Intermittency and C[O.sub.2] Reductions from Wind Energy.

AuthorKaffine, Daniel T.
  1. INTRODUCTION

    In light of the dramatic worldwide growth in renewable electricity, particularly wind, there is substantial interest in understanding the costs and benefits of these technologies. U.S. electricity generation from wind has grown from less than 1% in 2007 to more than 6% in 2017 and is likely to continue to grow as costs continue to fall. One longstanding area of concern is that renewable technologies such as wind and solar are intermittent, in contrast to conventional electricity sources that can be dispatched as needed. Intermittency can raise the social costs of renewable technologies (Gowrisankaran et al., 2016), and the need to balance renewable intermittency with conventional backup (e.g. coal and gas) may also affect the emissions savings potential of renewable technologies.

    In principle, renewable generation reduces emissions through offsetting of fossil generation that would have otherwise been used to meet a given level of electricity demand. However, matching the variability of renewables requires the emissions-intensive process of "ramping" of generation from fossil fuel generators, potentially undercutting the emission savings per unit of wind or solar generation. Given emissions reductions, C[O.sub.2] in particular, are a primary economic justification for the substantial policy interventions supporting renewable energy (Ambec and Crampes, 2015), it is crucial to understand the extent to which intermittency may undercut emissions savings from wind generation. As such, this paper asks: How does the grid respond to wind generation and intermittency? Does wind intermittency reduce the C[O.sub.2] savings associated with wind generation? What is the magnitude of this effect, and to what extent does it undercut the economic justification for renewable policies? How will intermittency affect C[O.sub.2] savings as wind capacity grows?

    A unique feature of this study is the use of 5-minute generation data from the Southwest Power Pool (SPP). (1) These 5-minute data provide a high-frequency look at the intra-hour evolution of the generation mix. In particular, they allow statistical comparisons of emissions in two otherwise identical hours (including the same level of wind generation), but with different levels of intra-hour wind intermittency. Given the plausibly exogenous variation in intermittency, we can interpret our estimates as the causal impact of intra-hour intermittency on the emission savings from wind. (2) Of course, how intermittency affects the emissions offset by wind will likely vary across different grids depending on the generation mix and institutional features. Nonetheless, our use of observational, historical data allows us to causally identify how intermittency affected emissions savings in SPP.

    To our knowledge, this is the first study to empirically identify the effects of intra-hour intermittency on emission savings. In a related study, Dorsey-Palmateer (2019) provides empirical evidence from Texas that intermittency over longer time spans (5 hours) shifts the grid from coal to natural gas, generating a reduction in emissions through a compositional effect. Wheatley (2013) examines 30-minute data in Ireland and argues intermittency substantially reduces emissions savings, but does not causally identify its impact. Di Cosmo and Valeri (2018) also examine the Irish market, but find no evidence of a strong negative effect on thermal plant efficiency and thus emissions. Graf and Marcantonini (2017) examine the impact of increases in intermittent renewables on thermal plant annual emission rates and find evidence of modest increases in emission factors, though they are unable to separate the specific effects of intermittency from other channels by which increased renewables may affect emission rates (e.g. heat rate changes from merit order effects).

    More broadly, there is a larger literature that has typically relied on a broad class of simulated dispatch models, with varying degrees of realism and optimizing behavior. These studies have typically found that increased renewable generation reduces emissions, with the magnitude of reductions depending on the mix of generation and other features of the modeled electricity system (Lamont, 2008; Benitez et al., 2008; Maddaloni et al., 2009; Lueken et al., 2012; Scorah et al., 2012; Gutierrez-Martin et al., 2013; Gowrisankaran et al., 2016). (3) However, quantifying the effect of intermittency itself on emission savings from renewables has received less attention. One exception is Gutierrez-Martin et al. (2013) who focus on the effects of wind intermittency on emission savings in Spain, and find little evidence intermittency substantially reduces emission savings. (4) As a complement to the above studies, our data and approach allows us to estimate and identify the impact of intermittency on emissions savings without making assumptions about grid operator behavior or plant operations. That said, a key limitation of our approach is that we can only examine the intermittency effects over ranges of wind penetration observed in the data for SPP, and the above methods may be more appropriate for considering the effects of intermittency on emission savings for higher levels of wind penetration in other energy systems.

