APPLICATION OF COST FUNCTIONS FOR LARGE SCALE INTEGRATION OF WIND POWER USING A MULTI-SCHEME ENSEMBLE PREDICTION TECHNIQUE - Energy Research Developments: Tidal Energy, Energy Efficiency and Solar Energy - page 357

Table 6. Optimisation Scenarios for the AESO area.

Optimisation Scenario 1 Scenario 2 Scenario 3 Scenario 4

Static Determin. Security Economic

Reserve FC Res. Reserve Reserve

[% rated cap.] [% rated cap.] [% rated cap.] [% rated cap.]

Reserve Predictor 75% reserve FC+/-11% max-min 0.7*(max-min)

Bias (FCR) 0.00 -0.85 -0.50 -0.80

MAE (FCR) 0.00 4.54 1.40 2.50

RMSE (FCR) 0.00 10.15 5.40 7.22

Correlation (FCR) 1.00 0.95 0.99 0.97

Required UpReg (7) 4.70 4.70 4.70 4.70

Required DownReg (8) 6.80 6.80 6.80 6.80

Predicted UpReg (9) 4.70 2.84 4.20 3.83

Predicted DownReg (10) 6.80 4.10 5.80 5.13

Unpredicted UpReg (7-9) 0.00 1.80 0.50 0.87

Unpredicted DownReg (8-10) 0.00 2.70 1.00 1.67

Unused Regulation 32.30 3.30 9.30 5.67

Effective cost 15.90 10.80 11.40 10.83

Hours covered by reserve 100.00 64.50 85.10 76.00

square error (RMSE), scenario 3 seems to also outperform scenario 2 and 4. However,
when comparing the unused regulation, then the security scenario has almost double the
amount of scenario 4, and three times as much unused regulation as scenario 2. Dependent
on the pricing structure of the market, which was excluded in this experiment, this could
even change the effective cost levels of the scenarios,
' The results of the Canadian example demonstrate once again that the statistical error
measures are not capable of providing a complete answer for an optimisation target. How-
ever, the results do provide an insight of the complexity of the optimisation of cost functions
of reserve to end-users requirements.

8. Summary and Discussion

The ensemble prediction method has a number of applications in wind energy integra-
tion and in energy in general. From the discussion in this chapter it can be concluded that
energy markets can benefit from using ensemble predictions. Table 7 shows a general
description of which forecasts from an ensemble forecasting system should be chosen for
minimal costs. We have assumed that the capacity of wind power is sufficient for wind to
be the “price maker” when the wind conditions are optimal. It can be seen in Table 7 that
the forecast with the lowest mean absolute error (MAE) is not always the forecast that will
generate the lowest cost of integration.
Although the original scope of ensemble prediction was to be able to conduct risk anal-
ysis of severe weather, it appears that the application in energy is not limited to grid security,
but extends to trading and management of weather dependent energy generation systems.
This includes all generation methods except nuclear power after the Kyoto protocol has be-
come effective. Nuclear power generators do not have a CO 2 problem and have therefore