Customization of a Mesoscale Numerical Weather Prediction System

Customization of a Mesoscale Numerical Weather Prediction System for Industrial Site Operations Anthony Praino & Lloyd Treinish Deep Computing Systems/Big Green Innovations IBM Thomas J. Watson Research Center Yorktown Heights, NY http: //www. research. ibm. com/weather/DT. html

IBM Research – Big Green Innovations Ÿ Model configuration Ÿ Weather sensitive industrial site operations Ÿ Current status Ÿ Example products and applications Ÿ Summary and future work

Deep Thunder -- Model-Based Forecasts for Local Weather-Sensitive Operations Ÿ Enable proactive decision making affected by weather Ÿ Customize & integrate for different users Ÿ Provide usable forecast products fast enough to enable timely decisions Weather Data Ingest Local Weather Prediction p. Server Model Observation s Advanced Visualization Ÿ Visualized results produced Ÿ Identification of time within a few hours per day of forecast Ÿ Couple to business processes & models: Load forecasting, dispersion, road surface, groundwater modelling, . . . Ÿ Past forecasts useful for scenario planning Usable Graphical Products "Business" Processing "Business" Data Ingest Ÿ Demographics Ÿ History Ÿ Dispersion Modelling Ÿ Road Surface Modelling Ÿ Load Forecasting Ÿ Analytics and location of "events" Ÿ Analysis of impact on production Ÿ Expense vs. effectiveness of countermeasures for severe weather

Deep Thunder Modelling Geographies §Operational Locations – Austin-San Antonio, Atlanta, Baltimore-Washington D. C. , – Chicago, Kansas City, Miami, New York, Orlando, Philadelphia – Vermont-New Hampshire –Experimental Run Locations – Albany–Springfield, North Carolina, San Diego, San Francisco, Long Beach

Forecast Model Description • Deep Thunder: Highly Customized Version of RAMS • 24 Hour Forecast Period • Full Three Dimensional Non-Hydrostatic – 35 Vertical Levels • Triple Nested Configuration – 16 km, 4 km, 1 km – New York – 27 km, 9 km, 3 km – New England • Two Way Interactive Domains • Explicit Cloud Microphysics

IBM Site Operations Locations Burlington Center of Industrial Excellence - Essex Junction, VT East Fishkill, NY manufacturing plant site

Weather Sensitive Industrial Site Operations Ÿ Building Energy Management ŸOptimized use of electrical power and fuel sources and water – Daily accurate site specific forecasts could ensure that all of the site chillers/boilers are available and/or running before peak load – Forecast of afternoon conditions could enable operations personnel to shutdown unneeded equipment early and save energy – Accurate site forecast enables maximizing energy efficient operations by the use of free cooling with economiser controls or the use of flat plate heat exchangers – Energy/Cooling Storage Applications – Peak load shedding or avoidance strategies – Smart BMS applications – Alternative energy generation strategies

Weather Sensitive Industrial Site Operations ŸSafety –High winds/lightening can affect certain outdoor activities such as roof work, scaffold work, window cleaning or general construction –Heavy rain, ice, snow could contribute to accidents on site roads, parking lots and sidewalks ŸSite Operations – Weather sensitive manufacturing & development operations –Outdoor events and grounds work are affected by local weather conditions Employee work scheduling (arrival, departure, remote access) –Improved efficiency and effectiveness for cold season operations –(reduced cost and environmental impact) – Integrated supply chain management & logistics – IT operations management (Green datacenter) ŸSite Security –Infrastructure and asset protection have weather sensitivities

Center of Industrial Excellence Site Operations (Essex Junction, VT) • • • Operation of Chilled Water, Glycol Chilled Water and High Temp Hot Water System Building Heating / Cooling Optimization Load Response / Peak Shaving Program Emergency Response Team’s Early Warning System Site Electrical Single Source Monthly Emergency Generator Testing

Weather Sensitive Industrial Site Operations Ÿ Site Specific Forecasts –Burlington, East Fishkill, Hawthorne, Northcastle, Palisades, Poughkeepsie –Somers, Sterling Forest, Yorktown Ÿ Installation of IBM Weather Station Network –Provide current weather conditions for: site operations, forecast validation and improvement –Operational: Somers, Yorktown –Pending Installation: Burlington, East Fishkill, Poughkeepsie Ÿ Building Energy Management –Automated system creates and emails recommendations for plant operations based upon site specific forecast for each cycle Ÿ Site Operations –Burlington, East Fishkill, Somers, Yorktown Heights

Site Specific Forecast Example

Site Operations Forecast Data IBM East Fishkill [41. 5770 N, -73. 7942 W]

Site Operations Procedures Example • • • Somers Site Operations Actions based upon Deep Thunder Forecast Recommended Actions for Sunday - Friday Operations Valid Date: 200830806 • • Temperature is forecasted to be less than 60 F for more than 2 hours Run 1 chiller if CWR > 50 F Run 1 boiler before 8 AM Run 1 HW radiation pump before 6 AM • • • Humidity is forecasted to be greater than 50 percent for more than 2 hours Cooling Towers: Maintain CWR temperature Air Handler: Disable Economizer • • • Forecast Statistics High Temp: 53. 25219 Low Temp: 29. 85657 High Humidity: 99. 74673 Low Humidity: 73. 46073 Max Wind Speed: 6. 607601

Summary Ÿ Focus on IBM site operations –Multiple industrial site locations –Location and applications specific focus –Customization –Benefits and integration for site operations and business applications for environmental impact, energy use, safety, cost Ÿ Continued research and development –Multi-site operations –Additional customization and integration to support operational applications –Operational modes (standard, seasonal, storm, on-demand) –Better forecast validation using IBM mesonet data –Assimilation of IBM Mesonet Data For Improved Forecasts –Past results database for site climate studies –Additional visualization and dissemination techniques Ÿ Future work –Location specific customizations and integration –Improved computational performance and throughput –Extensions to other models and data products –Customized interfaces, products and packaging for new internal applications (e. g. , manufacturing, logistics, IT, ISC, environmental programs )