Increasing the Value of Weather Information in the Operation of the Electric Power System: Workshop Report

Increasing the Value of Weather Information
in the Operation of the Electric Power System

Environmental and Societal Impacts Group
National Center for Atmospheric Research

*The National Center for Atmospheric Research is sponsored by the National Science Foundation.

Hackney, J., 2003: Increasing the Value of Weather Information in the Operation of the Electric Power System. Report of Workshop held 6-7 November 2002. Boulder, CO: Environmental and Societal Impacts Group, National Center for Atmospheric Research.

Environmental and Societal Impacts Group
National Center for Atmospheric Research
PO Box 3000
Boulder, CO 80307-3000
Email: esiginfo@ucar.edu

The vulnerability of the electric power system to routine weather, as well as to weather extremes, modifies and amplifies the direct impacts of weather on society. These direct impacts, such as winds, precipitation, and temperature, are magnified in social and economic importance when they degrade the quality of electricity, including: reduced voltage, unstable frequency, interruptions in service, downed power lines, and abrupt price fluctuations. These secondary impacts can concentrate, extend, or shift the spatial and temporal extent of the weather event. Reliable and high quality electric service is central to public welfare and economic productivity, yet much of the electric power infrastructure, decision making, and consumption, remains highly exposed to the weather. Precise advance knowledge of the weather's influence on the electric power system may substantially reduce some of the most common and expensive societal impacts of weather.

It is a rare occasion when weather researchers and information providers sit down with the users of weather information in the electric power sector. Yet, it is clear that improved weather information is critical to mitigating physical risks in this sector, and it also became clear during this workshop that the industry would like to interact more often and more closely with the providers of weather information.

Many decision makers are involved in producing and delivering electricity, with different decision protocols, different time and space horizons, and different areas of concern. However, they all share the goal of maintaining service reliability while meeting necessary revenue or cost goals. Utilities, regulators, electricity consumers, and researchers need to know how to better use weather information in order to reduce the amplification of weather impacts while ensuring a healthy business. Policy makers who seek tools to align these goals also need access to weather information in order to recognize the limits and opportunities presented by weather's impact on the system. New weather information might also be needed, challenging weather researchers.

This workshop on improving the value of weather information for electric power operations outlined and prioritized research themes in meteorology that would be the most important in the electric power system. The theme encompassed reliability, information, and decision-making, in view of challenges and opportunities in the electric power industry's emerging deregulated environment. Restructuring of the industry may provide a sufficiently competitive climate for decision makers to accept new concepts and tools that enable compensated risks related to weather and communications, such as a forward market for reliability, forecasts of equipment operating capacity, and decentralized system optimization under uncertainty.

Thirty-five representatives from the electric power industry, electric power research, academia, and research meteorology were invited to participate in the two-day workshop. The first day's presentations were divided into four areas of business in the electric power sector: Demand, Supply, Delivery, and Trans-utility systems issues. The next day's discussions began with four groups, and then the groups mingled to exchange ideas.

Workshop results

The workshop was held on 6-7 November 2002, at the National Center for Atmospheric Research (NCAR) campus in Boulder, Colorado. Thirty-five experts from the electric power industry and the meteorological sciences attended to discuss the weather information needs of the electric power sector and the collaborations necessary to improve the value of this information.

The presentations emphasized five major aspects of meteorological information in electric power: its value; which weather information is used and how it is procured; how this information is summarized and passed on to decision makers; internal and external barriers to better use of the information; and, finally, desired improvements in meteorological information and meteorologically based decision tools:

The discussions were lively and productive. Several suggested action items require a partnership to enhance decision support systems. These included, in the near term,

Meanwhile, the industry needs to cultivate a culture that can utilize probability in decision-making, and in the slightly longer term, in collaboration with the government sector, it needs to develop business tools to use probabilistic forecast information.

The US Weather Research Program (USWRP), which sponsored the workshop, provides critical research underpinnings needed to improve forecasts required by this sector. However, this research must be focused in appropriate areas: improving forecasts of temperature and wind in the atmospheric surface layer; urban weather; cloud cover and quantitative precipitation; precise timing of fronts; seasonal deviation from climatology; and extreme events that could alter normal operations or threaten power outages.

3. Identification of enthusiastic extramural collaborators in industry and government

Gathering more information in small workshops on specific topics, e.g., the spatial optimization of information needs, the discovery of aligned programs in the topic, and focused stakeholder exchanges;
Assembling educational materials for the industry about the modern capabilities of weather forecasting;
A demonstration project in the potential value of probabilistic forecasts with application to a particular business or operational decision.

The USWRP aims to bring improved, more precise forecasts to vulnerable regions and weather-sensitive economic sectors. This process involves fundamental research, public-private cooperation and collaboration, and the development and evaluation of prototype forecast products and decision support tools. In the planning process, researchers, forecasters, and the users of forecast information identify the most relevant issues with high potential for scientific progress. Through collaborative workshops, priorities, milestones, and resource needs are developed that initiate a fast track for scientific progress and transition into operational forecasting and commercially available value-added products. Seven U.S. agencies currently support the US Weather Research Program: NOAA (National Oceanic and Atmospheric Administration), NSF (National Science Foundation), NASA (National Aeronautics and Space Administration), and the Departments of Defense, Transportation, Energy, and Agriculture.

The Environmental and Societal Impacts Group (ESIG) is taking a central role in establishing the research agenda for the "societal impacts" context of USWRP research. ESIG is suggesting an approach that provides useful scientific guidance to weather researchers by examining the decision making of prepared, sophisticated, formal decision makers who rely on weather information, whose actions affect cross-sections of society in a significant way, and who can adapt their decisions according to information about expected weather conditions.

A series of meetings or less formal discussions are envisioned with decision makers in different commercial and government sectors to analyze weather information needs, the technical capabilities of decision makers, barriers to the use of weather information, and improved meteorological research or services to increase the value of weather information. Follow-up on these discussions includes grant proposals to USWRP and other sources from discussion participants and other interested stakeholders to carry out collaborative research and demonstration projects.

This workshop establishes the following template for societal impacts projects by sector:

The electric power sector is the first set of decision makers to participate in this process.

Routine and extreme weather affect electricity consumption and the performance of equipment, both of which affect the cost and quality of electricity (frequency, voltage, reliability). Utilities use weather forecasts to interact with customers regarding impacts on service, to hedge risk in financial markets, and to plan maintenance, operations, infrastructure expansion, and recovery from damage. As deregulation and restructuring proceed, the industry will need to establish new responsibilities and relationships that require new tools and information for decision-making. This workshop explored how weather forecasts and improved weather information can provide resilience in the dynamic technological and institutional environment that characterizes the current electric power sector.

Previous workshops and studies have surveyed and documented the need for improved weather information in the electric power industry (see references). Tailored weather information and decision support tools are currently available for some applications, but weather research and information is not being used systematically in the electric power sector. Besides opportunities to improve tailored weather information, barriers in technology, economics, culture, and policy could be removed to improve the value and effectiveness of new weather information.

Recommendations from previous studies included changes to weather forecast variables, forecast communication methods, use of forecasts in decision support systems, observations and databases, data formats, data distribution systems, quality control procedures, interactions between researchers and users of weather data, and more effective use of the Internet.

Deregulation in the electric power industry is changing the allocation of responsibilities and risk in electric power system operations and management, as well as increasing the utility of enhanced weather forecasts. Incentives in the traditional regulated industry encourage utilities to investment in physical infrastructure to maintain reliability. The deregulated industry involves decentralized decision-making. There has been difficulty in assigning responsibility for maintaining reliable service, profitability for stakeholders, and the moderation of price fluctuations. In the deregulated environment, incentives tend to encourage making profit on transactions of the lowest-cost electricity, rather than investing in software or hardware to maintain reliability. Power transfers can place a burden the grid and degrade reliability, with the consumer potentially paying the biggest penalty. At the same time, reliable electric power of high quality is ever more important in the expanding digital economy. Increased reliability is in high demand. Capturing the value of reliability will require a new market strategy and technology to coordinate new consumer relationships centered on telecommunications and control engineering. Decentralized stakeholders, probably communicating online, will need reliable, customized weather forecasts to manage both infrastructure operations and financial risks.

Grid management involves many entities, and stakeholders with differing expertise all need the high quality information in forecasts and effective tools and decision support systems to integrate data, plan, and communicate their actions. New tools for representing and communicating grid conditions must emphasize speed, simplicity, and accuracy. Many tools in the industry are heading in the direction of visualization, as they are in meteorology. Visualization tools in the electric power industry will combine the physical world of meteorology and electricity infrastructure with abstract notions of system performance. Emerging tools like probabilistic grid flow models seem to be well suited to integrating certain kinds of weather forecasts. Ensuring the highest value application requires ongoing collaborations between meteorologists in both the public and private sector and experts in the management of electric power system operations.

Demand, or load, modeling is a key planning tool that is strongly weather-dependent, and has become increasingly crucial to operating efficiencies in the restructured industry. Collaborative research can help determine the most valuable representation of weather variables and their uncertainty for assimilation into demand modeling tools. Under competition, and with increased consumer involvement in the market, demand side management (DSM) alternatives will also alter demand models as prices change with load and as consumers plan their own consumption more consciously, with their own weather information and demand models. This complex interaction of social and physical variables has the potential to produce significant nonlinear dynamics in local and regional demand for electricity.

Some energy customers meet portions of electricity demand with local generation in order to more precisely anticipate energy costs and to maintain power quality and reliability. Military bases, hospitals, and companies in the credit-card industry, telecommunications, banking, and computer-chip manufacturing are most active in early adoption of distributed generation. Knowledge of accurate, location-specific weather, climate, and electricity market information is necessary to model the value of this investment over the long term. If retail electricity prices vary with load, and therefore weather, this calculation can be very complex.

Optimal operation of the distributed generation depends, at a minimum, on accurate 1- to 2-day weather forecasts. When co-generation of heat (cold) is combined with distributed generation installation, the importance of the weather becomes crucial to economic performance. These relationships must be studied and new tools for decision-making refined and developed. As fuel cells for distributed generation reach market maturity, effective weather forecasts will be part of shaping a transition to a more sustainable energy economy. The economic and reliable operation of the leading renewable resource, wind power generation, is especially demanding of accurate weather forecasts and tools for planning and operations.

The presentations took place after an overview of the USWRP from Lead Scientist Bob Gall and introductory remarks from Robert Harriss, which took note of the broad set of opportunities before the participants. Many speakers echoed the theme that better weather information's highest value is enabling more efficient and coordinated use of resources, which in turn increases the reliability and productivity of the electric power system. Long-range planning is governed by national and local economics and policy and by technologies used in the local service area; medium-range decisions are dominated by maintenance planning, which is affected by exceptional (not usually expected) weather events; and roughly 20% of the value of industry decisions in the short term is influenced directly by weather. Better weather information cannot guarantee total weather resilience of the power system, however, because there is an economic limit to the practical defense against weather extremes.

The financial stakes of an accurate weather forecast can be astronomical, even if reliability is not affected. The impact depends on the decision made, as well as the type of weather phenomenon. For example, the release of water from a dam in response to a forecast of 2 centimeters of rainfall in the drainage above the dam can cause a 7-figure loss if that rain doesn't fall. But the amount of precipitation has little effect on planning for electricity demand unless it is tied to strong changes in temperature or cloud cover. So, advance knowledge of the rainfall is of much less value for the demand-driven decisions than it is for the supply problem of running the hydro dam.

Speakers reinforced previous findings that weather information is difficult to use operationally in this industry, and that utilities face high costs (learning, hiring, computing) to make productive use of weather information. The expense is usually prohibitive for small utilities to use formalized weather information at all. Weather information is used across utilities for similar decisions. In different utilities, this is often done with different decision tools and different information sources. This is because of the differing influence of local weather patterns, terrain, local electrical technology (consumers and infrastructure), demand patterns, the electric market structure, the size of the utility, and its connections with other utilities. However, load planning, using demand (load) forecasts and models, are particular exceptions in which common products and common sources of weather data are used across utilities. To a first-order, the day-ahead peak of the dry bulb temperature is the most important weather variable in this forecast.

In general, industry decision makers using weather information are decentralized, have little meteorological training, occupy various positions in the company, are responsible for different regions, and come from various backgrounds. This situation will likely continue under deregulation. If consumers influence the electricity market in the future via innovations like demand management contracts, reliability service contracts, and distributed generation, accurate and symmetric weather information will be critical for efficient markets.

Culturally, the industry is conservative, partly because it has only been rewarded for "keeping the lights on," but not for innovative or operational efficiency, as was pointed out by more than one participant. Decisions are made in a deterministic framework, and there are long-standing practices to employ deterministic (expected value) forecasts. Operational decisions are proprietary secrets, and security of information is a concern. New concepts such as probabilistic forecasting, web-based communications, and increased automation will require cultural change before they are adopted.

Economically and politically, deregulation of the industry may force separate generation, transmission, and distribution companies to market themselves to customers by combining cost, reliability, and innovative services. Under pressure to compete, the decision-making culture may be flexible enough to adopt new tools and new modes of thought more rapidly than the vertical monopolies, in an effort to optimize operations and utilize infrastructure instead of investing in new infrastructure. Markets are incomplete, and technical and economic tools must continue to be developed.

Meteorologically, the needed improvements to routine weather forecasts focus on the accuracy of a few parameters (dry bulb temperature, dew point, wind speed and direction, cloud cover, and quantitative precipitation), and a few issues of time and spatial scale (urban and point locations, spatial grids under 1 km, hourly to 3-hourly time series forecasts, 1- to 5-day horizon, and correct timing of fronts). Forecasts of extreme events could help in recovery planning from damage, but may not prevent outage.

Maria Ilic delivered the keynote address, which set the context of weather discussions within the framework of reliability and risk under the differing incentives of the regulated and restructured industry. Technological and organizational developments in the industry may be leading it toward decentralized management of physical and financial risks, as well as adding to the difficulty in defining quantities like reliability and optimized operations. Technological advancement will enable, and industry restructuring will force, the development and merging of regional electricity markets with the physical electric power system. Integrating these two systems will require new technology, new definitions, and transparent information about the state of the electric power grid.

Competition will rely on decentralized stochastic optimization and feature separate functional and/or corporate entities for power supply, delivery, and consumption; decision making under uncertainty will be decentralized and iterative; temporally and spatially active price signals will close the loop to engage demand with supply; a potential will exist for valuing the right technologies for operations and control; and there may be issues with reliability and long-term system evolution. The role of information technology and its research is hypothesized to be critical in this communication between these actors and in controlling the operations of the system.

Open research problems include developing tools for moving from sequential operations/planning to interactive iterative decision making over various time horizons; the type of and rate at which information is provided; the realization that electric power reliability is a dynamic, not static, problem, and illustrates that market failure is related to a lack of adequate information. Reliability may become a private property, and reliability insurance may be a component of the new market. Weather information and weather tools overlap industry needs in developing smart components and smart controls in the distributed stochastic optimization of operations and planning.

James Weber emphasized the need for graphical depiction of grid loads and demonstrated the utility of graphical user interfaces in case applications by illustrating contours of local marginal price and line power transfer distribution functions (PTDF, the measure of where a "shipment" of power from point A to point B actually ends up on the grid). The systems previously built to serve local areas are now carrying long-distance shipments that could affect operations far from the origin and consumption of the electricity. The stakeholders interested in power issues are now therefore much broader, and a graphical interface showing derived parameters helps non-specialists to understand the system's performance. The tool could be useful to identify areas for optimal placement of new generation capacity, faulty transmission equipment, abuses of market power, and operator training.

Stephen Lee discussed the complexity of transmission grid operations and presented a probabilistic model of power transfers. Stakeholders would benefit from knowing the effect of transmission limitations on their activities in that community. The Community Activity Room (CAR) is a metaphorical concept for treating these limits as walls of a room within which market activities can freely move around in order to reach equilibrium. The ability to analytically describe and visualize the CAR in the state space of net interregional power exports will enable greater understanding of the need for more interstate transmission superhighways. This metaphor will also enable transmission planning and operation for a large interconnected power system to evolve to the next generation of probabilistic power system reliability assessment and integration of reliability and market efficiency. The weather also influences transmission capacities, outage probabilities, and scheduling of wholesale power transfers. Better forecasts of wholesale power transactions can lead to greater reliability and cheaper electricity prices. When an online probabilistic CAR is displayed for the entire community to see, it is quite conceivable that reliability can be turned into an equivalent energy product that has monetary value. It is the clear path to reliability insurance or reliability contracts.

Jane Preuss illustrated the close interaction between urban tree-trimming and land-reclamation practices with failures in distribution lines under natural conditions such as extreme weather and earthquakes. Urban and suburban policies to manage vegetation around distribution lines may have more effect on reliability than the weather. This project analyzed the potential role of external variables influencing recovery and restoration rates for an urban electric power utility, with a focus on recovery rates for the distribution system. Outage data have been obtained for five events in the area serviced by an urban utility that agreed to provide the principal investigators with basic data obtained from field logs that identifies outages by location (feeder) and cause. Data pertaining to feeders could be correlated with neighborhoods. Outages due to tree failures during both windstorms and the earthquake have been correlated with municipal boundaries, which are more restricted than the entire service district. Tree-related failure of power lines, poles and other distribution equipment occurs more often outside of the Seattle City limits. The City of Seattle Ordinance 90047 prohibits specific tree species with invasive root systems and species that are brittle and break up in wind. It also has adopted guidelines for how far trees must be planted and trimmed back when located in public areas where lines will be posted. Service restoration rates should be based upon empirical data rather than expert opinion to be realistic in planning for disaster recovery efforts.

Frank Monforte presented a cascading regression of load data versus weather variables to specify a load model, in which a base model independent of weather explains the day's load with a minimum absolute percent error (MAPE) of roughly 3%, which can be improved to 1.33% using recorded average dry bulb temperature. Additional improvement can be achieved by using time- and location-specific temperature. Other variables, such as dew point, cloud cover, and wind speed provided less predictable improvements to the models. Most of the demand can be forecast by knowing past demand and cultural and climatic indicators, such as the day of the week and date. Knowledge of the evolution of the day is more important than knowing just the coincident temperature.

Using forecast temperature as input to the model, which comes at 3-hour averages at airport locations, requires interpolation to hourly intervals and to local grid scales, and results in MAPE of 5% rather than the 1.33% obtained with historical hourly data. Better demand models will require more accurate forecasts of dry bulb temperature at the micro-spatial scale where the electric demand actually takes place, not at airports that may be miles away, and on hourly time intervals, though higher frequencies would be better. The secondary weather variables needed depend on the region in which demand is to be forecast (local technology, local weather). The critical forecast period is from 1 day to 10 days. Beyond 10 days a forecast of the weather patterns is necessary. Beyond two years, a weather (climate) forecast is not useful for demand modeling, as the state of the economy and adoption of technology by consumers takes over in significance. Specific changes that could be made focus on technology, data quality, and service. Doppler radar images are often in utility control rooms, and operators are able to convert this heuristically into an effect on their system load. How can we put this skill into an artificial neural network (ANN) or a load model statistic?

Michael Kintner-Meyer of PNNL explained the concept of load-as-a-reliability-resource (Consortium for Electric Reliability Technology Solutions [CERTS] research project), in which commercial customers are compensated for forgoing electric service. The peak load hours are brief, but the bulk of the infrastructure investment goes into maintaining reliability during these peaks. Load relief is an alternative to this investment, but it is hard to sell facilities managers on reducing consumption. Motivation for participation in compensated emergency demand reduction programs include the fact that it provides the image of good corporate citizenship, and it is clear how it saves money.

In determining the capacity to reduce demand, the solar radiance and a temperature record (dry, wet bulb) are the most important variables, in combination with the thermal response of the building (system inertia). Weather and thermal data are needed for at least 3 seasons in order to plan the demand reduction strategy. About 1 GW of the 820 GW in the US (0.1%) consists of bid load reductions that are used and compensated. A thermal load forecast would be part of a portfolio of tools and technologies to reduce system load, including: on-site generation, thermal/electrical storage, pre-cooling/pre-heating to shift power demand from peaks, shift of manufacturing away from a plant with high electric cost, and financial risk management like weather derivatives.

The loads (consumers) of tomorrow will be actively engaged in electricity and derivative markets in a pervasive information technology (IT) control and communications environment with central control of load and distributed intelligence. Weather adjustments are currently made after the fact. They need to be included in the system in time for the market decisions. The practical time horizon for weather information to be useful is half a day to tens of days.

Scott Samuelsen presented details about distributed generation (DG) of electricity and heat with gas micro-turbine generators and fuel cells, with scenarios for widespread use of DG in grid-connected office "power" parks and hybrid distributed generation systems which might include personal transportation, micro-turbine/fuel cell combinations, or PV/fuel-cell combinations. DG set close to a load can increase the reliability of the electric supply by eliminating the delivery system that is subject to overload and failure, by providing a local source of heat, and recently, as a source of income in demand response or grid voltage support strategies. Cleaner DG technology and falling grid reliability have made DG a more popular alternative. Quiet, clean DG presents a reversal of the trend of giant power generators away from urban regions and a return to generation in cities, potentially changing the utilization of transmission lines. Cogeneration of heat might also alter the system electric loads. Power parks are one strategy to propagate DG. They would be office parks along highways with good natural gas infrastructure and fuel cells to generate electricity and heat, nearing 80% efficiency and with low emissions.

IT is at the heart of the DG system for local/remote dispatch control and local measurement of dry bulb temperature and relative humidity. Forecasts of these variables are a critical input to load profile control to plan the use of electricity generated at the site versus the amount exported, and the use of the waste heat recovery (heat, cooling, processing). Solar insolation is important for renewable DG that uses PV and hydrogen storage, and forecasts would help manage the stored hydrogen.

Patrick Walshe is the staff meteorologist at TVA. He described temperature and demand forecasting in a large utility. TVA has the additional responsibility for maintaining acceptable water flows for recreation and natural preservation of the Tennessee River drainage.

TVA has one person making weather forecasts, which are converted to an energy plan, essentially for today and the next day, with a 10-day outlook. TVA purchases weather information from a vendor with whom Walsh is in frequent contact. With Walsh's corrections the ANNSTLF load model (TESLA model is in testing) is used to make the day's forecast for energy demand. This demand amount (MW) is passed on to energy schedulers who make generation commitments, after which the power system operator, the river system operator, and marketers plan efforts to accommodate this generation commitment.

TVA's weather information consists of a rolling hourly forecast of temperature and relative humidity for 5 cities in its service area, out to 240 hours from the present hour's forecast (10 days). The value of a correct or incorrect load forecast, especially combined with potential savings for efficiently scheduled maintenance and river operations, can be very high (6-7 figures) for precipitation and 5-6 figures for maintenance. The significance of errors in forecast temperature are less when the temperature is 65 degrees Fahrenheit, and increase if it is warmer or colder, up to 350 MW/degree. Better load forecasts would result from increased frequency of reporting current data, and improved accuracy of short-range meteorological models: MOS, local mesoscale effects, cloud cover, and precipitation forecasts (especially in summer convection, which affects peak load forecasts).

