NCAR is currently performing work that involves forecasting the power production at a variety of wind farms based on forecasted winds at each of the farms. As part of this work, different methods for converting forecasted winds to power were researched, with methods ranging from using the turbine manufacturer's power curve to creating static data mining conversion models from turbine level wind to turbine power. Preliminary results showed that using static data mining models for the wind to power conversion improves the accuracy of the power forecasts over using the turbine manufacturer's power curve. With the goal of moving to a fully automated forecasting system, new research is being done to be able to do the wind to power conversion dynamically, with the models being created routinely as part of the forecast system from recent turbine level data. Dynamic model creation presents more than engineering challenges, however, since data quality and turbine availability issues must be identified and handled automatically. This paper discusses our research into dynamically creating models to convert wind to power, including the challenges of identifying and handling data quality and turbine availability issues in real time.