3. Quantitative Analysis
3.1. Modelling, assumptions and limitations
The quantitative analysis relies on market simulations which help to support some conclusions of the study.
However the modelling relies on assumptions and has some limitations; which makes it difficult to derive direct and definite conclusions from raw numerical results.
The purpose of this chapter is to explain why numerical results should be considered carefully and to show the consequences of modelling assumptions.
Detailed quantitative results including all technical details related to modelling assumptions and limitations can be found in Appendix III.
3.1.1. Net coupling welfare
In chapter 2 of this report it was explained which aspects of the welfare can be modelled in the market coupling. The marginal total welfare gain is assumed to be adequately represented by the price difference in the market coupling. Of the marginal total welfare losses induced by the exchanges only those that are induced by losses on DC cables were included in the market simulations and respective calculations.
The welfare effect that is calculated from the simulations is:
{ ∑ ( ) ( )
}
{
∑
}
Where the producer and consumer surplus are calculated from the supply and demand curves in the order books and the market clearing prices.
This is called the net coupling welfare.
3.1.2. Gross and Net Congestion Rent
The second line of the formula in 3.1.1 represents the congestion income collected from market coupling.
This part is called gross congestion rent throughout this report. Note that for interconnectors with losses included in the allocation, the difference between sending end and receiving end volumes are the losses that are included. Because the included losses are added to the system balance constraint, the impact of
11
these losses on the producer surplus is fully taken into account. The second and third line of the formula together are called the Net Congestion Rent.
Gross congestion rents are not comparable between the runs as they contain for each run to a different extent DC cable losses that are procured within the market coupling. Only the Net Congestion Rents are comparable between the runs .
The third line of the formula in 3.1.1 refers to the costs of the losses which are not implicitly procured at the PX through a loss factor. These costs are a welfare loss that is not taken into account in the welfare as calculated by the market coupling algorithm, irrespective if these losses are procured explicitly on a PX (through a demand order) or bilaterally outside the PX (See Appendix VI –(D)).
Marginal total welfare losses induced by exchanges inside the AC network or on AC interconnectors were not included in the simulations. If in practice these would be in absolute value larger than the marginal welfare loss from the losses on the DC cables, the optimality condition for inclusion of a loss factor is not fulfilled. In this case it would not be valid to make any conclusions on total welfare effect from the Market coupling results. In the same case total welfare is likely to be decreased if loss factors on DC cables were included even if the net market coupling result would show an increase.
3.1.3. Simulations overview
Period of simulations and market data
Simulations cover full year 2011; results are available for 363 days (8712 hours)3. Market data are historical data from PXs order books. Network data are historical ATCs and ramping limits (except when losses apply).
Network and perimeter
The network is based on ATC interconnection (no flow-based); no tariff applied. Losses are applied only for some cables (see below). The perimeter covers the NWE bidding areas (including PL and Baltic areas).
List of Runs
No loss is applied on AC interconnectors for any run.
• Run #1 – No losses in the market coupling at all (loss factors applied in Run#3 are used to calculate external losses costs) - The output is the reference result in terms of welfare, prices and flow pattern
• Run #2 – Equal Loss Factor in the allocation on all existing DC cables (harmonized case)
• Run #3 – Individual Loss Factor in the allocation on all existing DC cables – These loss factors are assumed to be the actual loss factors which perfectly reflect the losses on the interconnectors
• Run #4 – Individual Loss Factor in the allocation on some DC cables (BritNed, IFA and Baltic)
• Run #5 – Equal Loss Factor in the allocation on some DC cables (BritNed, IFA and Baltic)
3 The inclusion of the ramping constraint with the flow of last hour previous day made two sessions fail, so that results were available for 363 days (8712 hours) only.
12
For all runs the costs of the losses which are not included in the market coupling are based on the difference between the actual loss factor in Run#3 and the loss factor which is included in the current run.
This is elaborated in Appendix V. under section a.
The only difference between the 5 runs is the modification of DC loss factors which are included in the algorithm. Every other characteristic (e.g. input data, algorithm parameters, network topology for each day) is identical for all runs4.
3.1.4. “Sending end” versus “Receiving end”: alterations of ATCs and ramping limits due to losses Since losses result in a lower flow at the receiving end of the cable than at the sending end of the cable, two options are possible when loss factors apply:
“sending end”
• The historical ATC is considered as the sending end ATC. Therefore the receiving end ATC is lower when losses apply.
Example: Baltic 610MW at sending end results into 595MW at receiving end when a 2.4% loss factor applies.
“receiving end”
• The historical ATC is considered as the receiving end ATC. Therefore the sending end ATC is higher when losses apply.
Example: NorNed 700MW at receiving end results into 729MW at sending end when a 4% loss factor applies.
4 Though being an input for a given day, the flow of last hour previous day through each interconnection with ramping constraint is an output of the day before and therefore can be different for each run.
13
The following DC interconnectors are modelled under the sending end option: Baltic, BritNed, and IFA5. The other DC interconnectors with losses are modelled under the receiving end option.
3.1.5. Ramping constraints
The following DC interconnectors are subject to a ramping constraint of 600MW6: NorNed; Storebaelt; Skagerak; Kontek; Kontiskan; Baltic; Swepol.
3.1.6. Topology description including SE splitting
The topology of the network takes into account the splitting of SE into 4 bidding areas after Nov 1st. Until Oct 31, the topology includes:
•
SEA virtual bidding area;•
SE is a single bidding area, with one single connection to FI in production, aggregating the DC line between SE and FI and the AC interconnection between SE and FI in the north7;The modelling of this topology in the frame of the simulations does not exactly correspond to the historical modelling in production as regards the parallel interconnections between SE and FI. Therefore corresponding results should not be considered as historical results, even for Run#1, but only as possible results if such a configuration were implemented.
After Nov 1st, the topology has changed:
• SEA no longer exists;
• SE has been split into SE1/SE2/SE3/SE4, so that there exists one SE3-FI Fennoskan DC interconnector and one SE1-FI AC interconnector;
Therefore yearly total indicators should not be compared to production yearly totals; the indicators related to these recent bidding areas and corresponding interconnections only concerns two months of simulations (61 days; 1464 hours).
Similarly, indicators related to the “old” topology are calculated and available only for 302 days (7248 hours).
The quantitative analysis always relies on comparisons between runs; no comparison between these different topologies can be envisaged or deduced from the results and such a comparison was never seen as a possible objective of the simulations.
5 IFA sending end ATCs are re-calculated from mid-channel reference – see Appendix VII.
6 Maximum variation (increase or decrease) of flow between two consecutive hours.
7 The modelling of this configuration is implemented by means of a virtual bidding area between SE and FI – see Appendix VII.
14