Skip to main content

How Much Does Wind Power Reduce \(\text {CO}_{2}\) Emissions? Evidence from the Irish Single Electricity Market

Abstract

This paper evaluates the effect of wind generation on \(\text {CO}_{2}\) emissions using 2008–2012 historical data for the Irish Single Electricity Market. Wind generation displaces \(\text {CO}_{2}\) emissions, as expected, in line with the average system emissions. Over the whole period, wind generation avoided about 8.8 million tons of \(\text {CO}_{2}\), equivalent to about 12% of total system emissions. To understand what drives the level of abatement we evaluate the results by technology and determine that wind generation has similar effects on total emissions from CCGT and coal plants, due to the higher carbon content of coal. Each MWh of wind, however, replaces more generation from CCGTs than from coal plants, in proportion to their generation. We also test the hypothesis that as wind displaces baseload plants it pushes them to generate less efficiently, but find no evidence of a strong negative effect of wind on CCGT or coal plant efficiency. Finally wind displaces about 2.5% fewer emissions when the pumped storage plant is on outage, suggesting that wind is more effective when paired with a flexible system.

This is a preview of subscription content, access via your institution.

Fig. 1

Source: a CCGT generation data: SEMO; wind data: EirGrid and SONI.

Fig. 2
Fig. 3

Notes

  1. As of early 2017 the SEM is expected to change significantly to comply with the EU Target Model by May 2018 through the adoption of the I-SEM (EirGrid 2016).

  2. The SEMO variable ‘load’ is not a good proxy for demand. For example it excludes imports and exports, includes pumped storage demand and excludes demand that is met by plants that do not bid directly in the SEM.

  3. Details for the Republic or Ireland are at http://www.eirgrid.com/operations/systemperformancedata/systemdemand/.

  4. www2.nationalgrid.com/UK/Industry-information/Electricity-transmission-operational-data/Data-Explorer/.

  5. Peat plants were historically subsidised in Ireland to maintain the employment of peat cutters. These subsidies are designed to be phased out over time.

  6. In our case, the estimated coefficients are very similar in the two specifications. Results for the dummy variable specification are available from the authors.

  7. We also check the results with an AR(5) process. The results are qualitatively similar (results available from the authors). In the results section, we present the AR(1) specification for consistency with the rest of the paper.

  8. The (average) wind coefficient is larger than the one reported in Wheatley (2013) for 2011. The results differ for a few reasons. In 2011 the system operated without the pumped storage plant and the interconnector (the latter for part of the year). However, our estimated wind effect remains larger even when we limit the analysis to 2011. The main driver of the different results is that Wheatley (2013) uses the load information from SEMO, whereas we use the information from EirGrid and SONI, which we argue in Footnote 2 is a better measure of demand. We also analyse the whole SEM instead of focusing on the Republic of Ireland and use a different econometric specification.

  9. We also run a robustness check including the aggregate availability of all coal plants as an explanatory variable. If lower availabilities of coal are statistically correlated with higher or lower displacements of \(\text {CO}_{2}\), we may respectively over or underestimate the effect of wind. Including the aggregate availability of coal plants does not substantially change the average effect of wind, which stays at −0.48, as in the main reported results.

  10. The poisson distribution can be used for continuous dependent variables, even with overdispersed distributions. In the latter case the standard errors of the coefficients must be corrected: in Stata the correct procedure is to use the vce(robust) command. See for example Wooldridge (2010), Cameron and Trivedi (2010).

  11. This is implemented with the STATA14 lincom command. The delta method takes the first order Taylor approximation of the mean of the considered variables, and then calculates their variance.

References

  • Benitez LE, Benitez PC, van Kooten GC (2008) The economics of wind power with energy storage. Energy Econ 30(4):1973–1989

    Article  Google Scholar 

  • Cameron C, Trivedi P (2010) Microeconometrics using Stata, revised edition, 2nd edn. Stata Press, College Station

    Google Scholar 

  • Cullen J (2013) Measuring the environmental benefits of wind-generated electricity. Am Econ J Econ Policy 5(4):107–33

    Article  Google Scholar 

  • DCENR (2012) Strategy for renewable energy: 2012–2020. Technical report, Department of Communication, Energy and Natural Resources

  • Denny E, O’Malley M (2006) Wind generation, power system operation, and emissions reduction. IEEE Trans Power Syst 21(1):341–347

    Article  Google Scholar 

  • DETI (2010) Energy: a strategic framework for Northern Ireland. Technical report, Department of Entreprise Trade and Investment

  • Di Cosmo V, Lynch MA (2016) Competition and the single electricity market: Which lessons for Ireland? Utilities Policy 41:40–47

    Article  Google Scholar 

  • EirGrid & SONI (2012). 2011 Curtailment report. Technical report

  • EirGrid & SONI (2013). 2012 Curtailment Report. Technical report

  • EirGrid & SONI (2015) Delivering a secure, sustainable electricity system (DS3): Programme Overview—2015. Technical report, EirGrid and SONI

  • EirGrid Group (2016) Quick Guide to the Integrated Single Electricity Market, version 1. Technical report, EirGrid

  • Greene WH (2003) Econometric analysis. Pearson Education India, New Delhi

    Google Scholar 

  • Howley M, Dennehy E, O’Gallachoir B, Holland M (2012) Energy in Ireland 1990–2011. Technical report, SEAI

  • Kaffine DT, McBee BJ, Lieskovsky J (2013) Emissions savings from wind power generation in Texas. Energy J 34:160–180

    Article  Google Scholar 

  • McInerney C, Bunn D (2013) Valuation anomalies for interconnector transmission rights. Energy Policy 55:565–578 Special section: Long Run Transitions to Sustainable Economic Structures in the European Union and Beyond

    Article  Google Scholar 

  • Meibom P, Barth R, Hasche B, Brand H, Weber C, O’Malley M (2011) Stochastic optimization model to study the operational impacts of high wind penetrations in Ireland. IEEE Trans Power Syst 26(3):1367–1379

    Article  Google Scholar 

  • NISRA (2013) Northern Ireland environmental statistics report. Technical report, Northern Ireland Statistics and Research Agency

  • SEAI (2014) Renewable Energy in Ireland 2012. Technical report, Sustainable Energy Authority of Ireland

  • Traber T, Kemfert C (2011) Gone with the wind? Electricity market prices and incentives to invest in thermal power plants under increasing wind energy supply. Energy Econ 33(2):249–256

    Article  Google Scholar 

  • Troy N, Denny E, O’Malley M (2010) Base-load cycling on a system with significant wind penetration. IEEE Trans Power Syst 25(2):1088–1097

    Article  Google Scholar 

  • Wheatley J (2013) Quantifying \({CO}_2\) savings from wind power. Energy Policy 63:89–96

    Article  Google Scholar 

  • Wooldridge JM (2010) Econometric analysis of cross section and panel data, 2nd edn. The MIT Press, Cambridge

    Google Scholar 

Download references

Acknowledgements

The authors thank an anonymous referee and the editor for comments that greatly improved the paper, in addition to participants to the 7th TransAtlantic Infraday conference and the 7th workshop in Empirical methods in Energy Economics. The authors are responsible for any remaining omissions or errors.

Funding  Funding from the ESRI Energy Policy Research Centre is gratefully acknowledged. Valeria Di Cosmo acknowledges funding from Science Foundation Ireland, Grant No. SFI/09/SRC/E1780. The opinions, findings and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the Science Foundation Ireland.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Laura Malaguzzi Valeri.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Statement of Exclusive Submission

This paper has not been submitted in similar or identical form elsewhere, nor will it be during the first three months after its submission to the Publisher.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1 (pdf 180 KB)

Appendices

Appendix A: Availability Data

We use data on plant-level availability from SEMO. We clean the data so that plant availability is:

  1. 1.

    Never larger than maximum capacity (allowing for 10% tolerance);

  2. 2.

    Never 0 when the plant is actually generating;

  3. 3.

    Interpolated from non-missing data when it is missing.

Some of the data is missing and some is registered as 0 even when a plant is generating, which can occur for a couple of reasons: a. availability is registered as 0 for system operation reasons. For example a thermal plant that is associated with a windfarm location is defined as unavailable according to the SEM. b. Some of the data might also be registered as 0 when data providers (plant operators) enter 0 instead of missing.

EirGrid publishes monthly availability for Republic of Ireland (ROI) plants (www.eirgrid.com/operations/systemperformancedata/availabilityreports/#d.en.797). We make sure that where information on availability is missing, the interpolated version is compatible with the EirGrid availability reports.

Appendix B: Daily Fixed Effects

In this section we report the results when the equations include daily fixed effects (and therefore exclude variables collected at the daily level, such as fuel and \(\text {CO}_{2}\) prices). This specification does not change the results significantly with respect to the results reported in the main text. In particular, the average effect of wind on system emissions is −0.49 tonnes of \(\text {CO}_{2}\) displaced, compared to −0.48 with month-year fixed effects.

As in the results reported in the main analysis, we implement fixed effects by taking the first difference with respect to daily averaged. To account for the remaining autocorrelation, we also impose an AR(1) specification on the residuals.

