Listen to the podcast:

This study examines how short-term variability in wind power—known as wind intermittency—affects real-time electricity system imbalances in U.S. regional power markets. The authors, Victoria Godwin and Matthew E. Oliver of the Georgia Institute of Technology and EPIcenter affiliates, analyze data from four major system operators: Bonneville Power Administration (BPA), New York ISO (NYISO), Southwest Power Pool (SPP), and PJM Interconnection. They focus on Area Control Error (ACE), a real-time metric used by grid operators to measure the mismatch between electricity supply and demand, adjusted for frequency deviations. Maintaining ACE near zero is essential for grid stability.

The authors find that a doubling of hourly wind generation variance increases average hourly ACE by 2% in BPA, 3.7% in NYISO, and 11.4% in SPP—equivalent to 1.2 MW, 1.8 MW, and 9.35 MW increases in system imbalance, respectively. PJM shows no significant effect, likely due to less granular data. They also show that sudden increases in wind generation are more likely to cause oversupply (positive ACE), while sudden drops lead to undersupply (negative ACE), confirming asymmetric operational impacts.

In SPP, moving from the 50th to the 90th percentile of wind variance increases hourly ancillary service costs by $505, relative to a median cost of $9,250. These costs rise faster at higher levels of wind variability—meaning the system becomes disproportionately more expensive to manage as wind becomes more erratic.

Godwin and Oliver conclude that while wind intermittency imposes measurable costs, these may decline as wind capacity expands and becomes more geographically dispersed. They recommend incorporating ACE-based metrics into policy and market design, expanding ancillary service markets, and investing in forecasting, storage, and demand response. As they note, “any increase in ACE is undesirable,” and minimizing it is essential for maintaining power quality and system reliability.

This summary was written with the assistance of Microsoft Copilot on June 5th, 2025.   Its content was edited and verified by EPIcenter staff and affiliates. The podcast was created with Google NotebookLM, May 16th, 2025

Read the full paper: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4745480