Integrated forecasts are based on NWP models, ensemble analysis, smart persistence, computational learning systems, and SCADA data.
Accurate, timely energy forecasting data is essential to utility operators for maintaining reliability and minimizing wind integration costs, to wind plant owners for optimizing operations and compliance obligations, and to power traders for determining trading strategy and optimizing positions. To meet these requirements, WindLogics forecasting methodologies are based on ensemble forecasting methods and advanced computational learning systems for superior performance.
For WindLogics, accurate wind energy forecasting techniques means combining a careful analysis of numerical weather prediction (NWP) models with on-site meteorological and operational data. These data typically include power data, permanent met tower data and SCADA (Supervisory Control And Data Acquisition) data from wind plant operations. At WindLogics, these data undergo a preliminary, automated screening followed by meteorologist review prior to ingest into our forecasting system.
Our forecast methods incorporate several physics-based, industry-standard NWP models, typically Rapid Update Cycle (RUC), North American Mesoscale (NAM) and Global Forecast System (GFS). To improve forecast accuracy, we use computational learning systems to apply a sophisticated statistical analysis to these NWP models, selecting and weighting the most applicable forecast data for the timeframe of concern.
To further improve forecast accuracy, we blend forecast model data with actual wind plant data (persistence) to predict ramp events while eliminating typical model errors that may occur in first 1 to 2 hours of forecasts. Our learning system approach provides continual improvement of accuracy as the system trains the forecast based on past performance.
The results of this integrated effort include forecasts hour ahead for spot market planning or load-following resource commitment, day ahead for wind plant operations, market trading or unit-commitment, and/or week ahead for operations and maintenance planning.
Forecast capability is often measured in terms of error rates. Mean absolute errors (as a percentage of a single wind plant’s rated capacity) for the WindLogics system are generally 5–12 % for hour-ahead forecasts and 12–20 % for day-ahead forecasts. Aggregated error statistics for dispersed plants are typically lower.
WindLogics is a leader in the wind forecasting community – wind integration studies, industry-sponsored R&D and involvement with UWIG, AWEA and others. In addition, we strategically invest in our technical infrastructure, expert operations staff and dedicated research group to provide a complete wind forecasting solution. For more on how we apply these methods, please see Wind Energy Forecasting.