Forecasting
Skip to:
- Example Interventions
- Tools and Methodologies
- Reading List and Case Studies
- Regulatory and Policy Examples
Introduction
To reduce the uncertainty inherent in demand and generation, system operators rely upon load and generation forecasts to balance electricity supply and demand. Accurate forecasts not only support the safe and reliable operation of the grid, but also encourage cost effective operation by improving the scheduling of generation and reducing the use of reserves. The better and more frequent the opportunities to use forecasts, the more impact that forecasts will have on systems operations.
Wind and solar forecasts are critical to reducing the uncertainty associated with variable renewable energy (RE) generation. To develop wind and solar generation forecasts, system operators utilize a combination of weather observations, satellite data, numerical weather prediction models, and statistical analysis to inform estimates of the level and location of generation in the near future. These forecasting techniques apply the following basic data to develop a wind or solar forecast:
- Precise location (latitude and longitude)
- Hub height (for wind)
- Current and historical metered power output of the generator
- Power supply curve
- Capacity in terms of plants or portions of plants that are operational (e.g., which turbines or strings of solar photovoltaic panels are online)
Obtaining accurate forecasts can be challenging. System operators need capable and available personnel to staff a sufficient forecasting process, which includes activities such as converting forecasting data into actionable system operation and potentially also data collection. Most system operators will handle load forecasting in house and some may also bring in outside help to combine multiple vendor wind and solar forecasts. Ideally, operators can utilize updated schedules with more accurate forecasts in periods closer to real-time deployment. In general, updating forecasts on a sub-hourly basis dramatically increases the benefit of these forecasts, so long as the system allows for corrective actions to be made within these same timescales.
Forecasts can be improved by:
- Increasing the amount and quality of solar and wind generation and forecast data to enhance the system operator’s ability to accurately monitor distributed variable RE systems on the grid (i.e., improve the visibility of these resources)
- More frequent measurements and observations of the weather and atmospheric conditions
- More frequent updates to Numerical Weather Prediction models
Centralized forecasting coordinated by a system operator provides several benefits in comparison to relying exclusively on forecasts provided by individual generators. Namely, by aggregating uncertainty across all generators in a balancing area, centralized forecasting smooths forecasting errors, which in turn provides more accurate information at the system level where the information is needed to monitor current conditions and schedule future generation. This helps to reduce system level risk and thus improve reliability.
At the same time, independent power producers can use their own forecasts separate of the central forecast to develop plant-specific schedules for delivering electricity. Independent power producers could be incentivized to develop forecasts that are intended to maximize value of the resource. Well-designed regulations and incentives can help discourage risky forecasting, facilitate the productive use of both centralized and decentralized forecasts to reduce uncertainty, and enhance reliability in systems with significant wind and solar generation.
Example Interventions
The following best practices can improve the accuracy and utilization of variable RE forecasts and the visibility of renewables to system operators:
Develop an enabling policy environment for forecasting
- Facilitate a national network of weather stations to create historical meteorological datasets that are useful for characterizing renewable resources and forecasting future power generation.
- Provide system operators with access to forecast-relevant data from generation resources or a national network for the development of centralized forecasts that are designed to minimize reliability risks.
- Adopt administrative or market rules that incentivize variable RE generators to submit accurate estimates of resource availability at all market time scales. When taken in combination with centrally aggregated forecasts, these resource-specific forecasts may help minimize the costs associated with reliably serving demand.
Enhance forecast accuracy and utilization in system operations
- Create visualizations that allow the operator to take more effective and efficient actions based on the aggregation of various information sources (e.g., forecasts, status of variable RE systems, ramping capabilities, geographic coverage of area control error sharing,etc.).
- Build historical datasets for time-synchronous demand and variable RE generation to understand the distribution of variability and uncertainty across the system.
- Shorten the time-step of forecasts to the minimum interval in which system operators can make actionable economic dispatch decisions.
- Improve visibility of distributed solar by identifying a methodology for forecasting generation from non-metered distributed generation and/or installing communications and remote controls on portfolios of distributed generation or on representative systems (upscaling).
- Consider using multiple forecasting sources, also known as ensemble forecasting.
- Exchange data and forecasts among balancing authorities within the same interconnection.
- Analyze whether current forecasting systems adequately forecast variable generation ramps that are anticipated under higher penetration scenarios.
- Collect technical information about ramp rates, start-up times, minimum on-line requirements, and efficiency curves. Use information in concert with sophisticated methods to translate forecasting data into information and trends of up- and down-ramping events and possible ramp times, durations, and magnitudes.
Sources: Bird and Lew 2012; Lew et al. 2011; Schwartz et al. 2012; GE 2014; and NERC 2011
Tools and Methodologies
Photovoltaic and Solar Forecasting
International Energy Agency, October 2013
This report discusses the links between weather forecasts and photovoltaic (PV) output, as well as the various methods for conducting forecasting. Also included is an example of “upscaling” in which representative systems are used to develop a forecast for a larger pool of systems. The authors also analyze the accuracy of tools and present a survey of tools used worldwide.
A Review of Wind Power Forecasting Models
International Conference on Smart Grid and Clean Energy, 2011
This review examines several wind power forecasting models, including Wind Power Management System, Wind Power Prediction Took, Prediktor, ARMINES, and Previento. These models use physical, statistical, and hybrid methodologies. The authors examine the accuracy of the models and possible causes of error.
A Survey on Wind Power Ramp Forecasting
Argonne National Laboratory, December 2010
This report examines wind forecasting of wind ramp events and possible metrics to evaluate ramps. The authors compare existing methodologies, including event detection and regression, to assess such ramps and find that additional work is needed to develop forecasting methodologies that are more accurate.
Reading List and Case Studies
Deutsche Gesellschaft fur Internationale Zusammenarbeit (GIZ), 2015
This report provides a brief overview of emerging best practices and state-of-the-art methods for variable RE forecasting. The authors differentiate between shortest term (0 – 6 hours), short term (6 – 48 hours) and medium-term (2 – 10 days) applications for solar and wind power forecasting. Topics covered include forecasting techniques (including data requirements, numerical prediction models, conversion of meteorological forecasts to power generation forecasts, ensemble forecasting, forecasting power plant outages and curtailments), forecast accuracy, and various approaches to RE forecasting. Chapter 3 provides examples of how systems around the world are approaching RE forecasting, with examples from Europe, the United States, India, South Africa, Brazil and Uruguay. Additionally, the report discusses general concepts for integrating RE forecasting into system operations.
The Value of Improved Short-Term Wind Power Forecasting
National Renewable Energy Laboratory, February 2015
This report quantifies the value of improved short-term wind power forecasting in the California Independent System Operator market and estimates savings from regulation and flex reserves, as well as production savings. The authors examine both low wind (8% penetration) and high wind (25% penetration) scenarios. Cost savings are estimated to be USD 1.27 to 17.1 million for flex reserves, USD 0.917 to 12.7 million for regulation reserves, and USD 2.87 to 116 million for production costs.
A Review of Variable Generation Forecasting in the West
Exeter Associates, Inc. for National Renewable Energy Laboratory, March 2014
Based on a series of interviews with thirteen operating entities in the Western Interconnection of the United States, this report summarizes practices, lessons learned, and priorities related to the implementation of wind and solar forecasting. The operating entities provided insights regarding cost assignment, forecasting accuracy and uses, data collection, curtailment and outages, probabilistic forecasting, distributed solar production, control room integration, and staff training, among other topics.
PJM, Exeter Associates, Inc., and General Electric International Inc., November 2012
This report describes the state-of-the-art with respect to variable generation integration, mostly focused on the United States but also providing a few international examples where particularly relevant. The report is predominantly based on an extensive literature review with input from General Electric (GE) and PJM. Pages 97–119 focus on forecasting.
 |  |
|---|---|
| Click Image for Full View | Click Image for Full View |
National Renewable Energy Laboratory, October 2011
This report summarizes Xcel Energy’s forecasting program for its Colorado service area. Xcel began partnering with the National Center for Atmospheric Research (NCAR) in 2008 to develop a new forecasting system. The report overviews the wind forecasting system, data acquisition, models, power conversion and outputs, results of the program and ongoing work.
Incorporating Wind Generation and Load Forecast Uncertainties into Power Grid Operations
Pacific Northwest National Laboratory, January 2010
This framework was developed to incorporate uncertainties associated with wind and demand forecast errors, unpredicted ramps, and forced generation disconnections into the energy management system (EMS) as well as generation dispatch and commitment applications.
Variable Generation Power Forecasting for Operations
North American Electric Reliability Council, 2010
The North American Electric Reliability Council (NERC) examines best practices for variable RE generation forecasting. The authors suggest using forecasts across timeframes and as close to real-time as is possible. The report focuses on forecasts from real-time to the coming 48-hours and how the forecasts can be best communicated to and utilized by system operators.
Regulatory and Policy Examples
Xcel Model Wind and Solar Power Purchase AgreementXcel Energy, 2017
Xcel Energy, a vertically integrated utility located in the United States, is a pioneer in incorporating state-of-the-art wind and solar forecasting into system operations. Xcel Energy’s Model Power Purchase Agreement (PPA) provides an example of a mechanism that enables the system operator to collect forecasting data (e.g., meteorological data, real-time power potential, and forecasted turbine availability) from wind and solar power generators. Sections of the Model PPA that relate specifically to data collection for forecasting include Article 10 (“Operations and Maintenance”) and Exhibit I, which includes a data list and protocols for automatic generation control for wind generators. In addition to forecasting considerations, the Model PPA also includes provisions related to compensation for wind and solar generators when Xcel Energy curtails renewable energy output (Article 8.3, "Curtailment").
FERC Order No. 764Federal Energy Regulatory Commission, June 2012
The Federal Energy Regulatory Commission of the U.S. (FERC) Order No. 764 is an example of a regulation that reduces barriers for variable RE by requiring transmission providers to offer sub-hourly transmission scheduling for their customers. The Order also requires new interconnection requests from customers with large variable RE to provide meteorological and forced outage data to transmission providers, in those cases where the provider conducts variable RE forecasting.
- A related report, FERC Order 764 and the Integration of Renewable Generation (EnerKnol Research, July 2014), summarizes the key changes to power system planning and operations that FERC Order 764 encourages. The authors outline the reasons why traditional electricity scheduling practices may lead to system inefficiencies in the presence of variable RE generation. While compliance costs for renewable energy producers are expected to be minimal, the report estimates that complete compliance by regional transmission operators and independent system operators will take many years to achieve due to challenges associated with cost allocation.
MISO Dispatchable Intermittent Resource Implementation Guide
Midcontinent Independent System Operator, 2010–present
In 2010, the United States’ Midcontinent Independent System Operator (MISO) began a process to work with stakeholders to design a new market mechanism that utilizes the benefits of wind energy and reduces the relevance of the non-dispatchability of the technology. This guide outlines MISO's approach to wind forecasting and practices to improve ramping capabilities.