Stretching the Duck: How Rising Temperatures will Change the Level and Shape of Future Electricity Consumption.

• Author: Rivers, Nicholas

1. INTRODUCTION

   Climate change will affect many social and economic outcomes. However, the magnitude, and in some cases the sign, of these effects is the subject of considerable debate. Thus, the need for more evidence-based empirical estimates of the potential effects of climate change is important to improve adaptation capacity as well as for understanding its costs. (1) This paper examines the effect of rising temperatures due to climate change on one such outcome: future electricity consumption.

   We consider how rising temperatures by mid- and end-century will alter both the level and timing of electricity consumption across Canada. While several studies have looked at the relationship between climate change and overall energy consumption (Isaac and van Vuuren, 2009; Davis and Gertler, 2015; De Cian and Wing, 2017; Wenz et al., 2017), a less explored area of the literature is how climate change will affect the intraday shape of consumption. The latter is of particular importance in electricity where supply must equal demand in every hour and storage is costly. Our study fills this void.

   Our results suggest a large increase across all provinces in ramping requirements--the range between minimum and maximum hourly consumption within a day--increasing the need for flexibility in future electricity systems. This finding from the demand side echoes a similar need coming from the supply side, where an increasing share of variable energy renewable resources is placing a greater importance on flexibility to meet larger ramping requirements.

   We find significant heterogeneity in temperature sensitivity across provinces and, correspondingly, regional differences in projected electricity consumption. This is most notable in projected changes to peak demand--the maximum hourly demand within a year. The largest increase is predicted to occur in Ontario, a province that is currently summer-peaking, where we project peak demand to increase by 38%. In winter-peaking provinces, such as Quebec, which rely predominantly on electric heating, we project declines in peak demand despite growth in summer demand. In most provinces, however, we project a seasonal shift from winter-peaking to summer-peaking electricity grids, turning Canada into a summer-peaking country by end-century.

   We find a relatively small increase of 4% in the overall level of electricity consumption across Canada, with rising summer demand offset by a reduction in winter demand. This result includes the effect of adaptation in the form of more air conditioner penetration and at the most extreme temperature scenario. This stands in contrast to much larger projected increases in future electricity demand from studies in warmer climates (Isaac and van Vuuren, 2009; Akpinar-Ferrand and Singh, 2010; Davis and Gertler, 2015). This result highlights the mitigating effect of a warming climate in a cold country such as Canada, whereby increases in summer cooling demand are largely offset by decreased electric heating demand from warmer winters.

   Our empirical analysis consists of two parts. First, we estimate the relationship between temperature changes and electricity consumption to create temperature response functions, i.e. the marginal effect of temperature on electricity consumption. We then combine our estimated temperature response functions with projections of future temperature from an ensemble of global climate models under various emissions scenarios to project changes to the level and timing of future electricity consumption by mid- and end-century.

   To estimate the causal relationship between temperature and electricity consumption, we draw on public and private data sources to construct an original dataset of hourly observations of electricity consumption for every Canadian province over the period 2001-2015. (2) We find temperature response functions characterized by a familiar U-shaped relationship: at colder temperatures, rising temperature leads to decreased electricity consumption due to less need for heating; whereas at warmer temperatures, rising temperature increases demand for cooling services and thus electricity. However, these estimates represent only the short run response, i.e. the assumption that future behaviour and technology matches that of today--an unsatisfactory result for long run projections.

   To incorporate one potential dimension of adaptation, we exploit the significant heterogeneity in temperature responsiveness across provinces. These differences correspond to key observed differences in underlying ways electricity is used across provinces--differences in air conditioner and electric heat penetration, and residential share of total consumption. Re-estimating temperature response functions based on these key observables, allows for the estimation of future temperature responsiveness at various counterfactual levels that reflect potential adaptation in the form of air conditioner uptake. We then inform our adaptation-inclusive scenarios by estimating a model of air conditioner adoption using household-level microdata. We find by end-century, under most emission scenarios, residential air conditioner penetration reaches well above 90% in most provinces.

   Combining our model of air conditioner adoption with the above temperature response functions delivers long run adaptation-inclusive demand projections--in effect, at higher levels of air conditioner penetration, electricity demand becomes more responsive (i.e. increases more) at hotter temperatures. Our results show a warmer climate leads to an increase in summer demand, an increase in peak hour demand in summer peaking regions and a shift to summer-peaking more generally, an expansion of the minimum to maximum intraday range of demand, and an overall--albeit small for Canada as a while--increase in average demand.

   It is important to note that our results throughout this paper take the form of ceteris paribus projections. That is, we estimate the impact of changes in temperature on electricity consumption and on air conditioner adoption, holding all else equal. Of course, over the long time horizons we consider, many other variables will affect electricity demand, and changing temperatures will affect many other variables in addition to electricity demand. Our results should therefore be taken as the marginal effect of changing temperature on electricity consumption holding other factors constant, not as unconditional predictions of future electricity demand.

   Our paper contributes to a new and growing literature, building on three recent studies that explore the effect of climate change on electricity demand. First, in terms of regional heterogeneity, our paper finds similar results as Wenz et al. (2017): rising temperatures do not significantly increase electricity consumption in a cold country, such as Canada. However, whereas Wenz et al. (2017) focus on regional heterogeneity driven by large climatic differences between southern and northern European countries, our paper finds differences in projected electricity consumption changes within Canada, despite relatively similar climatic conditions across provinces. Instead, we find heterogeneity driven by large variation in temperature-sensitive uses of electricity. Our finding emphasizes the importance of understanding underlying drivers of temperature-sensitive demand.

   Second, Auffhammer et al. (2017) emphasize the importance of looking beyond average effects in difficult-to-store electricity, projecting changes in both average and peak demand. We extend this by using hourly granularity to estimate changes in the intraday shape of demand. This aspect is particularly important for electricity systems already grappling with large swings in intraday supply from a growing share of renewable resources. Considerable attention has been paid to the electricity "duck curve", so-named due to the shape of intraday net demand characterized by a midday belly of low net demand when solar is generating at its fullest, followed by a steep ramp in the late afternoon having the appearance of a duck's neck (CAISO, 2016). (3) Our results provide evidence of the need for even more flexibility to manage greater intraday variance coming from the demand side as well.

   Third, we develop a tractable method to incorporate both the intensive and extensive margin of adaptation (in one dimension only--air conditioner adoption) into future projections of temperature-induced demand changes. Similar to Davis and Gertler (2015) we model the adoption of air conditioners in response to changes in temperature using household-level microdata, which can be used to project future air condition penetration in a warming climate. However, whereas Davis and Gertler (2015) use this information to project future consumption by assigning a temperature response function from a region with currently high air conditioner penetration levels, we estimate temperature response directly as a function of air conditioner levels and other temperature-sensitive observables. This innovation allows us to use the projected air conditioner penetration levels directly, while maintaining region-specific characteristics, to project future electricity consumption changes with adaptation.

   In a recent paper, Auffhammer (2018) exploits significant cross-sectional variation at the household level to estimate the relationship between temperature sensitivity and extant climate conditions. In doing so, this approach provides a reduced form method to incorporate adaptation by making temperature response a function of prevailing climate. This is a promising straightforward approach with the requirement of significant cross-sectional data. Our method is comparable and both papers seek the same thing: the effect of changing climate on electricity consumption, incorporating elements of adaptation. Our method unpacks the relationship by decomposing the change into its components: the direct...