Using Probability of Exceedance to Compare the Resource Risk of Renewable and Gas-Fired Generation

Publication Type


Date Published




Among the many risks surrounding long-term investments in power plants, resource risk is one of the most difficult to mitigate, and is also a risk that manifests differently for renewable and natural gas-fired generation. For renewable generators like wind and solar projects, resource risk is primarily a quantity risk—i.e., the risk that annual energy production will be less than expected due to a weaker-than-expected wind or solar resource. Conversely, for gas-fired combined cycle generators, resource risk is primarily a price risk—i.e., the risk that natural gas will cost more than expected. 

Most often, resource risk—and natural gas price risk in particular—falls disproportionately on utility ratepayers, who are typically not well-equipped to manage this risk. As such, it is incumbent upon utilities, regulators, and policymakers to ensure that resource risk is taken into consideration when making or approving resource decisions, or enacting policies that influence the development of the electricity sector more broadly.

This paper presents a new framework, grounded in statistical concepts related to “probability of exceedance,” to incorporate resource risk into utility decision-making processes.  Probability of exceedance is already widely used in the renewable energy industry to characterize the uncertainty around central or “P50” estimates of annual energy production at wind and solar projects, and within the natural gas sector to estimate the likelihood that natural gas prices will exceed certain levels in the future. The new framework developed in this report merges these two distinct applications to create a comparable range of probabilistic resource risk projections for both renewable and gas-fired generators.

Importantly, these probability distributions have markedly divergent characteristics. The distribution of projected natural gas prices is narrow over short time horizons but widens over longer time horizons, reflecting the fact that it is easier to predict where natural gas prices will be several months from now than several years from now. Conversely, the distribution of projected annual energy production from a wind or solar project is fairly wide in any single year but narrows over longer time horizons, due to mean reversion in the random inter-annual variability of the wind or solar resource.

As a result, even when gas-fired generation has a lower levelized cost of energy (LCOE) than wind and solar power on an expected or P50 basis—i.e., the basis on which these resources are most often compared—comparisons that are instead based on worse-than-expected outcomes often reach the opposite conclusion: that wind and solar have a lower LCOE than gas-fired generation. This is particularly the case the longer the time horizon considered and the greater the level of risk aversion.

Year of Publication



A webinar presentation of this report recorded on March 29, 2017, can be viewed here.


Research Areas

Related Files