The 2015 Paris Agreement represents the most recent step to forge a coordinated global response to reduce CO2 and related GHG emissions and limit the rise in mean global temperatures. Countries representing 97 percent of global GHG emissions submitted non-binding commitments in advance of the Paris conference. For example, the Obama administration pledged to cut domestic GHG emissions 26 to 28 percent below 2005 levels by 2025. It is widely recognized that existing commitments, both those ratified by nations and those that have not— including the US — are insufficient to meet the Paris agreement’s goal to limit the rise in global mean temperatures to well below 2° C.1-3
Energy-related GHG emissions account for approximately two-thirds of total global GHG emissions. There are many approaches to reducing the GHG emissions from the energy sector, including efficiency and conservation, fossil fuel switching (e.g., coal to natural gas), expanded use of nuclear power, and carbon capture and sequestration. It is also widely accepted that displacing fossil fuels with zero carbon, renewable sources of energy can and should play a significant role in addressing energy-related GHG emissions.4, 5 The yearly renewable energy potential is many orders of magnitude larger than current global consumption with direct solar radiation offering the greatest potential of all renewable sources.6
In 2016 renewable forms of energy represented just over 19 percent of global final energy consumption.7 Of all renewable energy sources, solar and wind have experienced the most rapid growth over the past several years. In 2016 solar photovoltaics (PV) achieved an historic milestone with capacity additions accounting for more additional net capacity than any other power generating resource. Net new generation capacity subtracts new fossil fuel capacity installed to replace decommissioned fossil fuel plants. In the same year solar accounted for 47 percent of newly installed renewable power capacity. Wind and hydropower accounted for 34 percent and 15.5 percent respectively of the total new renewable generation capacity added in 2016.7 The multi-year trend in the growth of installed solar and wind capacity is driven largely by declining costs, growing demand for electricity in some regions, and government incentives/programs to encourage investments in renewable generation.7 The prospects for continued price declines for wind and solar are good, thus it is likely that solar and wind deployment will continue to accelerate in the coming decade.
The Variability of Renewable Energy Production
Power grids are complex systems connecting a fleet of large electric power stations to households and businesses through a sprawling network of transmission and distribution lines. These systems evolved to reliably provide power to end-users continuously to meet the minute-by-minute variations in electricity consumption. This just-in-time production and delivery system requires a complex set of institutions and markets to assure reliable low-cost power to consumers when and where they demand it.
In the US, balancing authorities (including regional transmission organizations (RTO), independent system operators (ISO), or utility companies) serve this role dispatching generation based on forecasted demand for electricity and managing markets or generation resources to balance supply with demand in real time. Generation sources with different characteristics play different roles in this delicate balance between electricity use and production. Power plants that are slow to start and stop, taking hours or days, serve the role of providing baseload power. Fast response generation resources are well suited to meet the peak demand for electricity that often occurs only a few hundred hours each year. The one unifying characteristic of the vast majority of existing generation resources today is dispatchability, meaning they can deliver power to the grid when called upon by grid operators.
The rapid growth in the deployment of variable sources of generation, including solar and wind, is challenging the fundamental operational paradigm that has governed the management of electric power grids for over a century. It is generally understood that the grid must evolve to become more flexible using advanced smart grid control systems. This approach would leverage big data and information networks to both improve system reliability and power quality based on an evolving supply mix with increased production of low cost power from wind and solar generation sources.8-11
There are numerous strategies to create a more flexible grid on both the supply side and demand side. Supply solutions include geographical dispersion of variable sources of generation, larger balancing areas, fast responding generation with both fossil fuel-based (gas turbines) and renewable-based (hydroelectric) sources, and energy storage to name a few.9, 12 These approaches either reduce the overall variability in aggregate on a grid or provide buffer energy to offset variable output from solar and wind systems.