Rich Wilson presented the weather information and decision tools commonly offered to the industry by private vendors. Meteorlogix, WeatherBank, and WSI are the main providers. Typical products include automated load-forecast systems, integrated OMS/EMS technologies, enterprise GIS, improved local forecasts (2-5 day), improved prediction of high impact weather events, and long range seasonal and climate forecasts. The responsible individuals who purchase and maintain weather information tools are typically the vice president or manager of operations, dispatch manager, IT manager/analyst. Almost 95% of their business is with regulated utilities, where the trend is to use more GIS-integrated tools, especially for transmission/distribution applications.

Co-ops, municipalities, and other smaller gas and electric utilities are a potentially large market in need of weather information and decision tools, but they are generally priced out of the integrated systems that are developed for large utilities. Meteorlogix has not emphasized Internet services, though this seems to be an easy and promising medium to use, and it would be the method of access to smaller companies. Meteorlogix has a continued interest in open-architecture structures and web-based solutions. Decision tools are designed to suit market conditions above all.

Most Meteorlogix energy clients do not employ meteorologists directly, and consider Meteorlogix meteorologists as part of their company's team. Meteorologists in the private sector are trained with models and they often rely on them more than their judgment. The 2-5-day forecast horizon is becoming more important as forecasts improve. No significant improvement in 1-day forecasts has been seen recently.

The variables in which Meteorlogix customers demand improvement are: temperature, precipitation, dewpoint, and other specific and individual result from a consensus forecast. Standard products that nearly all utilities use are degree-day forecasts, historical data, real-time weather, short-term and long-term (climate) forecasts, and model data. To improve weather forecasts for utility decision making, Meteorlogix needs denser networks of weather stations, NWS preferred, and the frequency of reporting (hourly data) needs to be increased. Direct integration with user's GIS platforms would be valued by customers.

Bill Mahoney presented decision support systems (DSS) from a general research perspective, based on experience in aviation, range testing, water resource management, emergency management, and road maintenance decision tools produced at RAP/NCAR. RAP specializes in developing solutions to specific problems and transferring the technology from research to US government agencies, private sector, and foreign governments. Classifying decision tools by purpose can be helpful: strategic, tactical, operations management, incident management, "what if" scenario analysis, or training tool. A tailored solution is necessary and a bottom-up approach works better than top-down, beginning with stakeholder needs and available data and technological capabilities in the industry, and iterating the system with testing and customer feedback.

The weather information needs of the electric power community are highly specialized and not traditionally recognized in weather research. Probabilistic forecasts could help: what is the chance that temperature could exceed a certain threshold? The scientific challenges appear to be in boundary layer meteorology, thermodynamics, new probabilistic and statistics metrics, numerical modeling, and verification with limited data. Untapped IT and weather forecast capabilities offer significant opportunities to help. DICAST is a 6-model mesoscale forecast ensemble (probabilistic forecast) with fused data inputs developed and used at RAP. GIS systems used in road safety foreasts are integrated with RAP decision tools.

Bob Gough presented the potential for wind electricity generation in the vast Midwest wind resource, and suggested weather information that could help advance this technology. 75%-100% of the US electricity demand in the lower 48 states is available in the wind of the Midwest alone. South and North Dakota are downwind of much of the US electric demand. If wind power were used to its potential, Kyoto targets could be met (200-250 GW). "Basin Electric" provides 25% of the electricity in the Great Plains from 6 WAPA dams on the Missouri River. There has been a precipitation shift: old data may not be helpful for projections of future hydro production. WAPA has a 20-year planning horizon. Wind power could be used to complement H2O power. A participant suggested working with meteorologists to get economic value of wind and understand the system to get congressional support. Policy issues might be barriers.

Tim Hughes presented Oklahoma's progress in promoting wind generation, Oklahoma Wind Power Initiative (OWPI) (www.seic.okstate.edu/owpi/). He announced the Oklahoma State Energy Plan Meeting, which showed a big change in policy attitudes: earlier the State was only concerned with its gas and oil economy and did not want to make renewable energy plans.

The Oklahoma MesoNet has 114 stations with wind data at 10 meters elevation. Unfortunately not 100m. OWPI has made maps of Oklahoma's wind resource, based on MesoNet records and extrapolated to 50 m height, overlaid with transmissin lines. The next step is to work with people who have 100 m towers to verify the extrapolation to 50 m that is presented on the maps. NREL has worked on the low-level jet stream, and blades of wind turbines today reach over 100 m high. A participant asked what the cost of a MesoNet station was and if utilities could buy them? There were really good economies of scale for the 114 stations, about $7000 per station 10 years ago. Today they're over $10,000 apiece. OU/OSU did its own installation and instrumentation to save money. Law enforcement cooperation saves $1 million a year in communications costs by permitting the use of its radio frequency and repeater links to communicate automated measurements. The system has a 15-minute update including 5 min & 15 minute soil moisture averages.

The participants had challenging instructions for the discussions. First, they were to outline the current practices for using weather information in their area of expertise. Then they had to anticipate improvements that their colleagues could make, given improved weather information and/or decision support systems. And finally, they were to suggest improvements to weather information or its use and the means for achieving these improvements. Meteorologists made sure that the wished-for improvements remained realistic within bounds of scientific capability.

Participants were asked to develop well-defined recommendations, to think about specific issues, and to speak in a vocabulary that could be understood by all disciplines. They were to avoid issuing a "laundry list" of dreams and were asked to attempt to place a value on improved weather information. They were also asked to estimate necessary scales in space and time that would be needed in new weather information. Their thoughts were to be framed in terms of the USWRP research and technology transfer foci: operational test beds, fundamental research, demonstration projects, and case studies. They were asked to outline long-term issues, such as how to keep people involved in joint research, as well as short-term needs; to consider scenarios and to keep a holistic viewpoint, including policy variables. They were to consider their discussion as a contribution toward writing an RFP. The instructions are in an appendix to this document.

Participants were divided into four groups, representing the four distinct operating and decision-making groups in the industry today:

Before breaking out into groups, participants were asked to outline the top decisions and associated weather concerns in their area of expertise, in preparation for the breakout group sessions. On the whole, the needs written down by the participants before the group discussions agree with the needs brought out in the presentations.

