M. INTERPRETACION DE RESULTADOS DE CREMA PASTELERA DE
IX. ANEXOS
’s-Gravenvoeren is an H-gas exit border station in the south of the province Limburg. A total of 23 shippers operate on this network point, which together booked a total of 16,8·106 kWh per
hour exit and 1,52·106 kWh per hour entry capacity on December 31st 2014. Although gas can
only flow in the forward direction, the backward capacity is sold firm. The forward direction was totally sold out on this day. 22 of the 23 shippers allocated in more than 60% of the consecutive hours the same allocation for the period of their current booking. For these shippers it is assumed that they intentionally set their allocation for a period of time and that their behavior can be approximated with a discrete time Markov chain. For every individual shipper such a model is created. The sum of the individual results is the result of the network point as total. The single shipper that did not allocate the same amount in that much consecutive hours (0,4%) is neglected for the analysis. On its booking no oversell capacity will be offered.
0 200 400 600 800 −5 0 5 10 15 20x 10 6 ’s−Gravenvoeren December 2014 Time (hours) Energy (kWh) Allocation Exit Booking Entry Booking
Figure 9.7: The total bookings and hour allocations of shippers on ’s-Gravenvoeren in December 2014.
Allocation ’s-Gravenvoeren December 31st 2014
Confidence Max hour Oversell cap. interval allocation exit
(%) (106 kWh) (106 kWh)
99,9 14,9 1,9
99 14,0 2,8
95 13,2 3,6
90 12,9 3,9
Table 9.6: Upperbounds for the maximum hour allocation on December 31st 2014 for the network
point ’s-Gravenvoeren with corresponding oversell capacity for the one day auction that could be offered. Shippers booked a total of 16,8·106 kWh per hour capacity on this day.
For every shipper the range between its exit and entry booking is divided into 50 smaller ranges. All the allocations on hours of days with the same bookings as December 31th are rounded to the closest border of two of the smaller ranges, see figure 8.4 for an example. The borders of the ranges are the states of the Markov chain for the individual shipper. The realizations of the total period with the same booking as on December 31st are used as transition probabilities between
the states. The allocation for December 31st 2014 is simulated. One-sided confidence intervals
for the hour with the most allocation are created. The difference between the upper bound of the confidence interval for the hour with the most allocation and the bookings in exit direction can be offered for overselling in the forward direction. The results for the one day ahead auction can be seen in table 9.6. It is assumed that the behavior of a shipper is known when it has at least 7 preceding days with the same booking.
From the table 9.6 can be seen that 1,9·106 kWh energy per hour oversell capacity can be
offered at the network point in the exit direction on December 31st 2014, with 0,1% risk that in
one or more hours of the day the allocations surpass the reserved amount of 14,9·106 kWh per hour. That means that 11% of the technical capacity can be offered with that risk. Obviously, by taking more risk, more capacity can be offered. Because the maximum hour allocation on December 29thdid not exceed 80% of the technical capacity, GTS would offer 20% of the technical capacity for overselling on December 31st (see table 9.5 for the details).
The same analysis as for December 31st 2014 is done for every day of the last six months of 2014.
The results are plotted in figure 9.8. The green line, that represents the upper bound of the 99% confidence interval for the maximum hour allocation, fluctuates heavily sometimes. The main reason for that, besides the fluctuation in allocation, is the appearance of new bookings. For the shippers that change their booking is after 7 days assumed to be known what their behavior is. Up to that time no oversell capacity is offered on their bookings. Their total individual booking is then reserved for them for the next day. The same holds for day bookings. When a shipper does a day booking on top of its month, quarter or year booking, its behavior is unknown. Their total bookings, the original bookings plus the day bookings, will be reserved for them. It is probably not realistic to assume that when a shipper does a day booking on top of its original booking, its behavior over the original booking is the same as without an additional booking, because the orig- inal booking does not suffice anymore. The oversell capacity that can be offered with 1% risk in forward direction is the difference between the red and the green line in the upper plot of figure 9.8. In almost none of the hours of the last six months of 2014 was the total booking of all the shippers at ’s-Gravenvoeren higher than 80% of the the technical capacity. GTS would have offered 20% of the technical capacity for overselling on nearly all days, which is 3,36·106 kWh per hour. By
taking 1% risk, the capacity that can be offered for overselling fluctuates between 0% and 40% of the technical capacity. At the time of around 800 hours some new month bookings are done. This can immediately be seen the oversell capacity, which is almost zero at that point and a week
afterwards. 0 1000 2000 3000 4000 5000 −1 0 1 2x 10 7 Time (hours) Energy (kWh) ’s−Gravenvoeren Allocation Exit Booking Entry Booking 1% risk 80% Tech. Cap. 0 1000 2000 3000 4000 5000 0 0.5 1 1.5 2x 10 7 Time (hours) Energy (kWh) Exit Booking Overs. cap. 1% GTS overs. cap.
Figure 9.8: The total bookings and allocations of all the shippers in the last six months of 2014, together with a the upper bound of a one-sided 99% confidence interval for the maximum hour allocation on a day in the upper plot. In the lower plot the corresponding oversell capacity that can be offered at the one day ahead auction by GTS and by taking 1% risk.
10
Conclusions, recommendations and remarks
The end of this report is a brief retrospect of what was achieved and where there is room for further improvement. First the conclusions are presented and then some recommendations are made.