At great expense, balancing authorities are staffing wind forecast desks 24/7/365, structuring financial hedges, and adopting persistent wind curtailment strategies to mitigate or even eliminate the risk. To lower these costs and drive efficiency, better energy balancing decisions can be made by considering the actual ramp event risk in tandem with the external market signals and transmission constraints. Guidance on wind power ramp events is needed in probabilistic form in order to quantify the forecast uncertainty and afford the user the opportunity to take action only when the estimated risk is above the user's tolerance for it.
Specific examples are shown of tuning a best-estimate forecast based on an underlying forecast ensemble of wind power ramps at a U.S. wind farm over a several month period. The tradeoffs between frequency bias, hit rate, false alarm rate, miss rate, and misclassification rate are examined by maximizing contingency table scores such as the equitable threat score and the Gerrity Skill Score. In the end, the metric to be optimized depends on each user's cost for missed events and tolerance for false alarms, which likely varies substantially by market and power system conditions.