3.2 A Wind Extremes Forecast System (WEFS) for Outage Prediction

Monday, 13 January 2020: 2:15 PM
256 (Boston Convention and Exhibition Center)
Jeffrey M. Freedman, Univ. at Albany, SUNY, Albany, NY; and J. W. Zack, M. Berlinger, and C. Cheng

Under a changing climate, New York is experiencing an increase in the frequency of extreme weather events, many of which are accompanied by high winds. The rising number of extreme wind events is leading to an increase in the frequency of power outages, resulting in economic losses to utilities, businesses, and residential customers. Current power outage forecast systems have not satisfactorily addressed operational needs

Power outage forecast systems depend upon three components: 1) Numerical Weather Prediction (NWP) models that can accurately forecast the evolution of weather systems that produce (potentially) extreme wind gusts that reach the surface over varying terrain; 2) post-processing algorithms (e.g. Machine Learning, or ML) that can be trained on historical data to correct for local biases associated with surface features (e.g. terrain, land/water boundaries, surface roughness variations etc.) that are below the NWP resolution and systematic larger-scale NWP errors; and 3) observations that are dependable, representative, and of sufficient spatial and temporal density to provide a continuing depiction of atmospheric conditions and can be used to train the ML algorithm and evaluate the forecasts. Here, we present WEFS: a coupled NWP-ML forecast system that incorporates a state-of-the art 4D observation network—the New York State Mesonet (NYSM). The NYSM includes 126 surface stations, and more strategically, enhanced (profiler) sites that continuously sample the vertical profiles of temperature, humidity, and winds from the surface through the top of the atmospheric boundary layer (typically 1 – 3 km).

WEFS uses output data from a small set of Numerical Weather Prediction (NWP) systems (i.e. Global Forecast System (GFS), Rapid Refresh (RAP) model, the High Resolution Rapid Refresh (HRRR), and an in-house 1 km Weather Research and Forecasting (WRF) model run) as input to several configurations of ML algorithms and produces a final forecast from a composite of the NWP-ML ensemble members. Two nested grids (Figure 1), with the inner grid off-centered NW of the Con Ed service territory (to better capture the approach of most extreme wind features) are used. Although the initial focus of WEFS is on the Con Ed service area in southeastern New York (including New York City, Rockland, Orange and Westchester Counties), the system will be usable for any transmission service territory.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner