The seasonal wind profile is a poor match to the seasonal load profile, in part due to negative wind-load correlation. For wind to be the sole power source, seasonal storage would be necessary to match annual wind and load profiles.

The adjacent figure presents wind and load profiles for PJM 2016. Wind data is scaled so that annual wind production is equal to annual load. Both wind and load hourly data are smoothed using a 7 day moving average to illuminate the seasonal variation.

For a zero emission system, wind can be the sole generator if the system can draw upon 1.5 x 10⁸ MWh of stored energy available at the beginning of June. Wind energy would be stored during the preceding fall, winter, and spring, and released as necessary during the summer (shaded area).

If PJM were to build a reliable system based on this concept it would need at least 2 x 10⁸ MWh of seasonal stored energy to account for year-year variation and reserves. (200 million MWh, a 33% increase over the 1.5 x 10⁸ MWh.) Theoretically, this could be provided by using Lake Erie/Ontario for pumped hydro storage. Assuming 80% two-way efficiency, the annual Lake Erie level variation would be ~36 meters. This is not a practical concept.

Several academic studies claim that large penetration of renewables can be achieved without seasonal storage. For example Budaschak et al. argue that since seasonal storage is so expensive a least cost-system is achieved by overbuilding wind by 3x. However their system has huge seasonal power generation imbalances and does not include inter-annual variability or a credible assessment of cost and reliability. Other academic speculation (Jacobson et al.) claims that wind, hydro and solar can provide 100% of our energy needs. A primary flaw with this (and other studies) is that they are based on models not data. The models have not been validated to show that they realistically represent the variability of wind and solar across all time scales. Models often smooth variability by implicitly assuming that generation is stochastic random variable and do not correctly represent correlations with space, time and load.

A more thorough development of these arguments was developed by Ontario’s Council for Clean and Reliable Energy.