Table 6 Effect of wind on system \(\text {CO}_{2}\) emissions (tonnes) with daily fixed effects

Tables 6 and 7 show the results of the specification with daily fixed effects for CCGT and coal emissions and generation. The results do not significantly differ from those reported in the main text. Table 8 reports the results on the analysis of the effects of wind on average efficiency of CCGT and coal plants with daily fixed effects.

Table 7 Effect on generation (MWh), by technology with daily fixed effects)
Table 8 Effect on efficiency (%), by technology with daily fixed effects)

Appendix C: Other Generation Technologies

Combustion turbine, oil, distillate and OCGT technologies each account for less than 2% of demand during our period, on average. There are a large number of periods when these plants do not generate. Even when they do generate, the distribution of emissions is far from normal. This leads to challenges in identifying a robust specification to estimate the effect of wind on the emissions of plants using these technologies. In particular, specifications that do not account for the large number of zeroes in the dependent variable can lead to biased coefficients.

The following tables show the results of three different specifications: a simple OLS, an OLS with AR(1) residuals and a two-step hurdle model, that helps us address the abundance of zeroes and the highly skewed distribution of non-zero values. In the hurdle model the first step estimates the probability that each technology generates a positive amount of electricity and the second estimates the model conditional on there being positive generation, as shown by the following equation:

$$\begin{aligned} \left\{ \begin{array}{rl} Prob(Emissions>0|\mathbf {X}) = F(\mathbf {X})&{} \\ Emissions_{t}= G(\beta \mathbf {X'})+ \mathbf {\epsilon } &{} \text{ if } \quad Emissions > 0 \end{array} \right. \end{aligned}$$
(4)

where X is the matrix of explanatory variables and \(\epsilon \) is the error vector. X includes wind generation, load, the availability of other plants, net imports, fuel and \(\text {CO}_{2}\) prices, outages at Turlough Hill, Moyle and their interactions with wind generation.

In the first part, the probability of generating and therefore emitting \(\text {CO}_{2}\) is captured by a probit. The second part is modelled with a Poisson distribution with overdispersion, to account for the right skewness of the distribution of oil, distillate, CT and OCGT plants.Footnote 10

All other control variables for both the first and the second steps are the same as in Eq. 3, including month-year fixed effects. We include wind and load in levels instead of quartiles, since the Poisson specification is itself non linear, implying a non-linear effect of wind and load even when they are included in levels.

For the OLS and AR(1) specifications, the standard errors of the marginal effects are calculated using the delta method.Footnote 11

The following tables report the estimated coefficients for wind and load (in levels) for the hurdle model, the OLS and the OLS with AR(1) specifications. For the hurdle model, we report the estimates for the second step. Complete results are available from the authors upon request.

Table 9 Wind effect on tonnes \(\text {CO}_{2}\), other technologies, 2008–2012
Table 10 Load effect on tonnes \(\text {CO}_{2}\), other technologies, 2008–2012

We explore the effect on generation also for hydro, which is not associated with measured \(\text {CO}_{2}\) emissions. We do not study the effect on pumped storage, where generation decisions are more complex as they have to address both when to generate and when to recharge and therefore use electricity.

Table 11 Wind effect on generation (MWh), other technologies, 2008–2012
Table 12 Load effect on generation (MWh), other technologies, 2008–2012

Tables 91011 and 12 above show that across the specifications the estimated coefficients are fairly similar. For distillate and oil, the technologies with the lowest number of periods with positive generation, the two-part model shows a larger effect of wind and load, as expected. The second step of the hurdle model excludes periods with 0 generation. For all these technologies marginally increasing wind generation has a similar effect to marginally decreasing load. The largest effect of wind is on CT and oil plant generation and emissions.

When we examine the effect on generation of hydro, we see that wind has a smaller effect than load. The effect of wind on hydro generation is not estimated robustly. The two-step model suggests no effect, in line with expectations, whereas the OLS and OLS with AR(1) suggest respectively a positive and a negative effect. One explanation is that to model hydro generation more accurately, we would have to be consider additional variables, as it depends for example on rainfall.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Di Cosmo, V., Malaguzzi Valeri, L. How Much Does Wind Power Reduce \(\text {CO}_{2}\) Emissions? Evidence from the Irish Single Electricity Market. Environ Resource Econ 71, 645–669 (2018). https://doi.org/10.1007/s10640-017-0178-8

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10640-017-0178-8

Keywords

  • Carbon dioxide emissions
  • Electricity
  • Ireland
  • Wind generation
  • Storage