“The level of emissions savings provided by wind plants has never been conclusively determined, taking into account all the factors.”
Part I yesterday questioned the analysis and robustness of Joseph Cullen’s study, “Measuring the Environmental Benefits of Wind-Generated Electricity”.  Part II completes the commentary on this paper, covering:
The level of emissions savings provided by wind plants has never been conclusively determined, taking into account all the factors. Further, there is no published accurate, minute-by-minute, actual fuel consumption or emissions by individual plant, especially for systems with notable levels of wind present. Note the limitations in the Katzenstein and Apt paper looked to by Cullen for corroboration as discussed in Part I.
In general, government reported emissions are estimates based on calculations using assumptions and relatively simple algorithms. In some cases, actual measurements are taken but are no better than those calculated as reported by the International Energy Agency (see page 35).
“Commercial instrumentation is available for monitoring CO2 concentration and flue gas volume flows. Given the limitations of such instrumentation, the accuracy of directly measured CO2 release is probably no better than that derived by indirect calculation.” (emphasis added)
A report by The Sustainable Energy Authority in Ireland, “Renewable Energy in Ireland”, in Appendix 1 also refreshingly recognizes the limitations to existing reporting methods.
“The assumption underpinning this approach is that the renewable plant is displacing the last plants to be dispatched to meet electricity demand, i.e. the marginal oil and gas plants. There are clear limitations in this analysis but it does provide useful indicative results.” (emphasis added for “indicative”, which is taken to mean “suggestive”)
“The limitations and caveats associated with this methodology include that it ignores any plant used to meet the associated reserve requirements of renewables. These open cycle plants will typically have lower efficiency and generate increased CO2 and NOx emissions compared with CCGT and these emissions should be incorporated into the analysis. The purpose of presenting a simplified analysis here is to provide initial insights into the amount of fossil fuels that are displaced by renewables and the amount of emissions thereby avoided.” (emphasis added)
The issue raised in the last quote speaks to the comments made in the Robustness section in Part I.
The above comments point out some of the typical shortcomings of many current approaches in determining reported emissions offsets for wind. At best, such results are useful only as some indication and for rough comparison purposes, for example between jurisdictions and time periods, and not reliable for absolute levels.
The Cullen article admits that the modelling approach used, “…relies only on publicly available generator output and characteristics“. Emphasis has been added to the quote because there is no adequate, publicly available information on the constant cycling required of other generation plants mirroring wind plants’ highly random output on a short-term basis, as indicated in Katzenstein and Apt’s paper in Part I.
Cullen somewhat distances his results from the EPA system CEMS on the basis that CEMS reporting includes less than two-thirds of the generators in the ERCOT system, a notable comment in itself.
“As a robustness check, I estimate the same model with hourly emissions data from the EPA’s Continuous Emissions Monitoring System (CEMS) as the dependent variable. Using CEMS data may be able to account for the changes in the emissions rate due to efficiency changes, though it may exacerbate ramping effects.”
The latter part of this quote needs elaboration. Cullen shows that using the CEMS data in his model results in 4% lower CO2 emissions offsets for wind. He concludes that this indicates relative robustness, but the CEMS data “may” exacerbate ramping effects. This implies that Cullen has in fact captured the ramping effects, which is very questionable as already described.
However, CEMS reporting is subject to question as indicated above. Also note that the CEMS data is hourly based, which likely masks ramping effects on much shorter time intervals in which electricity system balancing must operate to ensure system reliability. In summary the CEMS information more likely understates the ramping effect.
So questionable or incomplete data is a problem in the determination of complete and accurate results, and any corroboration claimed by Cullen is questionable.
Interstate Trade in Electricity
This is a factor that is almost always overlooked in analyses of wind performance, and is not taken fully into account in the modelling here. The general reason for this is that electricity exports to or imports from another jurisdiction are a somewhat complicated matter, and unfortunately often are taken to be relatively inconsequential, which they usually are not.
When Cullen is talking about electricity imports, it is presumed he is talking about net interstate trade, as opposed to import/export of electricity in connection with another country. For simplicity here, the terms “interstate trade” and “exports/imports” will be used to refer to interstate trade only. Further it is important to be clear whether or not any reference to this is a net number of exports/imports as is often the case in reported values.
Cullen also says that he observes the flows of electricity over connection lines in neighboring grids in 15 minute intervals and claims that less than 1 percent of daily generation is exchanged with other grids, while wind accounts for approximately 2%, which allows him to restrict his analysis to within the Texas system. However, this does not agree with EIA reports,  as summarized in Table 1.
Table 1 – Wind Production and Net Interstate Trade as Percent of Total Texas Electricity Production.
Note that total wind production is at the same level as net interstate trade for two of the three years that Cullen analyzed, but the net of the interstate trade and could conceal larger amounts of export and import levels over the same period.
Further, Cullen talks in terms of imports, but in the three years analyzed, Texas was a net exporter of electricity. Note the balancing of Supply and Disposition in the reference shown for Table 1. Some clarification of this by Cullen would have been helpful.
In summary, notwithstanding the relative isolation of the Texas electricity system, interstate trade in electricity cannot be ignored in the analysis, and exports/imports even on a net basis could just as easily account for much of the reductions in fossil fuel plant emissions as wind.
To further examine misunderstandings in connection with inter-jurisdiction trade in electricity, in footnote 30 for imports, Cullen suggests that if it was assumed that the emissions offset profile of imports to be the same as in the models, this would change the results found.
It is preferable to assume imports carry no emissions, as it would be very difficult to identify the specific source of the electricity generation profile behind the exported electricity. For the purposes of simplicity in explaining this, assume that such distinction can be made in a couple of simple cases.
The exporting jurisdiction is exporting electricity associated with emissions. For the importing jurisdiction to also be charged with these emissions would be double accounting, unless the exporter took a balancing credit. Imagine the complex negotiation associated with this arrangement.
In the case of exporting electricity from non-emissions producing generation there is no need to associate emissions with either the export or import as well. However there is an example where this is not done. Denmark reports emissions in two ways: (1) as produced, and (2) after taking credit to reduce its actual emissions based on the amount of exported wind production.
The latter view is often cited in error. See Peeling away Onion of Danish Wind for more details. Here as above, this works if the receiving jurisdiction takes a balancing increase in reported emissions, which is unlikely, resulting in a double accounting for emissions reductions, an undesirable outcome.
Admittedly inter-jurisdiction trade in electricity is somewhat complicated, and the above descriptions are simplifications to illustrate the need to view associated emissions as staying within the electricity generating jurisdiction. Imports/exports from and to another jurisdiction should be treated as emissions free.
Perhaps the most important matter is that in most cases, even in Texas, imports and exports of electricity must be fully taken into account when analyzing the emissions impact of wind presence.
The paper contains appropriate caveats including:
These properly acknowledged limitations reduce the value of the paper in providing useful insights and reliable conclusions about the effects of deployment of wind power.
There are a number these, which further reduce the value of this paper.
The statement, “In fact, nearly all costs associated with wind power production are incurred during the construction and installation phase of a wind farm.” is not correct. Substantial additional costs are incurred by the presence of wind in (1) otherwise not needed, dispatchable capacity to balance wind’s persistently, erratic behavior within short periods of time (minutes or less) and unreliability on a longer term basis of hours and days, and (2) in substantial increases to the grid unique to wind to gather wind’s dispersed generation, transmit it to typically distant demand centers and support demand management in distribution systems. See the series on “Wind Consequences” for more information.
A modern 1 MW (not 1 MWh) wind turbine does not require only “roughly $1 million” to install. Overnight implementation costs are over twice this amount according to the EIA.
Cullen assumes a Wind turbine life of 20 years and that any change in operating efficiency over its lifetime is negligible. This is contrary to experience that shows substantial reductions in performance measured by load factor (aka capacity factor). Figure 2 shows experience in Denmark and the UK. 
Figure 2 – Performance Degradation of Wind Turbines in Denmark and the UK
The values shown here are weighted by capacity, because this better captures the effect of the most numerous wind turbine sizes, and produces more typical results than simple averaging. The years with no values are because of little or no experience. Outages due to mechanical failure are included in these statistics.
