Ameren Transmission Rate Stakeholder Presentation
Follow-up Questions from 10/17/2012 Meeting
1.
What detail can Ameren provide on the plant additions/modifications identified as AIC
Projections for 2013 totaling $358 million (Slide 30)? Please note that slide 30 shows total
capital spending in 2012 and 2013. Not all of these dollars will be in service by 12/31/13.
The table below shows that of the projects on slide 30, approximately $280 million is
expected to be in service by the end of 2013. Also note the addition of a new category for
wind farms. These projects will primarily be paid for by the wind farms and were previously
included with reliability projects.
Reliability/aging infrastructure replacement
Clearance for Planned Line Rating
Right of way expansion
Improved Reliability
GIA - Wind Farm
Total
$130
$94
$21
$23
$12
$280
million
million
million
million
million
million
2.
Does Ameren realistically expect to complete $358 million in plant additions/modifications
between 12/31/2011 and 12/31/2013 (Slide 30)? As noted in response to question 1,
approximately, $280 million will be in service by the end of 2013.
a. What is the capital expenditure split in the projected modifications between the
years 2012 and 2013? $146 million in 2012 and $212 million in 2013.
b. What amount of capital expenditure has been placed into service as of October
17, 2012 from the list of projected modifications? Ameren does not normally
track in-service during the middle of a month. However, assuming no unforeseen
delays, the current forecast shows $94 million in service by December 31, 2012.
c. What is the basis for the project plant modifications?
The projected plant is based on Ameren budget as of the end of August 2012.
d. If the modification basis is the 2012-2013 construction budget, can Ameren make
such construction budget available to its Customers?
No. Ameren does not share its budget. The elimination of ROFR has made details
regarding individual projects even more confidential. However, Attached is a list
of Ameren projects included in MISO’s current MTEP Appendix A, which is
publically available.
3.
What is the basis for a 77% increase in A&G expenses from 2011 (for ATXI) and a 14%
increase in A&G expenses from 2011 (for AIC) (Slides 16 and 24)?
a. Please provide the major reasons for the estimated increases.
ATXI - The projected 2012 A&G costs for ATXI were estimated based on the
approximate amount of A&G incurred in 2011 ($500,000). However, with more
activity occurring at ATXI this year, it is now receiving corporate allocations.
Through June 2012, actual A&G was almost $500,000. So year-end 2012 and
projected 2013 are now expected to be considerably higher.
AIC – Note that for AIC the comparison is between actual 2011 A&G versus
projected 2013 A&G. Therefore, more than half of the 14% increase is likely due
to cost increases in salaries over two year. Pension OPEB expense is also higher.
b. Can Ameren make available the estimate by FERC Account for the increased A&G
expenses for Customer review?
Ameren does not prepare an A&G forecast by FERC account.
4.
What is the basis for a 20% reduction in Account 565 credits from 2011 (for AIC) (Slide 24)?
a. Please provide the major reasons or brief details for the estimated decreases.
Account 565 is for payment to others for transmission service used by AIC to
serve its bundled customers. Most of this account is for payments to MISO for
Schedule 1 and 2 charges, as well as 26 and 26A. These payments are expected
to decrease in the future as Ameren Illinois is expected to serve less bundle load
due to municipal aggregation of residential load (i.e., more residential customers
selecting Alternate Retail Electric Suppliers).
5.
What is the basis for an overall 1% load reduction in demand (Ameren Illinois Divisor) from
2011 given that the hot weather summer 2012 load is not reflected in the June-12 load
divisor (for the 12 months ended 12/31/2011) (Slide 31)?
The current Attachment O effective June 2012 uses the 12 monthly coincident
peaks for 2011; which was generally considered warmer than normal. The
projected 2013 12 monthly coincident peaks are based on normal weather
conditions.
a. Can Ameren describe how the 2013 load forecast was prepared?
Please see explanation of the AIC load forecast methodology at the end of this
document.
b. Is a weather-normalization factor being applied to the 2013 load forecast
process? If so, please describe the methodology.
Yes . See response to a.
6.
Can Ameren provide a projection of the AMIL Schedule 9 Rate Calculation in much the same
form as Slide 31, but for year 2014?
While Ameren does not publish specific forecast data, it seems reasonable to
provide some guidance for 2014. Below is the rate calculation shown on slide 31
and a similar rate calculation comparing 2013 and 2014. Note that the 2014
should be considered preliminary and for informational purposes only. The actual
2014 rates will be based on a new internal Ameren forecast created during 2013.
Ameren does not intend to provide a future analysis comparing the preliminary
2014 calculation with the actual 2014 calculation.
