Smart Curtailment: Improving Efficiency by Using More Than Wind Speed

Industry members of the American Wind Energy Association have voluntarily agreed to curtail rotation of turbine blades below manufacturer cut-in speed (MCI: the wind speed at which turbines operating normally are able to send... [ view full abstract ]

Industry members of the American Wind Energy Association have voluntarily agreed to curtail rotation of turbine blades below manufacturer cut-in speed (MCI: the wind speed at which turbines operating normally are able to send power into the grid) to reduce bat mortality. In some areas, however, regulators require additional curtailment at wind speeds higher than MCI during the peak period of bat migration, further reducing bat mortality, but incurring unanticipated losses in revenue. We use data collected at a facility in the Midwestern U.S. to ask whether curtailment during this period can be relaxed under certain conditions, e.g., high precipitation, low temperatures, high humidity, dropping barometric pressure, etc., to reduce loss of revenue without compromising the reduction in bat mortality. Vegetation within a 90m circle was cleared below 15 turbines at the site. Carcass searches were conducted daily at these turbines from early April through mid-November with the objective of relating bat mortality to weather conditions. A conundrum arises because the fatality events are not directly observed hence the actual conditions under which they occur are unknown. Typically, analysis is carried out relating carcass counts to statistics summarizing the conditions of the fateful night, e.g., median wind speed or average temperature. A further conundrum arises with this approach: even if a strong relationship were discovered, because the median wind speed or average temperature of a night cannot be calculated until the night has passed, its usefulness in predictive models of mortality related to weather variables is limited. To address this, we used in our analysis only summary statistics that could be calculated within an hour of sunset to relate to the night’s fatality count. We used categorical and regression tree (CART) analysis to determine the combination of conditions under which the highest and lowest mortality occurred. We compared loss of power production under the classic scenarios of curtailing at 4, 5 and 6 mps vs under the combination of conditions identified by CART as having the highest mortality. We present results with the caveat that further study is needed at other sites in other years to ascertain the effectiveness of the CART algorithm in reducing bat mortality as well as revenue loss.