IBM Big Green Innovations Environmental R&D and Services Smart Weather Modelling Local Area Precision Forecasting for Weather-Sensitive Business Operations (e.g. Smart Grids) Lloyd A. Treinish Project Scientist, Big Green Innovations IBM-USA http://www.ibm.com/green Helmut Ludwar Chief Technologist IBM-Austria 2010 IBM Corporation
Classes of Weather Models for Different Geographic Scales, Time Ranges & Applications US National Weather Service uses a synoptic-scale model on a 12 km grid covering North America and more, but only a subset are available on three-hour intervals 40% physics, 50% dynamics, 10% microphysics Thunderstorms wind shear, land-sea breezes, circulation induced from topography cannot be seen at this resolution Like trying to catch small fish with a net that has a large mesh Regional models employ meso- and cloud-scale grids (e.g., 1-10 km spacing), including nesting 50% physics, 20% dynamics, 30% explicit cloud microphysics Can resolve storms, winds, etc. Cast a finer net over coarse net
Deep Thunder Forecasts for Weather-Sensitive Operations Problem: weather-sensitive business operations are often reactive to short-term (3 to 36 hours), local conditions (city, county, state) due to unavailability of appropriate predicted data at this scale Energy, transportation, agriculture, insurance, broadcasting, sports, entertainment, tourism, construction, communications, emergency planning and security warnings Solution: application of reliable, affordable, weather models for predictive & proactive decision making & operational planning Numerical weather forecasts coupled to business processes Products and operations customized to business problems Competitive advantage -- efficiency, safety, security and economic & societal benefit
Approach Testbed implementation for multiple metropolitan areas Deep Thunder End-to-end process (user to meteorology) tailored to business needs Operational infrastructure and automation with focus on HPC, visualization, and system and user integration Operational 24-hour forecasts to 1-2 km resolution Experimental 84-hour forecasts at 2 km resolution New York City, Chicago, Kansas City, Baltimore/Washington, Atlanta, San Diego, Fort Lauderdale/Miami and others Business applications with actual end users to address usability and effectiveness issues It is not about weather but integrating forecasts into decision making to optimize business processes
Short-Term Weather Event Prediction and Observation Forecasting (Modelling) Nowcasting (Sensors) NWS / Commercial Providers Forecast for longerterm planning where decisions require days of lead time, but may not have direct coupling to business processes Deep Thunder Forecast for assetbased decisions to manage weather event, pre-stage resources and labor proactively Remote Fine-tune approach based upon extrapolation from Doppler radar and satellite observations In Situ Near-real time revision Continental Scale Local Scale Local Scale 84-168 18-72 3 0 Time Horizon for a Local Weather Event (Hours of Lead Time)
Development and Deployment of a Weather and Outage Prediction Service for Electric Utility Operations The operation of the distribution system of an electric utility, particularly with an overhead infrastructure, can be highly sensitive to local weather conditions What is the potential to enable proactive allocation and deployment of resources (people and equipment) to minimize time for restoration? Ability to predict specific events or combination of weather conditions that can disrupt that distribution network with sufficient spatial and temporal precision, and lead time Can highly localized, model-based forecasts (e.g., IBM Research Deep Thunder ) be adapted to address these problems and reduce the uncertainty in decision making?