Demand-side resources have become increasingly important in managing peak power demand. In many jurisdictions, homeowners and businesses are compensated for decreasing demand for electricity during peak load events when reserve margins fall below a certain reliability-based threshold. Demand response programs designed to reduce peak demand for power could evolve to allow demand-side resources to respond to variable production of energy from wind and solar resources.13 Increasingly, these systems are being automated so that no individual needs to physically turn off and on equipment in response to a signal from a grid operator.14 The emerging vehicle-to-grid (V2G) concept represents a hybrid of both a supply- and demand-side approach to increase grid flexibility. Electric vehicles (EVs) can provide load shaping by scheduling charging during times of maximum generation from variable sources; with V2G grid-interactive systems EVs can supply power from the onboard battery in response to shorter-term fluctuations from variable sources of generation.15 Both demand- and supply-side approaches to grid flexibility can be made more effective when control strategies are informed with renewable energy generation forecasts.16, 17
The Practice of Renewable Energy Forecasting
Weather forecasting is an advanced science that serves a vital role in our daily lives and the economy. The process referred to as numerical weather prediction (NWP) utilizes complex models with embedded algorithms to predict the weather. These models are typically run on large, powerful government computers. The U.S. Global Forecast System (GFS) is the weather forecast model run by the National Center for Environmental Information, which is within the National Oceanic and Atmospheric Administration. Another leading weather model is the European Center for Medium-Range Weather Forecasting (ECMWF) model run by an independent international organization supported by 31 European countries. These models are “initialized” using actual weather conditions gathered from millions of sensors and recording devices scattered around the globe, including weather balloons, geostationary satellites, buoys, radars, sensors on commercial aircraft and ships, coastal and river gauges, and networks of ground-based observing stations. Mesoscale weather models provide greater horizontal resolution to generate smaller-scale details of weather forecasts. The Weather Research and Forecasting Model (WRF) is one of the leading mesoscale weather models in use today, which was developed in the later part of the 1990s.
Physical wind and solar production forecasting methods require wind speed and direction and ground-level irradiance and temperature respectively derived from weather models like those referenced above and input into engineering models. These models use specific technical and physical characteristics of particular wind farms or solar photovoltaic (PV) arrays to forecast energy production. These forecasts are improved over time using statistical methods based on historical real-time generation data in an effort to “tune” forecasts produced from inputs generated from NWP models.18 For shorter-term forecasts, a basic statistical model known as persistence forecasting is commonly used, which assumes that current levels of renewable energy generation will remain unchanged in the near-future. This and other purely statistical methods do not rely on physical system characteristics and instead rely on numerous data sources to developed a forecast using autoregressive and artificial intelligence predictive models.19 Short-term solar power production forecasting often makes use of cloud images obtained from either geostationary satellites or total sky imaging devices. The cloud movements, both direction and speed, are used to predict the impact of cloud cover on the short-term variability of a solar PV array production.19
Renewable energy generation forecasts can be produced for varying time frames. Although there is no standard characterization of renewable energy forecast time frames, it can be broken down as follows:
- Ultra-short-term forecasting: From few minutes to 1 hour ahead.
- Short-term forecasting: From 1 hour to several hours ahead.
- Medium-term forecasting: From several hours to 1 week ahead.
