41 / 52
41
Master Thesis
Department of Electrical
Engineering
Analysis of End Conditions
for a Stochastic Short Term
Model
By Håvard Hansen
Supervisor: Magnus Korpås
Background
A new model for short term planning of hydro power (SHARM-model) has been developed
and is being tested in an ongoing research project at SINTEF Energi. SHARM aims at
improving the daily production planning when the uncertainty of important parameters
increases. The SHARM-model accounts for uncertain market price and inflow to the
reservoirs, and will give a better basis for decisions and more robust planning when multiple
strategies must be weighed against each other.
The ongoing research project aims at finding out if a stochastic
formulation gives a possibility of increased economic earnings or
other operational improvements, as well as how such a model can be
operationalized at the producers.
The short term model is coupled to the seasonal model through water
values that are used to assess the water in the timeframe of the
short term model. The water value represents the resource cost of
the water and it is important that the coupling between the models
is consistent. In SHARM the water value may be descirved as either a
constant end value per Mm3 or by use of cuts from the seasonal model.
The use of cuts is especially suited in a stochastic model because the
uncertainty makes multiple end states possible. In a deterministic
model where only one course is given, there is a rather narrow band
of possible end states given by things like maximum and minimum
production. Because of this it is interesting to look at the use of cuts
in a deterministic versus a stochastic model. What cuts are chosen by
the deterministic model? What cuts are chosen by the stochastic? Does
the stochastic model need more cuts as input? In that case, what does
it depend on? Does more extreme variations in price and inflow require
more extreme cuts?
Hydro scheduling hierarchy