Load Forecasting Presentation

Load Forecasting
Tao Hong, PhD
Mohammad Shahidehpour, PhD
Disclaimer
This report was prepared as an account of work sponsored by an agency of the United
States Government. Neither the United States Government nor any agency thereof,
nor any of their employees, makes any warranty, express or implied, or assumes any
legal liability or responsibility for the accuracy, completeness, or usefulness of any
information, apparatus, product, or process disclosed, or represents that its use would
not infringe privately owned rights. Reference herein to any specific commercial
product, process, or service by trade name, trademark, manufacturer, or otherwise
does not necessarily constitute or imply its endorsement, recommendation, or
favoring by the United States Government or any agency thereof. The views and
opinions of authors expressed herein do not necessarily state or reflect those of the
United States Government or any agency thereof.
The information and studies discussed in this report are intended to provide general
information to policy-makers and stakeholders but are not a specific plan of action and
are not intended to be used in any State electric facility approval or planning
processes. The work of the Eastern Interconnection States’ Planning Council or the
Stakeholder Steering Committee does not bind any State agency or Regulator in any
State proceeding.
2
Load Forecasting
• Short-term
• Long-term
– Minutes to weeks
– Operations
– Weeks to decades
– Planning (incl. rates)
Integrated Load Forecasting:
3
Three Questions
• Why load forecasting now?
• How to develop and evaluate load forecasts?
• What are the insights and findings?
4
Why NOW?
• More COMPLICATED than ever before!
5
Why NOW?
• More UNPREDICTABLE than ever before!
6
How to Develop?
•
•
•
•
•
•
Use HOURLY data!!!
Benchmark model
Recency effect
Weekend effect
Holiday effect
Macroeconomic indicators
7
How to Evaluate?
• Ex post accuracy
– actual economy info
– actual weather info
• K-S statistic
• Probabilistic scoring
– Pinball loss function
– Winkler score
–…
8
How to Evaluate?
• Pinball loss function
– used in GEFCom2014 www.gefcom.org
qa:
y:
forecast at the ath percentile;
observation
Example #1: observation is 10MW above 90th percentile forecast
L = 0.9 x 10 = 9
Example #2: observation is 10MW above 10th percentile forecast
L = 0.1 x 10 = 1
9
Case Studies - Territories
10
Case Studies – Weather Stations
11
Case Study – One Example
• ComEd (2011 – 2013 forecasts)
12
Insights and Findings
• One size no longer fits all: variables
(ISO-New England)
13
Insights and Findings
• One size no longer fits all: weather stations
14
Insights and Findings
• Recession was real…
– NCEMC (2009 – 1013 ex ante probabilistic forecasts)
15
Insights and Findings
• Recession was real…
– NCEMC (2009 – 1013 ex post point forecasts)
16
Takeaways
• It’s time to modernize your load forecasting
process
– Embrace high granular data and recent
advancements in forecasting
• One size no longer fits all
– Spend your efforts customizing the models for
each region
• All forecasts are wrong
– Be realistic about the accuracy, especially in the
long run
17
Acknowledgement
• This material is based upon work supported by the
Department of Energy, National Energy Technology
Laboratory, under Award Number DE-OE0000316.
• Key project members
– Dr. Zuyi Li from IIT
– BigDEAL Students from UNCC: Jingrui Xie, Bidong Liu, Jiali Liu and Lili
Zhang
• Contributors
– Members form EISPC studies and white paper workgroup
– Members from IEEE Working Group on Energy Forecasting
– Colleagues in the energy forecasting community
18
Key References
• Tao Hong (2010). Short term electric load forecasting. PhD Dissertation,
North Carolina State University.
• Tao Hong (2014). Energy forecasting: past, present and future. Foresight:
The International Journal of Applied Forecasting, issue 32, 43-48.
• Tao Hong, Pu Wang and Laura White (2015). Weather Station Selection for
Electric Load Forecasting. International Journal of Forecasting, 31(2), 286295.
• Tao Hong, Jason Wilson and Jingrui Xie (2014). Long Term Probabilistic
Load Forecasting and Normalization with Hourly Information. IEEE
Transactions on Smart Grid, 5(1), 456-462.
19
Thank You
Dr. Tao Hong
www.drhongtao.com
http://blog.drhongtao.com
www.otexts.org/book/elf
20