- Long-term forecasting: From 1 week to 1 year or more ahead.20
The forecasting method that yields the most accurate results is linked to the time frame under consideration. For example, forecasting approaches based on NWP data produce more accurate forecasts for day-ahead production versus a persistence forecast and the reverse with regard to short term intra-hour forecasts. The intended use of the forecast dictates the forecast time-frame that is most appropriate. For example, grid operators produce day-ahead dispatch schedules based on forecasted load and thus can integrate day-ahead renewable energy system forecast into the scheduling process to avoid starting and stopping conventional generation unnecessarily, saving fuel and money.18 Energy imbalance markets and real-time dispatch decisions require intra-hour renewable energy forecasts in ten-minute time steps for up to 6 hours ahead allowing market participants and grid operators to avoid costly under or over generation and maintain system reliability.18
Another important distinction between different forecasting approaches is between area forecasts and point forecasts.21 An area forecast aggregates estimates from numerous renewable energy systems to produce an estimate of aggregate output within a broader geographic control area. In contrast, a point forecast is produced for one particular renewable energy system be it a single wind farm or solar array. It is widely understood that area forecasts lower uncertainty due to the canceling effect of individual system variability due to aggregation.18 Most wind generation resources are connected at the transmission level and thus are typically included in the system dispatch planning enabled by day-ahead and hour-ahead forecasting. In contrast, much of the solar deployed today is distributed on the customer-side of the meter and thus ultimately impacts the load that appears on the system. As a result, solar energy forecasts play an important role in determining estimates of net load day-ahead, hour- ahead and in real time for regions with high penetration of distributed solar installations.22
Today numerous companies provide renewable energy forecasting services. The main market for their services is for grid operation and renewable plant management. Many grid operators in areas with high penetration of wind and solar either subscribe to these services or produce their own internal renewable energy forecasts to improve the management of grid resources given the variable output of renewable resources within the network. Wind forecasting is more mature and in wider use given the greater scale of wind energy deployment historically relative to solar.22 As both utility-scale and distributed solar installations expand, there will be growing interest in and use of solar forecasting services for grid operations and renewable energy plant management. A yet untapped use for renewable energy forecasts is by energy consumers allowing for individuals and organizations to plan energy use around renewable energy production. This is referred to as a bottom-up approach to managing the variable nature of renewable energy production. This stands in contrast to the top-down approach whereby grid operators seek to manage the impacts of variable sources of generation on system reliability.
Openly Accessible Renewable Energy Forecasts for the Public Good
As discussed earlier, societies must transition back to a point whereby the majority of energy supplies come from renewable sources. In essence, society needs to reconnect with renewable energy flows that served as the foundation for most of human existence.
We argue that making renewable energy forecasts publicly available to consumers is an important first step toward reconnecting society with renewable energy flows. With access to this information, consumers could consciously shift the use of energy based on the availability of renewable energy production. This could occur actively such as making a deliberate choice to delay the use of a household appliance like say a dishwasher or washing machine to a day in the week when the forecast predicts significant renewable energy production. An approach with perhaps more potential would be to leverage smart appliances and information networks to passively schedule appliance use based on embedded algorithms that are informed by a customer’s preferences to match household energy use with renewable energy generation.
National survey data indicates strong support for renewable energy by the general public. A 2016 in-depth survey by Pew Research Center found that over 80 percent of Americans support expansion of solar and wind farms.23 This level of strong support suggests that individuals may be willing to change behaviors if they were given the necessary information for the efficient integration of more solar and wind on their regional grid. Another survey found a significant portion of respondents indicated motives beyond saving money underlie energy efficiency behavior changes; these motivations include reducing impacts from climate change, acting morally, and feeling good about themselves.24 There is ample evidence that consumers respond to feedback on energy use and consciously alter behaviors when given the appropriate information and clear instructions on the appropriate actions.25, 26 Thus, it is reasonable to anticipate that if consumers are given access to user-friendly renewable energy production forecasts and the necessary information and tools they will adjust energy use accordingly.
Consumers that voluntarily shift energy use in response to renewable energy production are providing a benefit to society by allowing the efficient integration of low carbon sources of energy. One of the sectors that could gain economic benefit from this include grid operators, many of which already subscribe to a renewable energy forecasting service or generate their own forecasts internally. Thus initially, these organizations should be encouraged to make their forecasts available to the public in a user-friendly manner. While initially providing an area forecast for renewable energy production within the regional grid, the systems could be developed to allow for point forecasts that provide consumers with the anticipated output of renewable energy systems on their home or within their communities. Consumers who actively use the publicly available forecasting information may be eligible for financial incentives depending on the value their actions via reducing the cost of top-down efforts to manage the variable output of renewable sources of generation.