Storm Impact and Response Prediction (ConEd) Weather causes damage Damage require restoration (resources) Restoration takes time, people, etc. Build predictive model from environmental observations, storm impact and related data Weather predictions Damage prediction Damage location, timing and response Wind, rain, lightning and duration Demographics of effected area Infrastructure impacted in effected area Ancillary environmental conditions Couple this model to weather predictions to enable a forecast of impact and response Resource requirement prediction Restoration time prediction
Challenges of Building an Impact Model Damage forecast model inputs Which weather inputs are important for damage forecast? Most weather variables are correlated Multicollinearity may cause invalid interpretation of weather predictors Weather forecast calibration Forecasted variables (e.g., wind speed) may differ in meaning vs. observations used in the damage-forecast-model training How should physical model outputs be calibrated so that they can be used as the inputs of damage forecast model? Gust speed calculation Exploratory data analysis indicates that gust speed has a stronger relationship to damages vs. wind speed General meteorological models do not provide a direct gust forecast How should gust speed be calculated based on limited weather data? Model integration How should damage forecasts, multiple spatial resolution interpolations and calibration be integrated in one framework? Uncertainty quantification and visualization Uncertainties come from various data sources
13 March 2010 Rain and Wind Storm Coastal storm with strong winds and heavy rains Gusting between 40 and 75 mph observed in the afternoon and evening Innumerable downed trees and power lines Local flooding and evacuations Electricity service lost to over 600,000 residences and businesses in the New York City metropolitan area Widespread disruption of transportation systems (e.g., road and bridge closures, airport and rail delays) Other forecasts during the late morning on 11 March: rain may be heavy at times, east winds 20 to 25 mph with gusts up to 40 mph Wind advisories issued (gusts to 45-50 mph) at 1619 EST, 12 March High wind warnings issued (gusts to 55-60 mph) at 1343 EST, 13 March
13 March 2010 Nor easter Deep Thunder Weather Forecast Initiated with data from 1900 EST on 3/10 with results available after midnight on 3/11. High winds shown in forecast available more than two days before the event and 37 hours before the National Weather Service advisory
13 March 2010 Nor easter Deep Thunder Impact Forecast Actual Outages (Repair Jobs) Estimated Outages (Repair Jobs)
Web Interface for Consolidated Edison Site-Specific Forecast Plots Surface Precipitation Animation
Wind and Solar Energy Issues
Challenges for Wind and Solar Power Wind and solar power intermittency creates significant barriers to expanding utilization Consider for wind Ramp events Spinning reserve Better forecasting and smarter dispatch can alleviate these barriers Ensemble forecasts Stochastic programming Dynamic reserves
Example Wind Farm Forecast for a Ramping Event 15 84-hour wind forecast showing speed and direction at a height of 10m and at the location of a hypothetical wind farm off the coast of New York City along the blade extent as well as wind trajectories
Example Wind Farm Forecast for a Ramping Event 16 84-hour wind forecast showing speed and direction at a height of 80m (top, i.e., hub height for turbines at the location of a hypothetical wind farm off the coast of New York City) and at height of 10m (bottom) along the blade extent as well as wind trajectories Peak wind speed before the ramp down is close to the cut-out speed of some turbines There is not lot of shearing from the 10m to 80m because the offshore flow is relatively laminar
17 Wind Farm Siting Configure and run a high-resolution weather model over the target area to create a climatology to determine the power potential Match weather model to the wind farm scale (horizontally, vertically, temporally) Include local, detailed geography such as terrain, vegetation, etc. Couple result to power estimation, includes assessment of variability Evaluate more than the maximum wind If there is too much variability or a sufficiently large variation between the surface and the turbine height, then the equipment will be subject to stress and fail relatively quickly Consider optimization driven by the physical constraints as well as economic and logistical ones Assess environmental impact of the turbine operations In aggregate, a large wind farm does take energy out of the atmosphere Estimate changes in local weather to a climate scale due to wind farm operations It is not about wind, but rather the nature of the weather in general (e.g., wind, storm intensity and duration, ground saturation, etc.), which you have to do correctly at the appropriate scale in order to determine wind correctly
Example Site-Specific Cloud Forecast (1 km Resolution)
19 Solar Farm Siting Configure and run a high-resolution weather model over the target area to create a climatology to determine the power potential Match weather model to the solar farm scale (horizontally, vertically, temporally) Include local, detailed geography such as terrain, vegetation, etc. Couple result to power estimation, includes assessment of variability Evaluate more than the maximum solar irradiance If there is too much variability for integration into a grid or operations Recognize three components: global horizontal, direct normal and diffuse Understand efficiency impacts due to temperature variation Determine vibrations from wind affecting focus, and wind-induced ventilation removing heat Assess environmental impact of the solar farm operations In aggregate, a large solar farm changes the surface albedo and increases turbulence in the boundary layer Estimate changes in local weather to a climate scale due to solar farm operations Nature of the weather in general (e.g., solar irradiance, moisture and clouds, storm intensity and duration, ground saturation, wind, etc.), must be done correctly at the appropriate scale
20 Example Operational Applications Accurate timing of shut down due to severe weather Lead time required to take preventative measures (e.g., when a storm hits the turbines must be shut down, which takes several minutes) The shut down causes a loss of energy generation The more precise the weather forecasting the more optimal can the time of shutdown be determined to minimize the loss Accurate reserve margins With improved power estimations and when losses would occur, the need for alternative sources (e.g., fossil fuel plants) can be determined better More cost effective management of all generators as the need for high margins on power reserves and standby production will decrease Accurate area for shut down Improved weather forecasting will allow for a more limited shut down of the facility since a subset of the effected areas could be determined Overall reduction in variability into the grid
21 Thank You