Most of today’s wind turbine installations have occurred since 2000. These are typically 1-3 MW and much larger than their predecessors. These larger wind turbines have a 100 ton  blade assembly and nacelle enclosing the generator on top of a 200 foot, or more, tower. They are reported to be the largest rotating structures in the world. A possible explanation of Denmark’s slower degradation experience may be because most were installed before 2000 and the majority are of a much smaller size.
There are many considerations behind these numbers, but a strong case is made for concern about considerably shorter than generally expected life times for wind turbines, as indicated by the quote from this paper:
“With such low levels of performance it seems very unlikely that large wind farms will continue in operation beyond 10 years of age, with a strong likelihood of re-powering at that point. The consequence is that large scale reliance upon wind power seems likely to involve a regular – and costly – commitment to upgrading major components of the wind turbines.”
There are other reports on frequent, major component failures requiring substantial costs in the order those for the initial implementation. 
On page 112, Cullen refers to a paper by Holland and Mansur, which describes how real time retail pricing would shape demand by reducing peaks and shifting load to low cost periods at night. Holland and Mansur explain that this can have an impact on emissions, but whether the impact is positive or negative depends on the generation plant profile of the jurisdiction. Cullen claims that wind plants likewise reshape demand, which is not obvious at all. Apart from the fact that there is no mention by Cullen of the necessary attendant real-time retail pricing, wind usually has higher production at night, and it thus tends to shape low cost production base load generation (not demand) in this period. Further he states that this re-shaping of demand necessarily leads to emissions offsets, which is different from Holland and Mansur’s conclusions which allows for the possibility of increased emissions as well.
As described above in the Analysis Approach section, a major effect of wind is its impact on short term residual demand variances, a totally different type of demand shaping. This shapes the residual demand in a non-beneficial way and arguably increases emissions overall.
Cullen states, “However, for both economic and environmental reasons, hydro facilities are unlikely to spill water over dams without generating electricity.” This is clearly an inappropriate assumption and demonstrates his emphasis on market participants determining electricity system operations. The Bonneville Power Administration (BPA) in the Pacific Northwest provides an instructive case study of system operational issues and associated contortions required here and here.
Another related comment by Cullen is, “Both nuclear and aggregate hydropower production will be largely unaffected by the roll out of wind farms.” The use of “largely” is acknowledged, but the statement can be read to suggest non-problematic impacts. The BPA experience above speaks to the hydro aspects. Admittedly nuclear is an unlikely form of electricity generation to be offset by wind, but it does happen as discussed here.
A final very questionable claim is, “First, technology advancements in wind turbines have reduced the cost of wind power by 80 percent over the past 30 years (Wiser and Bolinger 2007).” I could not find confirmation of this claim in the referenced paper. A chart on page 21 does show project costs per kW reducing by about 65% between 1982 to 2000 and then increasing thereafter. This has a related comment in footnote 35, ” Limited sample size early on – particularly in the 1980s – makes it difficult to pin down this number with a high degree of confidence.”
Further to claim technology advancements have been the cause is questionable. The cost per kW of a wind turbine has been largely influenced by simple mechanical considerations, primarily the size of the blades and height of the tower, both of which have increased dramatically over this period.
Given the very large portion of national wealth (multiples of $trillions) necessary for the contemplated extensive deployment of wind plants, and the associated longer term risks to electricity system viability and reliability, we simply have to do better at complete and conclusive analysis of the impact of wind plants with much improved public availability of operational data on generation plants, including real time fuel consumption and accurate emissions. Fuel consumption may be the only realistic way to assess emissions.
It is questionable that sophisticated econometric modelling is appropriate for this task.
 As indicated in Part I, there is an earlier version which may be accessed at http://www.u.arizona.edu/~jcullen/Documents/measuringwind.pdf , but no attempt has been made to confirm that it is an identical document.
 Hughes, Gordon (2012). “The Performance of Wind Farms in the United Kingdom and Denmark” http://www.ref.org.uk/attachments/article/280/ref.hughes.19.12.12.pdf Chart is based on calculations from Figure 2.
 This is not widely reported, and some links to sites previously dealing with this topic appear to have removed the documents in question. Here are some currently available reports relating to this matter.