1. Independent Load Forecast Workshop
July 25, 2016

2. Stakeholder Comments

3. Stakeholder Comments
Documents regarding multicollinearity, heteroscedasticity, first difference models, and comparison of 2015 and 2016 models are provided separately
Inclusion of 2015 data on the load forecast comparison results and results of Louisiana sales+CHP modeling are included in this presentation

4. Stakeholder Comments
We were asked a question regarding incorporating a significant change in load if LSEs provide feedback explaining the change
We can make a load adjustment to the forecast based on such changes. This should be done with care to avoid double counting (if the macroeconomic forecast already reflects it) and to ensure the allocation factors assign the load change to the proper LRZ in the state

5. Forecast Comparison 2016-2025 2015 ILF vs. Module E vs. 2015 Actual

6. Loads Comparison
Comparison charts have been made to show 2015 actual zonal peak/annual loads and projections from ILF
The historical data source (data 1) we used for the 2014 and 2015 ILF studies were discontinued
We have been provided with historical data from a different source (data 4)
Discrepancies between data 1 and 4 exist and may or may not be significant.

7. Historical Hourly Loads
Data 1
Data 4
Metered load with EE/DR/DG adjustment;
Does not include voluntary LMR data.
Used for 2014 and 2015 ILF peak analysis;
Historical hourly loads for LRZ 8,9 and 10 (prior to LRZ formation) were derived;
Contains 2010 to 2014 historical hourly data.
SERVM Historical Hourly Load;
Contains voluntary LMR data starting May 2014;
Contains 2010 to 2015 historical hourly load data;
The recorded hourly and annual loads do not exact match data 1’s records;
Peak demand may not occur at the same time as in data 1.

8. Summer Peak Discrepancies Between Data Sources (2014)
LRZ
Peak Demand (1)
Peak Demand (4)
Difference
Load Factor (1)
Load Factor (4)
2014(1) PT
2014(4) PT 1
17,018
16,937
-0.5%
67%
66%
7/21/ :00
7/21/ :00 2
11,730
11,720
-0.1%
63%
7/22/ :00 3
8,283
8,244
65%
9/4/ :00 4
9,563
9,588
0.3%
60%
8/25/ :00 5
8,487
8,507
0.2%
58%
8/25/ :00 6
17,170
17,204
68%
9/5/ :00
9/5/ :00 7
19,293
19,246
-0.2%
59% 8
7,058
7,172
1.6%
8/25/ :00
8/26/ :00 9
19,173
19,220
8/22/ :00 10
4,297
4,277
57%
8/6/ :00
8/20/ :00
MISO
114,709
114,285
-0.4%
Red indicates shift in time of peak demand

9. Forecast Comparisons ILF Peak MISO Module E Peak
Allocates state-level forecasts to LRZ levels;
Uses zonal non-coincident factors and coincidence factors to estimate zonal peak loads and system-wide peak load;
Net load: with EE/DR/DG adjustment;
Gross load: without EE/DR/DG adjustment.
Aggregates asset owner level loads to LRZ levels;
Applies intra-zonal/MISO-wide coincidence factors to convert asset owner level peak loads to zonal peaks and system-wide peak load;
Does not include DR/DG adjustment, may include EE adjustment.

10. Comparison Charts Module E – solid black line
ILF Gross – dashed black line
ILF Gross high/low bands – hashed area
ILF Net – dashed red line
ILF Net high/low bands – solid area
2014 actual value from data 1 – blue diamond; 2014 and 2015 actual loads from data 4 – red square;
Metered loads, transmission losses excluded
Historical data based on actual weather conditions; forecasts assume normal weather

11. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 1

12. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 2

13. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 3

14. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 4

15. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 5

16. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 6

17. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 7

18. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 8

19. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 9

20. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 10

21. SUFG Peak Gross & Net Forecast vs. Module E vs. 2015 Actual—MISO

22. Module E Data PY
Module E forecasts provide a series of projections of seasonal loads at the asset owner level (24 months followed by 8 years of summer and winter loads);
Winter loads covers December, January, February, March, April and May;
Summer loads cover June, July, August, September, October and November;
Transmission loss is included;
January to May 2016 monthly loads were from PY

23. Forecast Comparisons ILF MISO Module E
Allocates state-level retail sales forecasts to LRZ levels
Adjust zonal retail sales to metered level sales by adding zonal distribution loss
Net load: with EE/DR/DG adjustment
Gross load: without EE/DR/DG adjustment
Forecast loads by calendar year
Aggregates asset owner level loads to LRZ levels
Are on a planning year basis that does not correspond to calendar year
Are at the asset level (includes transmission losses)
Does not include DR/DG adjustment, may include EE adjustment

24. Module E Data Adjustments
Winter forecasts cover two calendar years (December-May)
We assumed 5/6 of the winter load was in the 1st calendar year and 1/6 was in the 2nd
Forecast was reduced by the transmission loss percentage to represent metered load level (as in ILF)
Loss percentage is for peak demand, not annual energy
This likely overstates losses, resulting in too large of a reduction