    To motivate the empirical analysis, we first begin with a conceptual model of emission savings and intermittency and how they may be measured. Empirically, we find coal and natural gas are the primary sources of generation offset by wind, whereby 1 megawatt hour (MWh) of wind on average offsets 0.52 MWh of coal and 0.37 MWh of gas. (5) Next, we show intra-hour intermittency in wind generation (measured as the intra-hour root-mean-square of changes in 5-minute generation levels) is also primarily balanced by intra-hour variation in coal and gas, and this intra-hour variation in coal and gas increases C[O.sub.2] emissions. Finally, our key parametric estimation finds 1 MWh of wind reduces C[O.sub.2] emissions in SPP by 0.726 tons holding intermittency constant, but an increase in the intra-hour intermittency of wind generation offsets emissions reductions to some extent. Similarly, our semi-parametric approach finds that in the lowest decile of intermittency, 1 MWh of wind generation reduces C[O.sub.2] emissions in SPP by 0.773 tons, while in the highest decile of intermittency, wind generation reduces C[O.sub.2] emissions by a smaller 0.703 tons per MWh.

    Evaluating the parametric point estimates at the mean values of wind generation and intermittency, marginal C[O.sub.2] emissions savings from a MWh of wind are reduced by 6.5% due to intermittency in a dynamic model that considers lagged effects, compared to a more modest 3.8% intermittency effect in a static, contemporaneous model. In terms of a simple per MWh subsidy for C[O.sub.2] savings from wind, this represents the difference between a $28.31/MWh subsidy and a $26.50/MWh subsidy for the dynamic estimates. Furthermore, while intermittency concerns will grow as wind's share of total generation increases, we find that the overall effect on emission savings is likely to remain modest in the near-term (for example, increases in wind's average share of generation from the current 10% in SPP to 20%). Thus, at recently observed wind generation shares, the concern that intermittency reduces C[O.sub.2] emissions savings is borne out, but the average magnitude of the effect on emission savings from wind is modest.

  2. EMISSIONS SAVINGS AND INTERMITTENCY

    2.1 Measuring emission savings

    This paper contributes to a growing empirical literature that measures the emissions savings from various renewable technologies, which are often supported through a variety of subsidies and other policy supports. Economic theory suggests correcting pollution externalities via a Pigouvian tax on emissions or a Pigouvian subsidy on emissions avoided can yield equal and efficient outcomes, at least to a first-order approximation. And while there has been substantial work exploring when that equivalence breaks down from a theoretical or behavioral perspective, a perhaps less obvious distinction between the two policy instruments is the issue of measurement.

    Standard theory shows the efficient tax should be set equal to the marginal external damages of emissions, which is then applied to the measured level of emissions. Though determining the marginal external damages may be challenging, the measurement of the emissions themselves is typically straightforward (from the perspective of economists)--that is to say, the measurement of emissions generated is primarily a matter of physics, chemistry, and engineering, and not something economists have much to contribute towards. (6)

    By contrast, in the context of an efficient subsidy policy, one must be able to measure the emissions avoided, and this is no longer quite so straightforward from the perspective of measurement. While one can measure the carbon dioxide (C[O.sub.2]) emissions from a coal-fired power plant's smokestack and apply a carbon tax, there is no smokestack to measure the "non-emissions" from a wind turbine or solar panel. One must determine the counterfactual level of emissions, which depends on market processes and behavioral responses, and this is a task economists are better-suited to consider. Similar challenges arise in measuring energy consumption avoided through energy efficiency adoption. As a policy-relevant, practical consequence, issues arise when policy decisions are made based on aggregating or averaging (for example, a flat subsidy per MWh) across marginal emissions savings that can vary in real-time with market conditions (Klotz et al., 2017).

    A substantial literature has emerged to measure the emissions and energy consumption avoided from various technologies, and the policy implications that follow. This is likely driven in part by the fact that subsidies are viewed as more politically palatable and more...

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT