## .
Christopher NicholsSenior Analyst, National Energy Technology Laboratory, Pittsburgh, PA Gavin PickenpaughEconomist, National Energy Technology LaboratoryPittsburgh, PA
The Annual Energy Outlook 2015 (AEO’15), the premier energy market forecast produced by the U.S. Energy Information Administration (EIA), forecasts that just under 300 GW of new electricity generation capacity will be required to be built in the next 30 years. Historical capacity additions have averaged around 20 GW per year, or double the rate projected in the AEO’15 and most other energy forecasts. Given that myriad organizations – many without intimate knowledge of the energy industry - use these forecasts in their planning efforts, it is imperative that the assumptions and methodology behind these results are as accurate and reasonable as possible. Several factors indicate that the amount of new generation capacity forecasted in the AEO’15 may be much lower than would be needed to replace retiring baseload electricity generation units and to support even moderate economic growth. Although the results of the AEO’15 are the sole focus of this article, most other energy forecasting models either use AEO results for calibration or have similar methods, which are also likely to minimize the need for new generation. The National Energy Modeling System (NEMS) is the model platform the EIA uses to forecast the need for new electricity generation capacity builds, among a myriad of other model outputs related to the energy sector. NEMS uses an algorithm, based on the existing capacity and the projected growth in electricity demand, to forecast how much new capacity will be needed. Exhibit 1 shows the projected electricity generation in billion kilowatt-hours (KW/h) by fuel source from the AEO’15 Reference Case. In this Case, the total electricity generation grows at a roughly linear rate. Electricity generation from coal and nuclear, traditionally the suppliers of baseload power, remain relatively constant over the entire projection period.
Almost all of the growth in generation is provided by natural gas (due to a relatively low price path for gas) and to a lesser extent, renewable (due mainly to renewable energy standards at the state level). As seen in Exhibit 2, this generation profile is supported by new capacity builds of natural gas combined cycle and renewable energy plants (which are backed-up by the combustion turbines) later in the projection period. Also of note in Exhibit 2, is the relatively low magnitude of new nuclear and coal builds, which implies that nearly all of the electricity generation projected for these fuel sources through 2040 will be from existing power plants. The age of the average nuclear or coal plant is already around 40 years old. The combination of generation and capacity projected in the AEO’15 imply capacity factors in the range of 80-90 percent for coal and nuclear plants by 2040, when these plants will be nearing an average of 70 years old.
Although life extensions are possible for both nuclear and coal units, sustained operations at such high capacity factors are unrealistic for a fleet of power plants advancing into old age. In Exhibit 3, the average capacity factor is shown by unit age for coal power plant operations over the period of 1998 to 2012. As seen in the figure, the average capacity factor drops drastically around age 50 for coal units due to maintenance, operational issues, and profitability concerns. Nuclear units do not show a similar profile, likely because coal units with operational issues can often run at lower capacity, while nuclear units must often shut down for extended outages to deal with problems. However, it is exceedingly unlikely that the current baseload fleet will be able to perform at the levels projected in the AEO’15. The recent announcement of the early retirements of the San Onofre and Crystal River nuclear power plants demonstrates the vulnerability of nuclear units to premature and permanent shutdowns. Around 40 GW of coal-fired generation – 15% of the total capacity – has already been announced for retirement also, with NEMS only forecasting 10 GW more of coal retirements by 2040. In order to forecast the potential loss of electricity generation based on unit degradation with age, a model was constructed based on the performance of units historically. Exhibit 4 employs kernel density estimators[1] to approximate the density (i.e., distribution) of capacity factors for the various age groups.[2]
Regression analysis was used to estimate the rate of decline for units 30 years old or greater. A simple regression of capacity factor on age for units in the 30 to 60 year age group results in a statistically significant coefficient estimate of 0.9425 (see equation 1 in Exhibit 5). This implies units that are 30 years of age and older see a decline in capacity factor of around 0.94 percentage points per year. Note that regression analysis conducted for units under 30 years of age did not result in a statistically significant coefficient for age; thus the analysis was only conducted for units in the 30 and over age range. To provide a more conservative estimate of the decline in capacity factor over time, a multiple regression analysis was conducted in which capacity factor is the dependent variable and the explanatory variables include age, nameplate capacity, and the ratio of average fuel cost of the unit to the Henry Hub natural gas price. The regression results for units over 30 to 60 years of age imply that for each year a plant ages beyond 30 years, the capacity factor declines approximately 0.56 percentage points. Regression formulas (equations 1 and 2 in Exhibit 5) and results (Exhibit 6) are displayed below for units ranging in 30 to 60 years of age.