25. Annual Load Discrepancies Between Data 4 and Data 1
2010
2011
2012
2013
2014 1
4.88%
-1.02%
-0.90%
-0.89%
-1.01% 2
-0.40%
-0.42%
-0.91%
-0.73% 3
-0.63%
-0.81%
-0.88%
-0.65%
-0.66% 4
-0.32%
-0.18%
-0.45%
-0.43%
-0.34% 5
-0.21%
-0.11%
0.05%
0.17% 6
3.76%
-6.14%
-3.15%
-3.12%
-2.37% 7
-1.92%
-1.06%
-1.71%
-1.82%
-1.10% 8
-6.73%
-6.26%
-6.51%
-9.68%
-1.04% 9
-1.42%
-1.08%
-2.34%
-2.86%
-0.44% 10
-1.30%
-1.78%
-1.73%
-18.12%
-2.25%
MISO
0.13%
-1.93%
-1.79%
-2.62%
-1.00%
Negative values indicate that Data 4 are smaller than Data 1

26. Notes on Data Differences
We have not had an opportunity to determine source of the discrepancies
The 2014 and 2015 ILFs were calibrated to Data 1; calibration to Data 4 would lead to generally lower projections with similar trajectories

27. Comparison Charts Module E – solid black line
ILF Gross – dashed black line
ILF Gross high/low bands – hashed area
ILF Net – dashed red line
ILF Net high/low bands – solid area
2014 actual value from data 1 – blue diamond; 2014 and 2015 actual loads from data 4 – red square;
Metered loads, transmission losses excluded
Historical data based on actual weather conditions; forecasts assume normal weather

28. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 1

29. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 2

30. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 3

31. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 4

32. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 5

33. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 6

34. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 7

35. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 8

36. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 9

37. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—LRZ 10

38. SUFG Annual Gross & Net Forecast vs. Module E vs. 2015 Actual—MISO

39. Draft State-level Forecasts

40. State Forecasts
The state forecasts shown here are gross forecasts in that they do not include an adjustment for EE/DR/DG
2015 and 2016 forecasts are provided to show how different/similar they are

41. Arkansas Retail Sales (GWh)
2015 Forecast projected CAGR %
2016 Forecast projected CAGR %
actual CAGR -2.29%
CAGR – Compound Annual Growth Rate

42. Illinois Retail Sales (GWh)
2015 Forecast projected CAGR %
2016 Forecast projected CAGR %
actual CAGR %
CAGR – Compound Annual Growth Rate

43. Indiana Retail Sales (GWh)
2015 Forecast projected CAGR – 1.29%
2016 Forecast projected CAGR – 1.41%
actual CAGR %
CAGR – Compound Annual Growth Rate

44. Iowa Retail Sales (GWh)
2015 Forecast projected CAGR – 1.60%
2016 Forecast projected CAGR – 1.70%
actual CAGR %
CAGR – Compound Annual Growth Rate

45. Kentucky Retail Sales (GWh)
2015 Forecast projected CAGR – 1.09%
2016 Forecast projected CAGR – 1.20%
actual CAGR – 1.07%
CAGR – Compound Annual Growth Rate

46. Louisiana Retail Sales (GWh)
LA will be discussed later

47. Michigan Retail Sales (GWh)
2015 Forecast projected CAGR – 0.88%
2016 Forecast projected CAGR – 0.98%
actual CAGR – 0.95%
CAGR – Compound Annual Growth Rate

48. Minnesota Retail Sales (GWh)
2015 Forecast projected CAGR – 1.67%
2016 Forecast projected CAGR – 1.68%
actual CAGR – 1.58%
CAGR – Compound Annual Growth Rate

49. Mississippi Retail Sales (GWh)
2015 Forecast projected CAGR – 1.76%
2016 Forecast projected CAGR – 1.63%
actual CAGR – 1.81%
CAGR – Compound Annual Growth Rate

50. Missouri Retail Sales (GWh)
2015 Forecast projected CAGR – 1.18%
2016 Forecast projected CAGR – 1.25%
actual CAGR – 1.86%
CAGR – Compound Annual Growth Rate

51. Montana Retail Sales (GWh)
2015 Forecast projected CAGR – 1.82%
2016 Forecast projected CAGR – 1.90%
actual CAGR – 0.30%
CAGR – Compound Annual Growth Rate

52. North Dakota Retail Sales (GWh)
2015 Forecast projected CAGR – 1.08%
2016 Forecast projected CAGR – 1.55%
actual CAGR – 4.06%
CAGR – Compound Annual Growth Rate

53. South Dakota Retail Sales (GWh)
2015 Forecast projected CAGR – 2.02%
2016 Forecast projected CAGR – 2.17%
actual CAGR – 2.82%
CAGR – Compound Annual Growth Rate

