Sign in Subscribe
Technical institutes

Factor-Based Portfolio Optimization for Risk Management in Electricity Markets with Renewable Energy

Strategic Insights

— 1 min read

A team of researchers from Massachusetts Institute of Technology (MIT) and Lehigh University has developed a novel factor-based portfolio optimization approach to manage risk associated with variable renewable energy (VRE) forecasting errors. The approach considers the geographical correlation of VRE resources and reduces aggregate VRE shortfall compared to traditional single-asset methods. The researchers propose integrating this factor method into unit commitment formulations and demonstrate its effectiveness in a New York Independent System Operator (NYISO) case study. By using this approach, they expect to reduce expected system wholesale costs by 11-22% and unserved demand by 16-22%.

Key Takeaways:

  • The team developed a factor-based portfolio optimization approach to manage risk associated with VRE forecasting errors.
  • The approach considers the geographical correlation of VRE resources and reduces aggregate VRE shortfall.
  • The researchers propose integrating this factor method into unit commitment formulations.
  • A NYISO case study demonstrated significant benefits, including a reduction in real-time prices in most locations (on average 9-20%).
  • Expected system wholesale cost reduced by 11-22%.
  • Expected unserved demand reduced by 16-22%.
  • The approach was highlighted as part of the ABSCoRES Project, which aims to develop novel risk management strategies for coherent risk measures in electricity systems.
  • Researchers used data from 160 existing/planned solar sites in NYISO, covering a year with 2/3 for training and 1/3 for testing.

Statistics:

  • The factor model reduced daily real-time wholesale cost by 22% ($153M vs. $119M).
  • Load-weighted average real-time LMP across the network reduced by 19.9%.
  • The factor model reduced load shedding by 16.6% (362 MWh vs 434 MWH).
  • Single asset score vs. factor model: daily wholesale real-time cost reduced by 11% ($134M vs. $119M).
  • Average LMP 1-2pm ($/MWh) load-weighted average real-time LMP across the network reduced by 9.2%.
  • Single asset score vs. factor model: The factor model reduces load shedding by 21.6% (362 MWh vs 462 MWH).

Sources:

  • Y. Liu, A. Sur, A. J. Lamadrid, and A. Botterud, "Factor-Based Portfolio Optimization for Risk Management in Electricity Markets with Renewable Energy," Working Paper, Available at SSRN: http://dx.doi.org.sdpl.idm.oclc.org/10.2139/ssrn.5123750
  • ARPA-E PERFORM, DE-AR0001277
  • Proceedings ACM e-Energy, Orlando, FL, June 2023
  • IEEE Transactions on Energy Markets, Policy, and Regulation, Vol.2, No. 1, pp. 132-145
  • EJOR, Dec 2024

Read more