As in the presentations, there is considerable variation between participants' opinions with respect to what weather data would be helpful for making what decisions. This is consistent with the different operating environments within the utilities and the responsibilities held by the participants. There is also no specific mention of the need for probabilistic forecasts, though the concept of a "distribution" of forecasts is mentioned several times. The value of combined physical weather variables arises repeatedly: wind/temperature, humidity/temperature, wind speed/direction, temperature/duration, reflecting the need for indices that contain more information than single-variable reporting. The matrices, as they were filled out by participants, are in an appendix to this document.

Each group was given the same questions. Questions 1-5 focused on bringing out the weather-related decisions and their value in each area of decision-making, and were to be answered in the first breakout session. Questions 6-11 for the second session demanded consideration of what changes in weather information would be most useful, how useful, and why. Questions 12-17 were intended for the third breakout session and focused on how to go about making improvements in weather knowledge, reporting, strategies, and decision tools:

The groups were eager to address the issues as a big picture rather than in individual questions, so they tended to answer multiple questions in a single answer. As a result of this rapid progress, the final (third) group session was replaced by a plenary discussion of "Next Steps" to pursue this research and development topic.

This first group report focuses on questions 1-5. The reports emphasize that weather forecasts vary in value according to the weather phenomenon, the decision, the timing of the forecast, and the ease (cost) of using the forecast in a decision protocol. In general, weather forecasts are used to improve productivity by avoiding failure or recovering more quickly from it, to avoid planning maintenance that would compromise the system's ability to cope with unexpected demand spikes, and to schedule the lowest-cost service at a particular time (grid management and scheduling generation). The latter application is recognized as having the highest value and most common application of weather information, because the short-term nature of these decisions means they are well supported by accurate short-term weather forecasts. Weather events that can cause grid failures such as icing, highly localized winds, or lightning, are more difficult to predict, and their forecasts are useful but less valuable because of their lack of specificity. Longer-term forecasts that might be useful for planning maintenance are also less accurate and therefore less valuable than day-ahead forecasts.

The industry's decisions are characterized by conservatism and the desire to minimize loss because of the high social consequences of failure. The weather forecasts used are based on NWS forecasts, either directly or as they are presented on television or in local papers. This information is often supplemented with local observations and corrected according to experience. None of the groups reported widespread use of automated decision support systems. The group participants did not identify any method for regular knowledge or technology transfer from research meteorology to operational forecasting at a utility, regulatory bureau, or large electricity consumer.

The demand group considered the problem of predicting electricity demand (load) from the point of view of a system operator and also from the point of view of a building energy manager, giving two interesting and distinct sets of weather needs and values of this information. System-wide load estimates are made on the time scale of one-half to 10 days, with hourly updates. Building (facilities) controls need more frequent updates in order for the heating and cooling system, as well as service curtailment contracts, to be used efficiently. Weather information for system-wide load estimates are NWS forecasts, often supplemented with local measurements, and/or purchased at higher spatial resolution from private vendors. Facilities need to supplement local forecasts with site-specific data that they measure themselves. The two kinds of load estimates use the same driving variables, however: typically the day's average dry-bulb temperature and relative humidity. Corrections are sometimes made for wind. In each case, the load model must have better than 5% accuracy to have value, but every 1% improvement in accuracy below 3% is extremely valuable, particularly in temperature extremes. One participant remarked that, for demand forecasts up to 5% accuracy, weather information is of practically no value, whereas every % improvement below 2% accuracy is worth millions of dollars. Accurate forecasts where the temperature is more than 10°F above or below 65°F are far more valuable than forecasts when temperatures are near 65°F. System-wide and facility-based load models would benefit from the same kinds of improvements in weather forecasts: microclimate adjustments; common data format (XML was suggested); more and better representation of urban weather; and improved forecast of precipitation.

Discussion focused on how this presentation emphasizes that weather information needs are user-specific. In addition to the needs of utilities and the facility managers outlined here, ISOs and regulators (e.g. NARUC) could also use demand models to monitor or influence electric markets and reliability. It is important that the ISO understand the market and what the motivations of the ISO and other members of the electricity community are for using this information. Dynamic game theory with weather information may help model load curtailment bids for advance knowledge of the markets, for instance.

Another line of comments and questions were aimed at the demand forecasting activities at a utility: "How good can the demand forecast get with perfect weather information? How much would it be worth? Who makes the adjustment to demand forecasts: meteorologists, the financial sector decision maker, or the power scheduler?" The value of the forecast is hard to estimate and it depends on the market and the action that the utility takes in response to the model. The ANNSTLF model, which is not a regression but a neural network, has been tested at TVA. Its accuracy depends on the system in which it is used. 1.5%-2% accuracy is typical with perfect temperature information. But people's habits change in a power crisis. It's therefore hard to get a regression to work well in such cases because of the limited data available for regression. Finally, the adjustments to demand forecasts are made by the staff meteorologist at TVA up to 2pm for the following day, but this specific activity varies by utility.

To reach new audiences, or to introduce new weather forecast tools to demand forecasters, participants suggested that research meteorologists attend facility managers' association meetings. The importance of nice graphics should not be underestimated in this endeavor. It was also recommended that research meteorologists come watch operational decision making in a utility and see the difficulty with using weather forecasts in practice.

Transmission and distribution both involve wires, and both kinds of wires are threatened directly by very high winds and ice storms. But each network is threatened by different secondary effects from weather, has different consequences if it fails, and their managers have different responses.

Transmission networks are subject to failure when elements are overheated by large transactions of power. This practice of sending electricity over long distances is increasing at an alarming rate as deregulated systems enable energy traders to buy the lowest-cost power and sell it for more, regardless of its origin and destination. Overheated transformers fail structurally, while wires can expand and sag into obstacles, causing short circuits of the very high voltage. Otherwise, transmission wires are high enough that they are not normally endangered by ground-based obstacles such as trees. In addition, transmission towers are often remote and difficult to monitor and repair. They can be subject to secondary weather-related hazards like wildfires, flash floods, and landslides (Hawkins, CAISO). Failures of transmission grids result in power outages for thousands to millions of consumers. On the positive side, the power-carrying capacity of a transmission line or transformer is increased by wind, as this cools the wire.

