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 352

Table 2. Results of the 3 scenarios of using and not using spinning reserve predictions
in the wind power production forecasts for the western part of Denmark in the period
January to April 2005. All statistical quantities are given as [%] of installed wind
power capacity of Western Denmark

Scenario no. 1 2 3

EPS configuration optimal* optimal average

Number of Hours 2331 2331 2331

Forecasted Mean 27.01 26.6 26.7

Bias -0.77 -1.19 -1.02

Mean Absolute Error 2.72 5.30 5.83

Stdev Error 5.39 7.60 8.32

Correlation 0.977 0.953 0.946

It can be seen in Table 2 that (scenario no. 1) using the optimal forecast and reserve
prediction for next day's wind generation almost halves the mean absolute error (MAE) of
both scenario no. 2 and scenario no. 3 and hence the costs for expensive spinning reserve for
unexpected events (errors). The root mean square scores are slightly lower, but still of the
order 30% and 40% better for scenario 1 than for scenario 2 and 3, respectively. The lower
improvement in the root mean square (RMSE) is related to the fact that the prediction of the
forecast error removes the smaller errors in a band-like way around the optimal forecast.
Since the RMSE is more sensitive to the larger errors, the improvement is lower.
Table 3 shows the percentage of reserve of installed capacity that is required, predicted
and non-predicted in scenario 1. It can be observed that the predicted up-regulation pro-
vided better results than the down-regulation. This is the result of optimising the forecast
towards the less costly reserve.

Table 3. Results of the verification of the predicted regulative power on a long-term
basis for the Western part of Denmark in the period January to April 2005. The
regulation magnitude is given as % of installed wind power capacity. The results
correspond to scenario no. 1.

Regulation Regulation description

type magnitude

Average Up 2.40 required regulation

Average Down 3.43 required regulation

Predicted Up 1.43 traded day ahead

Predicted Down 1.68 traded day ahead

Unpredicted Up 0.98 traded on the day or handled by flexible contracts

Unpredicted Down 1.75 traded on the day or handled by flexible contracts

Unused 0.91 unnecessary regulation capacity