AIC Adjusted Revenue Requirement
ATXI Adjusted Revenue Requirement
Total Revenue Requirement
Ameren Illinois Divisor
Annual Cost ($/kW/Yr)
Network & P-to-P Rate ($/kW/Mo)
Jun-12
105,160,840
7,179,601
112,340,441
7,256,406
15.482
1.290
Jan-13
133,641,823
7,357,158
140,998,982
7,175,041
19.651
1.638
Change
28,480,984
177,557
28,658,540
(81,365)
4.170
0.347
Percent
27%
2%
26%
-1%
27%
27%
AIC Adjusted Revenue Requirement
ATXI Adjusted Revenue Requirement
Total Revenue Requirement
Ameren Illinois Divisor
Annual Cost ($/kW/Yr)
Network & P-to-P Rate ($/kW/Mo)
Jan-13
133,641,823
7,357,158
140,998,982
7,175,041
19.651
1.638
Jan-14
146,200,000
7,400,000
153,600,000
7,281,889
21.093
1.758
Change
12,558,177
42,842
12,601,018
106,849
1.442
0.120
Percent
9%
1%
9%
1%
7%
7%
Page.Line
1.7
1.7
1.15
.
.
Page.Line
1.7
1.7
1.15
.
.
7.
Is it correct that Ameren is using a 13-month average balance for rate base items?
All rate base items are based on a 13-month average balance except for Accumulated
Deferred Income Taxes which are a simple average of the beginning and end of year
balances. (See Notes EE & FF on page 5 of AIC’s Attachment O.)
8.
What all is included in what is computed on that basis (13 month average balance):
a. Does it include plant in service and accumulated depreciation? Yes.
b. Does it include materials and supplies? Yes.
c. Does it include prepayments? Yes.
d. Does it include Accumulated Deferred Income Taxes? No.
The monthly balances are included as workpapers on additional tabs in the posted
Attachment O calculation.
9.
If any of these items are not computed on the basis of a 13-month average balance, what is
the basis for computing any such items; that is, computed based on the average of the
beginning and end of year balances?
Please see responses to 1 & 2.
10.
Can and will Ameren provide its estimate of the 13-month average balance for gross plant in
service and accumulated depreciation for transmission plant, general plant and intangible
plant for 2013?
For AIC, please see the Attachment O calculation posted on the Ameren OASIS. WP 1 Plant includes
the 13 month gross plant balances by function. WP 2 Accum Depr includes the 13 months
accumulated depreciation balances. For ATXI, please see the Projected Rate Base tab.
11.
What is the maximum prior level of plant additions for transmission facilities in the last 5
years?
$63 million.
AIC load forecast methodology from procurement plan and submitted to MISO
The monthly peak forecast for AIC’s eligible customer retail load was performed at the total Ameren
Illinois level. Historical hourly data from 2010 to 2011 was collected while the corresponding daily
temperatures were used for building the regression models. The daily temperatures are calculated by
averaging the daily high and low values. The loads were at transmission level and excluded wholesale
load.
Methodology:
Using the hourly input data from 2010 to 2011, a daily peak regression model and a daily energy
regression model were constructed. A peak and energy model for every DS class (namely DS1, DS2,
DS3A, DS3B, DS4 and DS5) was built. This is because each of these DS classes has a different weather
response function. For example, DS1 is the most weather-sensitive class. Year 2010 was taken as a
reference calendar year. The actual load for 2010 was weather normalized using the daily peak and
energy models, by adopting the Unitized Load Calculation approach. This approach is briefly
discussed below.
Unitized Load Calculation:
Using the actual hourly load data estimate the daily peak and daily average load.
Calculate the Unitized Hourly Load using the equation shown below:
Daily peak designated as: PK t (0)
Daily energy designated as:
AVG (0)
t
Unitized Hourly Load:
D
ht
(0) 
MW (0)  AVG (0)
PK (0)  AVG (0)
ht
t
t
t
The same regression coefficients are used to run-through the normal weather for daily peak and
energy.
Weather normalized daily peak designated as:
PK (0)
'
t
Weather normalized daily energy designated as: AVGt (0) '
Normalized hourly load:
MW
ht
(0) ' 
AVG (0)  D
'
t
ht
(0).( PK t (0) '  AVGt (0) ' )
Daily Peak Model
Daily peak loads were modeled using regression within the MetrixND software package. Daily peak
load was the dependent variable, and the independent variables included temperature based
variables, seasonal variables, day-type variables, calendar variables, and energy growth trend
variable. Average daily temperature, defined as the arithmetic mean of the day’s high and low
temperatures, is the basis for all of the weather variable constructions. Temperature splines are
then created from the average daily temperature variable to allow load to respond to temperature in
a non-linear fashion. These temperature splines are also interacted with seasonal and weekend
variables to allow the temperature response of load to change with respect to these variables (i.e.