Distribution wires are lower voltage and lower to the ground. They are threatened by weather primarily through the action of falling or breaking trees. They affect hundreds to tens of thousands of consumers from the neighborhoods level to portions of cities.

The transmission and distribution group noted that weather-related decisions in this area currently focus on avoiding blackouts and restoring service after an interruption. One use of weather forecasts is mobilizing crews ahead of time to perform repairs in the event of a major weather disturbance, such as a hurricane or a storm that could produce icing (which can pull down both

kinds of lines). Regular maintenance must also be planned by taking unusual weather into account, so that the system is not undergoing major maintenance during a time of high electric demand. A scale called a "line rating" relates the maximum amount of power a line can carry at a given air temperature, and forecasts are sometimes used to curtail power shipments before they can exceed these ratings. This practice reflects the risk-averse decision making in this industry.

The group looked to the future to dynamic line ratings, based on weather forecasts and local temperature and wind observations, in order to make the most of the existing infrastructure. Accepting probabilistic forecasts of loads means that lines may be permitted to operate for short times at capacities over their fixed rating. Better weather information in the hands of transmission and distribution operators would also let them better plan the use of curtailment contracts, giving more advance notice and choosing the most advantageous contracts to activate. In this respect, distribution and transmission decision makers had needs in common with the demand forecasting group.

Currently, weather forecasts come from the mass media. There is a culture of conservatism because the positive rewards for innovation are outweighed by the negative impacts if the system fails. Therefore, acceptance of new forecast tools will be slow because of a perceived risk with no apparent benefit. Research money is also unlikely to be allocated to this kind of research within utilities unless managers can be convinced of the value of forecasts. If accepted, the useful weather information would include local temperature and wind forecasts, and forecasts of icing. Weather conditions across the grid might also be of interest, because this would tell operators where the demand and supply, as well as transmission bottlenecks, could be in the next day or two. Also, outlooks up to 2 weeks ahead would help re-schedule maintenance to avoid weather disturbances or weather that would drive high electric demand during the maintenance window. Long-term wind information might help municipalities with tree planting and pruning guidelines to protect distribution networks.

Discussion reiterated the importance of a consistent data format in accomplishing any goal that would integrate weather information with decision making. The fact that many transmission decisions are fundamentally probabilistic (likelihood of outage), and that future weather information may also be available in a probabilistic format, was also a common theme in the discussion.

Probabilistic information could be very valuable for planning the state of the system and for forecasting its reliability. Estimating the probability of a storm outage allows an ISO or utility to plan for and warn about the likelihood of curtailments, which are inconvenient, in order to prevent widespread blackouts, which are disastrous. Probabilistic models would help in siting transmission with respect to cooling winds, or with respect to distributed generation that is influence by weather (e.g. wind turbines), and to anticipate the distributed generation that will be running in certain weather conditions, since these large consumers will buy from the grid if that is more expedient. Expressions of both spatial and temporal overlap of weather and infrastructure would be needed. A "risk portfolio" based on demand, generation, and transmission would result.

The adoption of probabilistic forecasting faces barriers. Forward-looking individuals must be identified and convinced of the value added by such forecasts. One participant noted that there is inertia to overcome, and that transmission operators don't use weather information at all today except for dispatching crews after a failure. The need to see a link between a weather forecast and an application that they make regular use of. At the root of it, a participant remarked, any forecast is essentially a probability of an outage. A participant suggested that using probabilistic forecasts, but expressing only what is highly certain about the forecast to a decision maker, could diminish the cultural resistance to using forecasts of reliability, say the dynamic thermal rating. However, it was pointed out that, ultimately, someone would have to take the responsibility for making a decision based on the forecast instead of on the old static limit, which would hinder innovation.

Participants noted that EPRI already has technology for real-time dynamic line ratings, and that this investment is relatively inexpensive compared to replacing wires with high-tech alternative materials that sag less when heated. A 10% increase in rating that may be possible with dynamic line ratings could mean a very large financial return. Such a demonstration may be what is needed to usher in probability and forecasting into this sector of the industry.

The most important application of weather forecasts in electricity supply decisions is estimating the price of electricity ahead of time to gain a market advantage. This estimate is strongly based on the estimate of the demand for electricity, since supply availability and costs seem to be well known within a utility. A participant noted that forecasting is the principal element of business economics, and the demand estimate is the "lifeblood" of such a forecast. Suppliers must balance air quality, fuel consumption, electricity price, operating cost, maintenance, and the management of hydrological resources, to make profit. Failing to correctly anticipate the necessary supply forces a supplier to buy or sell on the volatile spot market, which greatly reduces the predictable revenues. Miscalculating supply is considered to be less a problem for reliability, and more a problem of controlling operating costs or losing advantage in the power market. Occasionally however, badly timed generator maintenance can result in blackouts.

The weather forecasts are the official local NWS forecasts for the nearest locations, and forecasts from private vendors. At TVA, they are modified by local temperature measurements according to a regression model. When hydrological forecasts are valuable, these forecasts are usually supplemented with a system of rain and flow gauges maintained by the utility. Precipitation forecasts are notably unreliable, however. 4-10 day forecasts are used to decide whether or not to defer planned maintenance, while seasonal climatologies are used to plan hydro resources. The highest value weather information for supply is next-day to 10 days.

Suppliers use commercial demand models and supplement them with their own information to make a marketing plan one day ahead of when the power will be needed: dry bulb temperature (the key index), dewpoint, wet-bulb temperature, winds, cloud cover, and precipitation are important variables. TVA uses a deterministic value for the forecast, and does not employ distributions. To learn about new meteorological knowledge and tools, industry meteorologists take advantage of invitations to demonstrate and evaluate new decision tools or new forecasts. This system relies on the staff meteorologist "keeping an eye" on the AMS newsletter, for example. Weather information could be more useful if it were not necessary to first have a middleman render NWS forecasts into a useful decision-supporting format. Automation would be helpful, for which a consistent reporting format would be necessary. Accurate precipitation forecasts would be extremely valuable for managers of hydroelectric dams.