Over the long term, we imagine the federal government playing an important role in providing local weather forecasting data. The U.S. Department of Energy’s National Renewable Energy Laboratory (NREL), which currently maintains the Renewable Resource Data Center (RReDC), provides access to an extensive collection of renewable energy resource data, maps, and tools. NREL could serve the role of developing and providing publicly accessible renewable energy forecasts for all locations across the country. Similar to the federal government’s role in providing weather forecasts, public access to renewable energy forecasts serves a public good and could lower the costs of integrating renewable sources of energy into our existing grid system. Through a mobile app or online interface (see Figure 1) consumers would be able to tailor the forecasts they receive based on geographic location, type of forecast (area or point), timeframe (day-ahead, hour-ahead, or real-time), and intended use. This service should be accompanied by information about strategies consumers could adapt to alter behaviors that would allow for the low-cost integration of increasing penetration of variable sources of generation on regional grids. This bottom-up approach would serve to complement the top-down approach to grid integration of variable sources of generation by regional grid operators.
A likely first use case for publicly available renewable energy forecast is to link electric vehicle (EVs) charging with renewable energy production.27 As EVs represent a large, flexible new load emerging on regional grids, they are well suited to align charging with renewable energy production. EV owners are often motivated by the environmental benefits that are possible when using electricity for transportation. However, the environmental benefits are directly related to the sources of generation and thus EV owners may prefer to charge during periods of high renewable energy production on their regional grid. Many EV owners also invest in a home solar photovoltaic system to assure that they are charging from a clean source of energy. A survey of EV owners in California found that 39 percent of EV owners also invested in a home solar energy system.28 Net metering arrangements allow homeowners to get credit for energy that is produced and not used and thus the timing of vehicle charging does not necessarily align with the output of a home solar array. At work charging, however, could leverage the fact that EVs are stationary when solar production is at its peak. General Motors did an early demonstration in collaboration with Google transmitting renewable energy production data from PJM, the regional transmission organization covering a large part of the mid-Atlantic and mid-western states, through its OnStar vehicle communication systems to allow the charging of a fleet of 17 Chevrolet Volts to be coordinate with times when renewable energy production on the PJM grid were at their highest.29 Smart grid solutions more broadly will play a central role with the integration of renewable energy sources. Numerous technology firms are entering this space and developing solutions that provide the grid flexibility that is necessary to cost-effectively manage the variable nature of wind and solar production.30
The potential for bottom-up approaches to grid integration extends well beyond EV charging. Thermal and cooling applications for residential, commercial, and industrial customers are also well suited for load shaping given the latent energy storage capacity of heating and cooling systems. Furthermore, a bottom-up approach to grid integration through load shaping could reduce the amount of renewable energy curtailment, which occurs most often during periods of low demand.31 Initially, this strategy would have a negligible impact on load shapes, but over time load forecasting techniques would need to evolve in anticipation of increasing levels of bottom-up load shaping to support the overall goal of low-cost grid integration of variable sources of generation. Research is needed to characterize suitable end-uses for load shaping, estimate the intermittency management potential of a bottom-up approach, and gauge customer receptivity to altering energy use patterns in response to renewable energy forecasts.
Conclusions
Carbon dioxide emissions continue to rise and are now well above pre-industrial levels leading to an inevitable rise in mean global temperature. The 2015 Paris Agreement represents the first time in the multi-decade multilateral process to foster a global response to climate change with a general agreement to keep the future mean global temperature rise to well below 2°C. The years leading up to this global agreement have seen the renewable energy industry experiencing exponential growth. Renewable sources of energy now represent the lowest cost new power supply in many regions. In recent years, the deployment of renewable sources of generation has surpassed the deployment of new conventional sources of generation.
The technical challenges of managing a grid with increasing supplies of variable sources of generation have been studied now for at least a decade, and grid operators are currently deploying various strategies.32 The increased grid penetration of variable sources of generation requires a more flexible grid using many different approaches both on the supply side and demand side. Most strategies place grid operators and utility companies at the center of managing the grid impacts of increasing variability of electricity supply. Renewable energy forecasting has evolved rapidly with numerous companies providing forecasting services that are being increasingly used by grid operators to manage the variable output of wind and solar generation. Emerging sensor technologies and data aggregation techniques will likely provide the data necessary to radically improve forecast accuracy for individual renewable energy generators.
Global societies have largely become disconnected from renewable energy flows that have supported humans for most of their existence. This paper argues that public access to renewable energy forecasts represents an avenue to reconnect society to renewable energy flows, and could provide a valuable bottom-up approach to managing variable sources of generation. Experience in the efficiency field demonstrates that people respond to energy information to reduce consumption through altering behaviors not only for economic gain but also for altruistic reasons.33 Thus, it is likely that given the public support for wind and solar and past experience with energy behavior changes in the efficiency field, that individuals may voluntarily alter behaviors in an effort to shape loads to accommodate increasing amounts of variable generation from wind and solar resources. Additional research is needed to assess the costs and benefits of making renewable energy forecasts publicly available and to determine what institutions would be best equipped to provide these services.
References
1. United Nations Framework Convention on Climate Change (UNFCCC). (2015). Paris Agreement. Bonn, Germany.
2. Natural Resources Defense Council (NRDC). (2017, November 1). Issues Brief, The Paris agreement on climate change.
3. United Nations Environment Programme (UNEP). (2017). The emissions gap report 2017. United Nations Environment Programme, Nairobi.
4. Intergovernmental Panel on Climate Change (IPCC). (2012). Special report on renewable energy sources and climate change mitigation. ISBN 978-92-9169-131-9.
5. National Academy of Sciences (NAS). (2010). America’s climate choices: Report in brief. National Academies Press, Washington, D.C.
6. Perez, R. and M. Perez (2009). The world’s energy reserves: A fundamental look. International Energy Agency, Solar Heating and Cooling Programme, Solar Update Vol. 50: April 2009, Retrieved March 07, 2018, from https://www.iea-shc.org/Data/Sites/1/publications/2009-04-SolarUpdate.pdf.
7. Renewable Energy Policy Network for the 21st Century (REN21). (2017). Renewables 2017 global status report. ISBN 978-3-9818107-6-9.
8. Denholm, P., J. Novacheck, J., Jorgenson, & M. O’Connell. (2016). Impact of flexibility options on grid economic carrying capacity of solar and wind: Three case studies. National Renewable Energy Laboratory, NREL/TP-6A20-6685.
9. Lund, P., D. Lindgren, J & M, Jani, & S., Jyri. (2015). Review of energy system flexibility measures to enable high levels of variable renewable electricity. Renewable and Sustainable Energy Reviews. 45, 785-807.
10. Denholm, P. & M. Hand. (2011). Grid flexibility and storage required to achieve very high penetration of variable renewable electricity. Energy Policy. 39(3), 1817-1830.
11. Mai, T., D. Sandor, R. Wiser, & T. Schneider. (2012). Renewable electricity futures study: Executive summary. National Renewable Energy Laboratory, NREL/TP-6A20-52409-ES.
12. Lovins, A. (2017). Reliably integrating variable renewables: Moving grid flexibility resources from models to results. Energy Policy. 30(10), 58-63.
13. Swisher, J. (2012). The Role of Demand-side resources in integration of renewable power. Proceedings of the 2012 ACEEE Summer Study on Energy Efficiency in Buildings. American Council for an Energy Efficient Economy. Washington, DC.
14. Watson, D. et al. (2012). Fast automated demand response to enable the integration of renewable resources. Lawrence Berkeley National Laboratory, LBNL-5555E.
15. Kempton W. & S. Letendre. (1997). Electric vehicles as a new power resource for electric utilities. Transportation Research-D. 2(3), 157 – 175.
16. Minhas D., R. Khalid & G. Frey. (2017). Load control for supply-demand balancing under Renewable Energy forecasting. IEEE Second International Conference on DC Microgrids, Nuremburg, Germany, 365-370.
17. Perez, R., S. Kivalov, T. Hoff, J. Dise, & D. Chalmers. (2013). Mitigating short-term PV output intermittency. Proceedings of 28th European Photovoltaic Solar Energy Conference and Exhibition. Paris, France.
18. U.S. Agency for International Development (USAID). (2016). Forecasting wind and solar generation: Improving system operations. Greening the grid fact sheet, Retrieved February 06, 2018, from http://greeningthegrid.org/resources/factsheets/copy_of_ForecastingWindandSolarGeneration.pdf.
19. Pelland, s., J. Remund, J. Kleissl, T. Oozeki, & K. De Brabandere. (2013). Photovoltaic and solar forecasting: State of the art. A report by the International Energy Agency Photovoltaic Power Systems Programme. IEA?PVPST14?01.
20. Chang, W. Y. (2014). A Literature review of wind forecasting methods. Journal of Power and Energy Engineering. 2, 161-168.
21. Letendre, S., M. Makhyoun, & M. Taylor. (2014). Predicting solar power production: Irradiance forecasting models, applications and future prospects, Smart Electric Power Alliance: Washington, DC.
22. U.S. Department of Energy (USDOE). (2016). Solar forecasting: Maximizing its value for grid integration. Sun Shot Program, System Integration, Retrieved February 6, 2018 from https://energy.gov/sites/prod/files/2016/08/f33/Solar%20Forecasting%20White%20Paper_SunShot_2016_v1.pdf.
23. Funk C. & B. Kennedy. (2016). The politics of Climate: 2. Public opinion on renewables and other energy sources. Pew Research Center, Retrieved February 6, 2018 from http://www.pewinternet.org/2016/10/04/public-opinion-on-renewables-and-other-energy-sources/.
24. Leiserowitz A., E. Maibach, & C. Roser-Renouf. (2008). Saving energy at home and on the road: A survey of Americans’ energy saving behaviors, intentions, motivations, and barriers. Yale Project on Climate Change School of Forestry and Environmental Studies, Yale University.
25. Ehrhardt-Martinez, K. (2012) A comparison of feedback-induced behaviors from monthly energy reports, online feedback, and in-home displays. Proceedings of the 2012 ACEEE Summer Study on Energy Efficiency in Buildings. American Council for an Energy Efficient Economy. Washington, DC.
26. Faruquia, A., S. Sergicib, & A. Sharif. (2010) The impact of informational feedback on energy consumption: A survey of the experimental evidence. Energy. 35, 1598–1608.
27. Letendre, S. & M. Perotti. (2012). The business case for matching renewable energy production with vehicle charging. Proceedings of the EVS26 (26th Electric Vehicle Symposium), Los Angeles, CA.
28. California Center for Sustainable Energy (CCSE). (2012). California plug-in electric vehicle owner survey. A report prepared for the California Air Resources Board, Retrieved on February 8, 2018 from https://energycenter.org/sites/default/files/docs/nav/policy/research-and-reports/California%20Plug-in%20Electric%20Vehicle%20Owner%20Survey%20Report-July%202012.pdf.
29. General Motors Corporation (GM). (2012). Volt owners may soon get charged with renewable energyOnStar using Chevrolet Volts in Google’s “Gfleet” to demonstrate technology. Retrieved on February 8, 2018 from https://media.gm.com/media/us/en/gm/home.detail.html/content/Pages/news/us/en/2012/Jan/0123_onstar.html.
30. Kempener, R., P. Komor, & A. Hoke. (2013). Smart grids and renewables: A guide for implementation. A working paper prepared for the International Renewable Energy Agency. Abu Dhabi, United Arab Emirates.
31. Bird, L., J. Cochran, & X. Wang. (2014). Wind and solar energy curtailment: Experience and practices in the United States. National Renewable Energy Laboratory, NREL/TP-6A20-60983.
32. Weiss, J. & B. Tsuchida. (2015). Integrating renewable energy into the electricity grid: Case studies showing how system operators are maintaining reliability, prepared for Advanced Energy Economy Institute.
33. Mahone A. & B. Haley. (2011). Overview of residential energy feedback and behavior?based energy efficiency. Prepared for the Customer Information and Behavior Working Group of the State and Local Energy Efficiency Action Network.