The regression results above were subsequently incorporated into a spreadsheet model, which calculates the annual generation of the remaining coal units per year based on their calculated capacity factors from 2014 to 2040. The model also removed the same capacity per year as the EIA’s 2014 AEO reference case forecasted retirements. The 2008 to 2013 average capacity factors of all currently operating coal units were used for the first year of the model (2014). The average capacity factors were used each year afterwards until the unit turned 30 years old. After age 30, a de-rate factor (as described above) was applied each year, therefore decreasing the units capacity factor, until the unit turned 70. After the age of 70 years, the unit was retired. Exhibit 7 shows forecasts of electricity generation from coal using the NETL model results compared with the AEO’15 results.
The gray bars on the graph show the annual generation using the higher capacity factor de-rate value (0.94 %pt/yr) and the blue bars show the annual generation using the lower capacity factor de-rate value (0.56 %pt/yr). The top line shows the EIA AEO 2015 forecasted coal generation for the reference case. The area in between the bars and line is the difference in generation from what EIA forecasted and the NETL model yielded. Estimated capacity installments to make up for the lost generation were calculated using an 80 percent capacity for these new builds. The model and graph show a potential for a significant capacity gap in 2040 of over 100 GW, assuming the NETL methodology is accurate. Although the exact causality between Gross Domestic Product and electricity demand is open to debate, it is clear that there is a strong relationship between the two parameters as seen in Exhibit 8.
Exhibit 9 shows the historical relationship between these two parameters and the projected data from Annual Energy Outlooks from 2005 to today. The projected GDP/electricity consumption path extrapolated from the historical relationship is continued in the AEO’05, but subsequent AEOs change the nature of this curve, with AEO’13 suggesting that far more GDP growth is possible without corresponding growth in electricity consumption.
Comparing the trends shown in Exhibits 8 and 9, it appears that NEMS does not maintain a consistent relationship between GDP and electricity generation, in spite of the overwhelming evidence to the contrary. Since the need for new capacity in NEMS is based on projected electricity demand, this reduction of the GDP-electricity relationship will mask the need for new capacity, assuming even modest economic growth; a stagnant economy will likely only require the new capacity currently projected in the AEO’15. Failure to account for an aging fleet and reducing the historical relationship between GDP and electricity demand do appear to downplay the need for new electricity generation capacity in NEMS, but it is important to examine the magnitude of the possible shortfall. In 2040, over 180 GW of the coal-fired fleet will be over 70 years old, with almost 70 GW of the fleet over 80 years old. At the same time in the nuclear fleet, using a lower assumed retirement age, almost 40 GW of existing capacity will be over 60 years old. Depending on assumed plant lifetimes, there could be 110-220 GW of baseload capacity by 2040 that simply cannot operate anywhere near the high capacity factors projected by the AEO’15, if at all.
Exhibit 10 shows the historical and forecasted growth of GDP and electricity generation, along with two “wedges”. The first wedge (labeled “Generation Gap”) projects the amount of generation that would be required if the historical relationship between GDP and electricity is maintained from the AEO’05, as shown in Exhibit 8, equivalent to around 121 GW of generation capacity at an average capacity factor. The second wedge (labeled “GDP Gap”) shows the additional electricity required if GDP grows at a yearly rate from the AEO’05. This “GDP “Gap” would require an additional 68 GW of new power plants on top of the “Generation Gap.” Combined with the shortfalls from probable age-related retirements of coal and nuclear plants, the AEO’15 could be underestimating the need for new capacity by up to 500 GW by 2040. Projections and forecasting models matter – their results guide the development of policy and regulations. The results of the major models have the power to influence the future themselves, creating a situation of the “self-fulfilling prophesy”. Optimism is typically a virtue – however, in the world of energy forecasting models, being too optimistic about critical inputs such as the lifetime of current power plants and the ability to grow GDP without electricity could lead to a critical shortage of electricity generation capacity. References United States Energy Information Administration. 2005-2015. “Annual Energy Outlook.”
Notes [1] Stata 13 was employed to construct the density estimates. Some reasons we chose to employ kernel density estimates rather than histograms are the ease in graphing multiple density estimates in one figure, as well as the smoothness of kernel density lines. For more information, see: http://www.stata.com/manuals13/rkdensity.pdf [2] Figure 4 excludes units with heat rates equal to 0 or greater than 35,000 and exclude units with capacity factors greater than 100 |
## Recent Issues |