dt.ENERGY has gained hiGhER pRofits in practice

dt.ENERGY has gained
higher profits in practice
DT.Energy is applied day by day for the
municipal cooperation power plant in
Hamm-Uentrop since 01.01.2008
The municipal energy company service provider Trianel
is using cutting edge methodologies in day-ahead power
plant dispatch. Since January 2008, when their CCGT
plant has started its commercial operation, Trianel is
using new methods of stochastic optimization implemented
in the tool DT.Plantval for dispatch planning. Due to the
structure of this “cooperation power plant”, which
belongs to 28 independent share holders, a practical
commercial comparison between traditional methods for
power plant dispatch (deterministic models or day-type
methods), and the new method of stochastic optimization
was possible.
The multistage stochastic optimization method
implemented in DT.Plantval is more and more used in different branches of industries where decision making is
influenced by significant risks of future impacts. As
opposed to Monte Carlo simulation (scenario-by-scenario
approach), in DT.Plantval the decision process to
manage limited resources over time is based on non-anticipatively, i.e., without assuming perfect foresight in each
scenario on uncertain impacts, such as the power price.
The stochastic optimization is based on the representation
of future uncertainties by scenario trees. Thereby, in particular the valuation and operation of path dependent flexible assets such as CCGTs with limited fuel contracts,
pumped storages and gas storages can be performed in
accordance with the real world decision makers’ situation.
We consider the 850 MW CCGT power plant in HammUentrop, which is the first German municipal cooperation
power plant and has started its operation in January
2008. Each of the 28 shareholders controls a structurally
identical slice of the power plant with an individual share
(7-150 MW) of total power. These slices can be
dispatched individually on a daily basis. Due to limited
fuel, long-term production planning has to be considered
for the day ahead schedules from the very beginning.
In order to achieve an economically optimal operation of
the power plant, on the one hand the day ahead spark
spread is the major impact for the day ahead dispatch.
On the other hand, the economic potential of the fuel in
the future – until the end of the fuel supply contract –
has to be considered, which is at significant risk due to
the volatility of future spark spreads. Moreover, next to
variable costs, fixed operation costs and start up costs
need to be represented. Some slice owners manage their
slice independently, however, the majority of slices is
managed by certain optimization model providers. Trianel
is currently managing 13 of the 28 slices through its
generation portfolio management team by using the tool
DT.Plantval. In 2008, there were six optimization model
providers managing different slices that had been competing against each other.
A comparison of the power plant utilization based on
normalized (w.r.t. the size of the slices) cumulated power
consumption for the different suppliers shows that the
slices that were managed by suppliers 1-3 and the slices
managed by Trianel have used up the available fuel to its
full extent, Whereby suppliers 4 and 5 haven’t. Note that
the theoretically optimal power plant generation schedule
is defined by the real evolution of the EEX spot prices in
the gas year 2007/2008.
The profits of the relevant schedules determined by
different optimization models are shown in the following
figure. The values are given as a percentage of the
theoretically maximum profit that could have been
achieved.
The stochastic methods integrated in the software
DT.Plantval are being applied by several utilities in
Germany and Austria. They are used for the operational
dispatch optimization of CCGTs, coal fired power plants as
well as generation portfolios comprising flexible fuel supply
contracts and gas storages. Furthermore, existing and planned power plants of Mark-E-Aktiengesellschaft in Hagen,
Germany, Rheinenergie AG in Cologne, Germany, Dong
Energy in Copenhagen, Denmark, Energieversorgung
Niederösterreich in Vienna, Austria, Statkraft Markets in
Düsseldorf, Germany and others have been valued by
models and software of DT.Plantval.
Version: 02.2012
Performance benchmark on day ahead dispatch of power plant
slices between Jan 1st, 2008 and Sep 30th, 2008
Theoretic optimum
100%
Normated profit per 10 MW capacity of plant
The comparison between all schedules that had been
determined by different optimization models shows that
Trianel’s optimization performed with the highest profit at
the end of the optimization period. Trianel has performed
with 97,23 % of the theoretically maximum profit, which is
an advance of 1,54 percentage points as compared to the
second best dispatch. In relation to the generation of the
whole power plant, this advance equals a benefit of
millions of Euros. The schedule that performed worst
resulted in a value of 94,04 % of the theoretically
maximum profit. We state that Trianel’s stochastic
optimization has performed excellently even in its very first
period of application. This emphasizes the efficiency of the
applied optimization model in DT.Plantval and the
general need of using probabilistic methods in volatile
energy markets. Therefore, Trianel is looking forward to
operate their current and future power plants in future with
DT.Plantval.
97,2
95,8
95%
95,2
95,0
94,7
94,0
90%
85%
0%
Other slice owners
(daily or weekly deterministic
optimization or day type schedules)
Trianel
(Daily stochastic
optimization with
DT.Energy)
DT.Energy has achieved 97,2% of the theoretically
optimal profit for the slices of the CCGT in HammUentrop that are operated by Trianel
Other optimization model providers have achieved
2.3% less profit in average
For the power plant as a whole, Trianel has gained
an increase in profits of multiple millions of Euro by
using stochastic optimization in DT.Energy