92nd American Meteorological Society Annual Meeting (January 22-26, 2012)

Monday, 23 January 2012
Economic Evaluation of Short-Term Wind Forecasts in ERCOT: Preliminary Results
Hall E (New Orleans Convention Center )
Kirsten Orwig, National Renewable Energy Laboratory, Golden, CO; and V. Banunarayanan, S. Nasir, J. M. Freedman, and M. Milligan

Economic Evaluation of Short-Term Wind Forecasts in ERCOT: Preliminary Results Kirsten Orwig, Venkat Banunarayanan, Saleh Nasir, Michael Milligan, Jeff Freedman Historically, a number of wind energy integration studies have investigated the value of using day-ahead wind power forecasts for grid operational decisions. These studies have all shown that there could be large cost-savings for grid operators when implementing these forecasts. To date, none of these studies have investigated the value of shorter-term (0-6 hour ahead) wind power forecasts. In 2010, DOE and NOAA partnered up to fund improvements to short-term wind forecasts and to determine the economic value of these improvements to grid operators, hereafter referred to as the Wind Forecasting Improvement Project (WFIP). This paper and presentation will discuss the preliminary results of the economic benefits of WFIP for the Electric Reliability Council of Texas (ERCOT). This is a supplemental paper to Jim Wilczak et al. and Jeff Freedman et al. who will be presenting on an overview of the DOE/NOAA project mission and an overview of the AWS Truepower/ERCOT project, respectively. The economic evaluation will be performed by: 1) developing new metrics to measure forecast error, 2) evaluating the overall forecast and ramp event forecast improvements, 3) determining deviations between scheduled and delivered energy and the associated costs, 4) investigating the benefits of selling wind energy in hour-ahead or day-ahead markets compared to real-time markets, and 5) determining the reduction in ancillary service costs (i.e. by reducing reserve and regulation requirements). Traditionally, Mean Absolute Error (MAE) and Root Mean Square Error (RSME) have been widely used and accepted by the forecasting industry as ways to quantify forecast error. Therefore, one approach to quantify the economic value of the forecast and any improvements would be to use a metric such as $/%MAE or $/%RSME. Additionally, ramp error metrics will be defined, such as hit/miss, ramp error, magnitude error, and phase error, and these metrics will be adapted to quantify the cost of such errors. Preliminary results of these metrics for all forecasts will be presented. To ascertain the costs associated with the various forecast approaches, a production cost model will be run to simulate the ERCOT system and their operational decision strategy. The base case will employ a security-constrained unit commitment and dispatch market for the existing forecasting technology. Subsequent cases will be run with the modified/improved forecasts. It is expected that unit commitment and dispatch, as well as ancillary service allocation will be better optimized with the new forecasts. In addition to the production cost modeling, the team will use of intra-day and intra-hour models to assess the impact of improved ramp rate forecasts on operating reserve requirements. The results of the modeling will be compared across all cases to determine the overall value of the new forecasts, and preliminary results will be presented.

Supplementary URL: