Climate Change AI Wiki

Electricity Supply Forecasting: Difference between revisions

Page Discussion

Electricity Supply Forecasting (edit)

Revision as of 07:27, 5 July 2022

3,211 bytes added ,  1 year ago
→‎Data
VisualWikitext
(take out courses)
 
(9 intermediate revisions by 3 users not shown)
Line 1:
''This page is about the applications of machine learning (ML) in the context of electricity supply forecasting. For an overview of [https://en.wikipedia.org/wiki/Energy_forecasting energy forecasting], [https://en.wikipedia.org/wiki/Wind_power_forecasting wind], or [https://en.wikipedia.org/wiki/Solar_power_forecasting solar forecasting] more generally, please see the Wikipedia page on this topic.''
Since variable generation and electricity demand both fluctuate, they must be forecast ahead of time to inform real-time electricity scheduling and longer-term system planning. Better short-term forecasts can allow system operators to reduce their reliance on polluting standby plants and to proactively manage increasing amounts of variable sources. Better long-term forecasts can help system operators (and investors) determine where and when variable plants should be built. While many system operators today use basic forecasting techniques, forecasts will need to become increasingly accurate, span multiple horizons in time and space, and better quantify uncertainty to support these use cases. ML can help on all these fronts.
 
The supply and demand of power must both be forecast ahead of time to inform electricity planning and scheduling. ML can help make these forecasts more accurate, improve temporal and spatial resolution, and quantify uncertainty.
To date, many ML methods have been used to forecast electricity supply and demand. These methods have employed historical data, physical model outputs, images, and even video data to create short to medium-term forecasts of solar power [32–40], wind power [41–45], “run-of-the-river” hydro power [19], demand [46–49], or more than one of these [50, 51] at aggregate spatial scales. These methods span various types of supervised machine learning, fuzzy logic, and hybrid physical models, and take different approaches to quantifying (or not quantifying) uncertainty. At a more spatially granular level, some work has attempted to understand specific categories of demand, for instance by clustering households [52, 53] or by disaggregating electricity signals using game theory, optimization, regression, and/or online learning [54–56].
 
== Background readingsReadings ==▼
While much of this previous work has used domain-agnostic techniques, ML algorithms of the future will need to incorporate domain-specific insights. For instance, since weather fundamentally drives both variable generation and electricity demand, ML algorithms forecasting these quantities should draw from innovations in climate modeling and weather forecasting (§7) and in hybrid physics-plus-ML modeling techniques [33–35]. Such techniques can help improve short- to medium-term forecasts, and are also necessary for ML to contribute to longer-term (e.g. year-scale) forecasts since weather distributions shift over time [57]. In addition to incorporating system physics, ML models should also directly optimize for system goals [58–60]. For instance, the authors of [58] use a deep neural network to produce demand forecasts that optimize for electricity scheduling costs rather than forecast accuracy; this notion could be extended to produce forecasts that minimize GHG emissions. In non-automated settings where power system control engineers (partially) determine how much power each generator should produce, interpretable ML and automated visualization techniques could help engineers better understand forecasts and thus improve how they schedule low-carbon generators. More broadly, understanding the domain value of improved forecasts is an interesting challenge. For example, previous work has characterized the benefits of specific solar forecast improvements in a region of the United States [61]; further study in different contexts and for different types of improvements could help better direct ML work in the forecasting space.
 
=== Solar power forecasting ===
▲== Background readings ==
 
* '''"A review and evaluation of the state-of-the-art in PV solar power forecasting: Techniques and optimization" (2020)'''<ref>{{Cite journal|last=Ahmed|first=R.|last2=Sreeram|first2=V.|last3=Mishra|first3=Y.|last4=Arif|first4=M.D.|date=2020-05|title=A review and evaluation of the state-of-the-art in PV solar power forecasting: Techniques and optimization|url=http://dx.doi.org/10.1016/j.rser.2020.109792|journal=Renewable and Sustainable Energy Reviews|volume=124|pages=109792|doi=10.1016/j.rser.2020.109792|issn=1364-0321}}</ref>: A review of work in solar power forecasting.
== Libraries and tools ==▼
* '''"Forecasting of photovoltaic power generation and model optimization: A review" (2018)'''<ref name=":0">{{Cite journal|last=Das|first=Utpal Kumar|last2=Tey|first2=Kok Soon|last3=Seyedmahmoudian|first3=Mehdi|last4=Mekhilef|first4=Saad|last5=Idris|first5=Moh Yamani Idna|last6=Van Deventer|first6=Willem|last7=Horan|first7=Bend|last8=Stojcevski|first8=Alex|date=2018-01|title=Forecasting of photovoltaic power generation and model optimization: A review|url=http://dx.doi.org/10.1016/j.rser.2017.08.017|journal=Renewable and Sustainable Energy Reviews|volume=81|pages=912–928|doi=10.1016/j.rser.2017.08.017|issn=1364-0321}}</ref>: A review of work in solar power forecasting.
 
=== Wind power forecasting ===
 
* '''"Current methods and advances in forecasting of wind power generation" (2012)'''<ref>{{Cite journal|last=Foley|first=Aoife M.|last2=Leahy|first2=Paul G.|last3=Marvuglia|first3=Antonino|last4=McKeogh|first4=Eamon J.|date=2012-01|title=Current methods and advances in forecasting of wind power generation|url=http://dx.doi.org/10.1016/j.renene.2011.05.033|journal=Renewable Energy|volume=37|issue=1|pages=1–8|doi=10.1016/j.renene.2011.05.033|issn=0960-1481}}</ref>: A review of work in wind power forecasting.
* '''"A review on the forecasting of wind speed and generated power"''' '''(2009)'''<ref>{{Cite journal|last=Lei|first=Ma|last2=Shiyan|first2=Luan|last3=Chuanwen|first3=Jiang|last4=Hongling|first4=Liu|last5=Yan|first5=Zhang|date=2009-05|title=A review on the forecasting of wind speed and generated power|url=http://dx.doi.org/10.1016/j.rser.2008.02.002|journal=Renewable and Sustainable Energy Reviews|volume=13|issue=4|pages=915–920|doi=10.1016/j.rser.2008.02.002|issn=1364-0321}}</ref>: A review of work in wind speed and wind power forecasting.
 
== Conferences, Journals, and Professional Organizations ==
 
=== Journals ===
 
* '''International Journal of Forecasting''': The official journal of the [https://forecasters.org/ International Institute of Forecasters]. Journal website [https://www.journals.elsevier.com/international-journal-of-forecasting here].
 
▲== Libraries and toolsTools ==
 
* See [https://docs.google.com/document/d/1tk9cF4O539TzaMaUufn9Ay4f6qKKEyoNKmzP03kbSDo/edit# this list] by Jack Kelly of Open Climate Fix for useful tools for processing and visualizing data during solar PV nowcasting workflows.
 
== Data ==
 
=== Future directionsGeneral ===
 
* See the [[Electricity Systems]] page for general electricity systems datasets.
 
=== Solar power forecasting ===
 
* See [https://docs.google.com/document/d/1tk9cF4O539TzaMaUufn9Ay4f6qKKEyoNKmzP03kbSDo/edit# this list] by Jack Kelly of Open Climate Fix for useful data sources for solar PV nowcasting.
* '''SubseasonalRodeo''': "A benchmark dataset for training and evaluating ''subseasonal forecasting systems''—systems predicting temperature or precipitation 2-6 weeks in advance—in the western contiguous United States." Available [https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/IHBANG here].
* '''American Meteorological Society 2013-2014 Solar Energy Prediction Contest''': Contest data for producing daily forecasts of solar energy, available [https://www.kaggle.com/c/ams-2014-solar-energy-prediction-contest here].
 
=== Wind power forecasting ===
* '''A collection and categorization of open-source wind and wind power datasets''': The paper gives an overview of open-source wind and wind power datasets, available [https://onlinelibrary.wiley.com/doi/epdf/10.1002/we.2766 here].
 
== Future Directions ==
 
* '''Hybrid physical modeling:''' As weather fundamentally drives solar and wind power production, and important direction could be for solar and wind power forecasts to draw on innovations in climate modeling, weather forecasting, and hybrid physical modeling techniques.<ref name=":0" /> These techniques might be able to improve short- to medium-term forecasts (e.g., by modeling fine-grained wind turbulence) and enable ML-based forecasts to deal with issues of weather distribution shift for longer-term (e.g. year-scale) forecasts.
* '''Decision-integration:''' As supply and demand forecasts ultimately need to inform power system optimization decisions, a fruitful direction may be to integrate knowledge of how these decisions are made into ML models. For instance, deep neural networks have been used to forecast electricity demand in a way that optimizes for electricity scheduling costs rather than forecast accuracy,<ref>Donti, Priya, Brandon Amos, and J. Zico Kolter. "Task-based end-to-end model learning in stochastic optimization." In ''Advances in Neural Information Processing Systems'', pp. 5484-5494. 2017.</ref> and this notion could be extended to optimizing for greenhouse gas emissions.
* '''Interpretable/explainable ML and uncertainty quantification:''' Techniques that explain or quantify the uncertainty of forecasts could help power system operators better integrate these forecasts into their operations, and facilitate applications such as robust optimization.
 
== Relevant Groups and Organizations ==
 
* '''OpenClimateFix''': A "non-profit research and development lab" working on projects including solar PV forecasting and PV panel mapping. Website [https://openclimatefix.org/ here].
 
== References ==
Retrieved from "https://wiki.climatechange.ai/wiki/Special:MobileDiff/184...6010"