actually, weather forecast is based in a numercial approach that take
Transcripción
actually, weather forecast is based in a numercial approach that take
Numerical Weather Prediction Services > Wind Forecast Ankara, 5th march, 2010 Weather forecast is no longer an empirical knowledge,… actually, weather forecast is based in a numercial approach that takes advantage of atmospheric science, met station data, satellite images and automatic self- training. 1 Can we anticipate the future ? Yes, with physical laws of the Yes atmosphere: with initial conditions, we can predict future conditions. We need atmospheric data through a 3D grid: id over the th entire ti globe l b ! 2 3 4 NWP: it is a world wide tool. Numerical Weather Prediction Services DIFERENTS ESCALES 5 INITIAL DATA NEEDED • SAT • Radar • Buoys • Rawsonda • Met station data 6 Wind forecast: ensemble techniques Global model forecast (GFS, ECMWF) - NWS FTP UPLOAD FTP 2-10 Mbytes ZIP 2 Mbits connection Met Tower Production Availability FTP NATIONAL WEATHER SERVICES Barcelona GRIBGRID FILES – 2 GBytes Mesoscale M l model d l fforecastt (MASS + WRF + ARPS) METEOSIM HEADQUARTERS UPLOAD TO CONTROL ROOM CONTROL ROOM WSTATION FORECASTS-VALIDATION DATA RESULTS WEB MET TOWER PRODUCTION AVAILABILITY METEOROLOGIST TRADING DEPARTMENT SCADA Ankara CONTROL ROOM WStation CONTROL ROOM - Backup SSH METEOSIM TRUEWIND TECHNICAL DEPARTMENT Turkey WEB MAPS WEB Barcelona 7 Ski resorts Numerical Weather Prediction Services Lon term forecasts. Numerical Weather Prediction Services 8 … and Real adjustment: Numerical Weather Prediction Services Sailing forecasts: tool decision based on wind Numerical Weather Prediction Services 9 Wind forecasts onshore and offshore. Numerical Weather Prediction Services Automatic procedures: how to use numerical weather forecasts ? Numerical Weather Prediction Services 10 PREVISIONES NÁUTICAS: http://sail.meteosim.com Numerical Weather Prediction Services Air quality control: services based in wind forecasts. Los modelos de dispersión analizan el terreno en alta resolución. Numerical Weather Prediction Services 11 Numerical forecasts can provide graphical plus tabular results. Numerical Weather Prediction Services Fire risk based in NWP. Fire Weather Index Numerical Weather Prediction Services 12 Media forecasts… Numerical Weather Prediction Services SCREEN… Numerical Weather Prediction Services 13 … and others: take care with quality Numerical Weather Prediction Services Final weather forecasts seems the similar ! Numerical Weather Prediction Services 14 Wind forecasts is applied in forensic analysis. Numerical Weather Prediction Services Wind wave forecasts… Numerical Weather Prediction Services 15 Even rainfall forecasts: not the same error Ejemplo de aplicación de modelos de simulación numérica para previsión de lluvias en puntos concretos. Numerical Weather Prediction Services PREDICCIONES PERSONALIZADAS: modelización en todo el mundo a las resoluciones necesarias Numerical Weather Prediction Services 16 Samples: climatic database managment. See: http://www.meteo.ad Numerical Weather Prediction Services SERVICIOS GENERALES Numerosos y variados clientes requieren boletines de predicción objetivos y sintéticos. La toma de decisiones debe partir de informaciones objetivas, por expertos p y con la elaboradas p información justa y necesaria. Numerical Weather Prediction Services 17 Statistics model Real data of wind and production are used to adjust computational forecasts. forecasts • MOS (Mean Output Statistic): empirical relation between historical data and forecasts parameters • Real time data are used to adjust final forecasts. Predict ors P1 ,P2 ,... SMLR ANN SVM Predict and F Training Algorithm F = f ( P1 ,P2 ,...) 18 Modelo estadístico Different regional models Wind speed Forecasts vs. Real Production 19 Modelo de validación: Typical forecasts adjustment (Single Wind Plant, 1h time step) • Absolute mean error increase quickly q y during g first 6 hours. • Staistical NWP adjusted it’s better than climatological values during 5 – 7 days. • Wind farm data required to improve persistence based forecasts 25% 20% 15% 10% 5% 0% 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 Forecast Time Horizon (Hours) 40% 35% 30% Persistence Climatology 25% 20% 15% Physical Model with MOS (no real-time plant data) 10% 5% Short-term statistical model (with real-time plant data) 0% Time (Hours) 20 THANKS ! Sr. Santi Parés Director Comercial Tel. +34 93 4487265, [email protected] 08028 Barcelona SPAIN 21