EU Energy Policy Blog

In a recent study on the impact of long term call auctioned by EDF on the French day-ahead market we found some quite unexpected results when we analysed the aggregated curves of offers to buy & sell power.

To introduce more competition into the French electricity market the European Union constrained the historic operator EDF to provide power to others through long-term call options called VPP (short for virtual power plants). VPP holders have the right to access to power at a predetermined strike price per MWh 24 hours per day, 7 days per week during the delivery period, after having paid an upfront premium.

In a recent paper we propose to model the impact of VPPs on the day-ahead market as a two-player game between the historic operator and the VPP holders (taken as a group), in which the VPP holders offer any unwanted power for sale on the day-ahead market if the price exceeds the strike of the VPP, and conversely, buy power that they want on the market if the price is below the strike. Consider a VPP with a strike price K that allows the new entrants to access a quantity V of power. Suppose that for a given 1 hour period the new entrants do not require all of V for their retail clients. They will send two orders to EpexSpot: one to buy the quantity required if the price is below K, and another to sell the unwanted quantity if the price is above K. These quantities will be referred to as BuyBelow and SellAbove.

If our hypothesis is correct, the aggregate curves will consist of the offers from VPP holders together with the offers from producers and consumers. The offers from VPP holders should be in a narrow slice on either side of the VPP strike price during the delivery period, but not at other times or for other prices. Offers from other buyers and sellers will occur at random in the intervals. To test this we consider baseload VPP in 2006 because their strike price was raised from 8 to 9 euro on 1 October. Figure 1 presents the amount offered for purchase at prices between 8.50 euro & 9.00 euro during 2006 for the 12th hour H12, (left) and the amounts offered for sale in the price range from 9.00 to 9.50 euro (right). The change from randomness to structure in 1 October coincides with the date when the VPP with a 9 euro strike price start to be delivered. Then we test the model on three peakload VPP with strikes of 61, 63 and 64 euro, and with three prices (60, 62 and 65 euro) which do not correspond to VPP strikes. These results also confirm our hypothesis.

Figure 1: Amount of power in the classes BuyBelow (above), & SellAbove (below), for the 9 euro reference price for H12 during 2006. Note the change in shape on 1 October when the VPP strike rose from 8 euro to 9 euros.

In addition to the amounts BuyBelow and Sell Above, we also computed the amounts BuyAbove and SellBelow. According to our model, these should be white noise. Much to our surprise they sometimes mirror the structure in the other two classes. Figure 2 presents the amounts BuyAbove and SellBelow as a function of time for H12. As before a marked change occurs on 1 October. The same occurred for some peakload VPP too. This raises two questions: Who could be selling at that price, and why?

Figure 2: Amount of power in the classes BuyAbove (above), & SellBelow (below), for the 9 euro reference price for H12 during 2006. We had expected these to be white noise but much to our surprise, the shapes are very similar to those given above

We are perplexed. If anyone has any ideas we would be very interested to hear from them…

Margaret Armstrong and Alain Galli, Mines PariTech

This entry was posted on Tuesday, June 1st, 2010 at 1:15 pm and is filed under Electricity. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.