Bidding strategies on international energy markets

Bidding strategies on international energy markets
Rocco Melzian, Niels Ehlers
Lehrstuhl für Energiesysteme, TU Berlin, Einsteinufer 25 (TA8), 10587 Berlin1
Tel: +49-(0)30-314 28163, Fax: +49-(0)30-314 26908, [email protected]
Key questions
Since the beginning of the liberalization of European energy markets when prices for electricity fell under the
monopolistic levels they rose again steadily within the last years. Especially wholesale prices at the European
Energy Exchange nowadays partly exceed retail levels which are still under supervision of the Federal Cartel
Office and thus price-capped, a situation that puts high economic pressure on local utilities. In accordance to
the rising public discussion this paper seeks a scientific approach for finding the cause of rising prices, the
question of applied market power and comments on the German and European market design.
Applied methods and approach
Our approach to the question of applied market power in the German wholesale electricity market consists of
three steps. At first, a database was set up including all relevant market data. It contains all available information
of the European Energy Exchange in Leipzig, including prices for electricity on the spot and long-term market,
prices for EU emission allowances, coal indices and as one of the key components, aggregated curves of supply
and demand bids for the electricity spot market. Since mid 2006, the EEX participants also publish the
availability and commitment of their power plants, which is also read into the database. Gas prices at TTF and
current electricity grid data imported from the TSO´s websites conclude the data collection.
The second and key component of this study was the design of an agent-based market simulation. In this system,
the German power plant fleet was modelled that allows assigning an individual strategy to each power plant
owner. Based on the power plant data and energy prices, individual strategies are simulated and bids are
submitted to the electricity exchange so the market clearing price can be calculated.
The third step was an empirical comparison of the results found by the simulations and the real market data. The
results for the German market were then compared to bidding strategies found in other markets based on
different designs. Since there is an ongoing discussion on state-ownership and the introduction of an independent
system operator the north-eastern American PJM market, which is based on an ISO, was analyzed.
Results
The analysis of the bidding curves at the EEX showed big differences to the bidding curves as to expect
following the theory of bids at marginal costs. The amount of limited orders was highly restricted with a large
amount of unlimited bids on the offer (0 Euro/MWh) as on the demand side (3000 Euro/MWh). Although some
parts of the unlimited offers could be explained by the physical settlement of financial futures and energy
delivered by must-run plants or renewables that has to be sold, the percentage of sometimes 80 percent of
unlimited offers of the traded volume was quite unexpected.
Price [Euro/MWh]
100
Offer bids
Demand bids
80
60
40
20
0
0
5000
10000
15000
20000
25000
30000
Volume [MWh]
Fig. 1: Bidding curves at the European Energy Exchange
1
http://www.tu-berlin.de/~energiesysteme
Since there is no large percentage of installed demand side management systems expected to be in service, the
electricity demand is widely assumed to be not elastic. In contrast, on the EEX market there was a high amount
of limited demand bids. Our findings show that this is due to a high competition between the exchange market
and bilateral contracts. Power plants that can be controlled very dynamically such as pumped-storage plants can
place the volume sold OTC as a demand bid on the EEX in order to re-sell power from the exchange if the price
is lower instead of operating the plant. The most important point was the finding of bids placed at prices that
exceed all possible marginal costs of conventional power plants.
500
70000
450
60000
400
350
?
40000
300
250
30000
200
150
20000
Price [Euro/MWh]
Power [MW]
50000
Wind
Oil
Other
Water
Gas
Hard coal
Lignite
Nuclear
EEX Spotprice
100
10000
50
0
0
07
20
1. 07
.0 20
31 1.
.0
07
30
20
1.
.0
07
29
20
1.
.0
07
29
20
1.
.0
07
28
20
1. 07
.0 20
27 01. 007
. 2
27 1.
.0
07
26
20
1.
.0
07
25
20
1.
.0
07
24
20
1.
.0
07
24
20
1. 07
.0 20
23 01. 007
. 2
22 01.
.
07
22
20
1.
.0
07
21
20
1.
.0
20
Fig. 2: Unit commitment and price at the EEX
In addition, the unit commitment found in the data of the EEX did not resemble a unit commitment expected by
marginal cost bids. Lignite plants that normally serve base load had high dynamic commitment and the amount
of gas plants exceeded the expected amount according to the merit order model used in the agent based
approach. The offer bids at the American PJM market showed a different bidding strategy more closely related
to the expected merit order with only a low partition of offer bids at zero prices.
Conclusion and discussion
This paper describes the comparison of an empirical analysis of market data of the European electricity market
with the results from an agent-based model simulating the German power market. The main finding was that the
conventional merit order approach is not suitable for a realistic model of the market. We found several
modifications that are needed to achieve a realistic market model which are as follows:
• Small market participants without big trading departments place unlimited bids to avoid unwanted
power plant switching operations if bids are not accepted in certain hours.
• Power plants with no marginal costs such as hydro power plants without pump-storages place bids at
shadow prices depending on the reservoir level
• OTC contracts are partly hedged by additional bids on the EEX market
• Depending on gas contracts such as Take-Or-Pay clauses some gas powered plants may move lower
within the merit order leading to higher commitment rates than expected by the normal spark-spread.
• Interdepencies between the Spot-, OTC and long term financial market have to be taken into account,
different hedging strategies and risk aversions on demand and supply side can lead to large price shifts
due to a change of marketed quantities.
The agent-based system described in this paper has the flexibility to adopt to these changes. In a next step, the
mere merit-order dispatching approach will be enhanced to include these elaborated strategies and find a more
realistic description of the German electricity market. This system will allow us to move closer to the answers if
market power is applied in the German market and how the European system performs in contrast to the
American system including an ISO.