
both higher time resolution than hourly and stochastics without losing computational
tractability. The high time resolution is crucial to correctly describe renewables,
such as wind power, and capture how they affect the system and the
system costs, since they are often fluctuating and hard to predict, also within the
hour.
The thesis consists of four chapters. The first is an introduction to the background
for the work with stochastic electricity market models with a high time
resolution. It is followed by three self-contained chapters.
The second chapter Short-term balancing of supply and demand in an electricity
system: forecasting and scheduling is on a balancing market model like in the Nordic
countries with high time resolution, and it takes extensive balancing rules into
consideration. We look into how wind forecast errors can be handled in a system
with a large and increasing amount of wind power and at what costs. The project
was done in collaboration with Jeanne Aslak Petersen, a PhD student at Aarhus
University and the Danish TSO Energinet.dk, and the chapter is identical to the
published paper Petersen et al. (2016) except that the reference list is part of the
common reference list for the thesis.
The third chapter A dynamic programming approach to the ramp-constrained intrahour
stochastic single-unit commitment problem considers a real-time market setup.
We describe two stochastic multi-stage single-unit commitment model in which
commitment decisions are made on an hourly basis and dispatch decisions are
made on a higher time resolution, e.g. 5 minutes. The stochastic input is the electricity
price modelled as a time inhomogenous Markov chain that the power producer
uses to maximise profits. To maintain computational tractability with such
high time resolution and stochastics the model is solved with dynamic programming.
The two models differ in the way the dynamic programming algorithm
handles the integer variables leading to two different non-anticipativity assumptions.
In the fourth chapter Open- and closed-loop equilibrium models for the day-ahead
and balancing markets we look into how power producers act in market which
is not perfectly competitive. Specifically, we look into the possibility of exercising
market power when the electricity market consists of two sequential markets
– the day-ahead market and the balancing market – and some power producers
have access to both markets whereas others only can participate in the first
market. The model is formulated with both an open-loop and closed-loop approach,
and we find that the solution to the more realistic, but also computationally
harder closed-loop model differs substantially from the open-loop solution.
Again the day-ahead market is assumed to have hourly time resolution, but the
balancing market has a higher time resolution, e.g. 5 minutes.
OriginalsprogEngelsk
ForlagDepartment of Mathematical Sciences, Faculty of Science, University of Copenhagen
StatusUdgivet - 2016

ID: 168783983