Load will respond more to an 80 degree day in July than in October, and more on a weekday than a
weekend).
..
The daily peak model also includes independent binary variables representing each day of the week,
each month of the year, and major holidays. This captures the change in load that is not due to
weather variation, such as load reductions due to industrial customers and businesses that may not
operate on weekends.
Statistical tests verify that the models fit the data quite well. The R-Squared statistic, which indicates
the amount of variation in the dependent variable (load) that is explained by the model, is around
88% on an average. The Mean Absolute Percent Error (MAPE) of the models is around 4.5% on an
average, indicating that over all of the years of the analysis, the average day has a small absolute
error.
Daily Energy Model
The concept for building the daily energy models is the same as that of daily peak, except that the
dependent y-variable is the sum of hourly loads. The R-squared statistic is around 90% on an average
for the daily energy models. The MAPE is around 4%.
Forecasting Normal Weather Conditions for the Daily Peak Model
AIC defines normal for a weather element as the arithmetic mean of that weather element
computed over the 10 year period from 2002-2011. Because daily average temperature is the
weather variable of interest for the peak forecast, the daily average temperature for each date must
be averaged over the 10 year period. Unfortunately, averaging temperatures by date (i.e. all January
1st values averaged, then all January 2nd values and so on) creates a series of normal temperatures
that is relatively smooth (i.e. no extreme values) and therefore devoid of peak load making weather
conditions. To ameliorate this situation, a routine known as the “rank and average” method is
used. In this method, all 10 years of historical weather data are collected. For each summer and
non-summer of each year, the respective degree day data is sorted from the highest value to the
lowest. Then the sorted data is averaged across the 10 years, with all of the hottest days in each
summer averaged with each other. Likewise, all of the coldest days in each non-summer season are
averaged, while the mild days are averaged together.
After the weather has been averaged by the degree day rank, the days are “mapped” back to the
actual weather of the reference calendar year, from each year for the historical period. For the
forecast period, an average weather shape is used to map the degree days. This way, the “normal”
degree days follow a realistic contour. The normal temperature series is run through the daily peak
and daily energy forecast models to produce a normal peak load and a normal energy load forecast.
The year 2010 is used as the reference year. We call it the ‘Planning Calendar’. Once we have the
normal peak and energy load forecast for 2010, using the unitized load approach discussed above,
the normal hourly loads are constructed. This profile shape is extended to the future time periods
(2013 to 2018 also called the ‘Actual Calendar’) after applying suitable calendar adjustments. In order
to do this, the first step was to simulate the normal weather (from rank and average technique
discussed above) from 2013 to 2018. The next step is to replicate the 24-hour profile shape
(considered separately for each month) for each day into the forecast period, by considering the
peak producing temperature, second peak producing temperature, and so on. Thus we have a profile
shape for each day from 2013 to 2018.
Using the peak and energy models, we forecast the normal daily peak and energy loads for the same
actual calendar time period. The unitized load formula is then applied to the forecasted values to
come up with normal hourly loads for all the years from 2013 to 2018.
Final Forecast Steps
The MetrixLT software is used to apply the hourly shapes developed above under the monthly
energy sales forecast. For example, for the month of January-2012 there are 744 hourly values and
one energy forecast value. The 744 hourly values are shaped according to the energy value. Suitable
loss factors are applied to the shaped values to arrive at final hourly forecast. This is done for each DS
class separately. The final hourly system values (and hence the monthly peaks) are obtained by
aggregating the values from each DS class.
Daily peak loads were modeled using regression within the MetrixND software package. Daily peak
load was the dependent variable, and the independent variables included temperature based
variables, seasonal variables, day-type variables, calendar variables, and energy growth trend
variable. Average daily temperature, defined as the arithmetic mean of the day’s high and low
temperatures, is the basis for all of the weather variable constructions. Temperature splines are
then created from the average daily temperature variable to allow load to respond to temperature in
a non-linear fashion. These temperature splines are also interacted with seasonal and weekend
variables to allow the temperature response of load to change with respect to these variables (i.e.
Load will respond more to an 80 degree day in July than in October, and more on a weekday than a
weekend).
The daily peak model also includes independent binary variables representing each day of the week,
each month of the year, and major holidays. This captures the change in load that is not due to
weather variation, such as load reductions due to industrial customers and businesses that may not
operate on weekends.