Some follow-up discussion focused on the types of generation and its weather-dependence. Hydrological generators and wind turbines are not the only weather-exposed costs in generating electricity. The efficiency of thermal generators is also affected by weather. Gas turbines are not monitored and rated in real-time. The industry is very resistant to this idea. Ramping rates and other secret information must be protected. A participant suggested that this idea could be turned around however. The performance that could be expected under certain weather conditions could be calculated, illustrating the value of a weather forecast to the generating companies with respect to the efficiency (fuel efficiency) of their generators. A compenent of fuel price could also be included, since the price of natural gas will also vary with the weather.

The systems group presented their discussion in a table format. The table is in an appendix on the website. The systems group has a longer time horizon in their point of view:

For the demand forecast decision, 1 day (operations) and 1-20 year (planning) forecasts are valuable. Temperature is the most important variable, along with cloud cover, at the sub-utility/utility/regional scale. The highest value improvements in weather forecasting in the electric industry would accrue here, over the 1-hour to 10-day range.

For wholesale purchases and trading decisions, 1 hour information is useful in the spot market, 1 month-1 year outlooks are useful for planning, and 1 day to 2-10 day forecasts are needed for most market decisions. The latter are of very high value. Weather forecasts have somewhat less value in the shorter and longer time frames. Temperature and duration of temperature is important, at the regional scale.

Precipitation forecasts are useful in the long-term > 1 year (for capital investment), and in operations (at outlooks of < 1 month). The shorter term forecasts carry high value, but the longer-term capital decisions are worth even more.

For planning recovery from severe weather, 1 hour to 1 month forecasts (ice storms/hurricanes) that help schedule and place crews have high value. Knowing the probability of extreme events over the long-term (years) has about the same value.

For managing fuel resources (e.g. gas storage), forecasts of temperature for > 1 day-1 month are most useful, and this decision has a high value.

This group sees weather-related decisions in transmission and distribution planning as less valuable, and the need of forecasts is long-term (> 1 year). The resolution ranges from the sub-utility (distribution), through utility, to the regional scale.

Discussion of the systems presentation focused on engaging utilities and decision makers in improvements to tools and information. Utilities are diverse across service areas and in the equipment that they and their customers use. GIS representations are very important. How utilities make decisions must be taken into account when changing weather data/information. With regard to setting up meetings with utilitites, since many were invited to this workshop and they showed less interest in coming than expected, attending the meetings of the associations might be a place to start. Finding interested individuals, perhaps starting with NARUC and EPRI as a conduits of communication, is important.

Instead of answering the remaining questions in order, groups chose to report weather information needs and decision support systems according to the highest need and best chance of successful improvement. Refer to the questions to understand the context of these reports. Finally, each group recommended research themes and even specific projects. Additional comments were made by participants in the plenary session. The results are recorded here.

The demand group provided answers to the questions for Session II. Their concerns focus on the need for a common data format with cleaner data, and tailored products geared toward urban locations and fine spatial scale (microclimate). Products with quality graphical representations would be valuable. Maintaining contact with the weather research sector should take place via education, collaboration, and demonstration projects.

The transmission and distribution group summarized the new decisions and decision tools that could be possible with improved weather information. They persistently emphasized that a standard data format for weather forecasts and other weather information would be very useful to their decision tools. This branch of the industry may also have to make more in-depth study of the impact of weather on its operations in order to make use of improved weather information, because the exact mechanisms of some impacts are unknown. There may be cultural barriers that have hindered more widespread study and use of weather information. A demonstration of the value of weather information for decision-making might be needed.

The generation and supply group looked at information needs over the short, medium, and long-terms, and brainstormed a series of 2-day workshops to delve deeper into the needs for new weather information and decision support tools. Information needs center on ways to manage uncertainty by understanding climate variability, and by using probabilistic forecasts. Value demonstrations are needed for weather forecasts.

1. When available, by invitation to demonstrate and evaluate
2. Availability of DICAST, for example, is not widely known
3. Watching AMS newsletters

Dry bulb temperature modeling (0-84 hours, 0-384 hours)
Boundary layer modeling improvements
Strengthening observational networks
- Explore available resources
- Extend as needed
Probabilistic range specification in dry bulb temperature
Climate variability

Sell managers of generation infrastructure the idea that we can predict generator performance with weather information.

The systems group emphasized engaging utilities and decision makers in making improvements to tools and information, because of the diversity across utilities. All utilities have different needs based on their geographic situation and the technology used by their customers. Case studies of the value of weather forecasts, and demonstrations of probabilistic forecasts, would be a way to get the attention of decision makers. They recommended focusing on the RTOs and the standards and regulatory agencies, NARUC and NERC.

The group gathered in a final plenary discussion of the ideas from the presentations and the breakout groups. Some brainstorms and many well-developed projects emerged which emphasize understanding, research, and technology transfer to enhance reliability tools in the electric power industry and illustrate the high value of improved weather forecasting for USWRP.

Fundamental knowledge and societal impacts case studies will be useful to help the industry realize the mechanisms of weather impacts on their operations, and to help identify which weather forecasts are most useful: when, where, and for what decisions. Products (tools) and demonstrations of leading-edge meteorological capabilities should have the context of an industry decision in order to show relevance and value. Given the variable needs of utilities and the different decision making entities involved in the industry, choosing a test bed for demonstrations or technology transfer is an important first step. The location of the weather-related problem, the decisions being made, and the weather forecast tools must be chosen in a complementary fashion. Continued outreach to this industry will be necessary to sustain the relationships begun in workshops, case studies, and technology demonstrations. Trade meetings, academic research, and regulatory commissions all offer avenues for continued contact.