54. Texas Retail Sales (GWh)
2015 Forecast projected CAGR – 1.91%
2016 Forecast projected CAGR – 2.00%
actual CAGR – 2.09%
CAGR – Compound Annual Growth Rate

55. Wisconsin Retail Sales (GWh)
2015 Forecast projected CAGR – 1.49%
2016 Forecast projected CAGR – 1.53%
actual CAGR – 1.45%
CAGR – Compound Annual Growth Rate

56. Elasticity at 2014 (weather at means)
Louisiana Model
Model presented in the last workshop
- Personal Income used as macro economic variable
Dependent Variable: ELECTRICITY_SALES
Method: Least Squares
Sample:
Included observations: 25
Variable
Coefficient
Std. Error
t-Statistic
Prob.
Elasticity at 2014 (weather at means) C
0.0000
@MOVAV(REAL_ELECTRICITY_PRICE,3)
REAL_INCOME
0.2503
CDD
0.0050
0.1747
HDD
0.0044
0.0839
R-squared
Mean dependent var
Adjusted R-squared
S.D. dependent var
S.E. of regression
Durbin-Watson stat
F-statistic
Prob (F-statistic)

57. Louisiana Model Problems with Previous Model
Personal Income has a relatively low elasticity (0.25) and a low growth rate (2.05% for the period of ).
The model produces an unreasonably low electricity sales forecast of -0.11% over the period of
Alternative approaches were tested and will be presented.

58. Elasticity at 2014 (weather at means)
Louisiana Model
Option 1
Electricity sales as dependent variable
GSP as macro-economic variable
2005 data dropped as an outliner (Hurricanes Katrina and Rita)
Dependent Variable: ELECTRICITY_SALES
Method: Least Squares
Sample:
Included observations: 24
Variable
Coefficient
Std. Error
t-Statistic
Prob.
Elasticity at 2014 (weather at means) C
0.0001
REAL_ELECTRICITY_PRICE(-2)
0.0000
REAL_GSP
0.4319
CDD
1.9495
0.0662
0.1765
HDD
0.0036
0.1269
R-squared
Mean dependent var
Adjusted R-squared
S.D. dependent var
S.E. of regression
Durbin-Watson stat
F-statistic
Prob(F-statistic)

59. Louisiana Retail Sales (GWh)
2015 Forecast projected CAGR – 1.87%
2016 Forecast projected CAGR – 0.62%
actual CAGR – 1.47%

60. Elasticity at 2014 (weather at means)
Louisiana Model
Option 2
Electricity sales + CHP as dependent variable
GSP excluding Mining GSP (oil and gas drilling)as macro-economic variable
Dependent Variable: ELECTRICITY_SALES+CHP
Method: Least Squares
Sample:
Included observations: 25
Variable
Coefficient
Std. Error
t-Statistic
Prob.
Elasticity at 2014 (weather at means) C
0.0017
REAL_ELECTRICITY_PRICE(-2)
0.0000
REAL_GSP_EXCLUDE_MINING
0.4617
CDD
0.1087
0.2010
HDD
0.0366
0.1212
R-squared
Mean dependent var
Adjusted R-squared
S.D. dependent var
S.E. of regression
Durbin-Watson stat
F-statistic
Prob(F-statistic)

61. Louisiana Retail Sales (GWh)
2015 Retail Sales Forecast projected CAGR – 1.87%
2016 Retail Sales Forecast projected CAGR – 0.95%
actual CAGR – 1.47%

62. Pros & Cons Option 1 Option 2
Consistent with approach in other states and preference for public data sources
Growth may be low considering expected new loads; some indication that all may not show up or be delayed
Option 2
Relies on arbitrary assumption that CHP will remain constant across forecast horizon
Relies on non-standard economic driver
Results in slightly higher forecast

63. Other Option
We could incorporate a load adjustment for new large loads that have a high degree of certainty
We would need information on the size and timing of those loads

64. Allocation Factors

65. Reminder
Allocation factors are used to assign state forecasts to particular LRZs
Small loads in Oklahoma & Tennessee that are in LRZ 8 are included with Arkansas
Kentucky is combined with Indiana and Montana is combined with North Dakota per MISO request
Illinois and Missouri were also analyzed at the Chicago and St. Louis Metropolitan Statistical Area

66. Historical Load Shares (MISO vs. non-MISO)
State
MISO Sales (Megawatthours)
Non-MISO Sales (Megawatthours)
2010
2011
2012
2013
2014
Average AR
34,008,348
13,071,953
70.57%
70.39%
70.52%
70.45%
72.23%
70.83% IA
43,920,136
3,281,717
92.92%
93.04%
93.22%
93.05%
93.03% IL
49,309,489
92,230,798
34.55%
34.80%
33.91%
34.59%
34.84%
34.54% IN
85,104,043
21,838,461
78.92%
79.00%
79.22%
79.58%
79.19% KY
11,400,836
67,438,088
11.87%
12.43%
12.82%
13.51%
14.46%
13.02%
IN+KY
96,504,879
89,276,549
47.49%
48.49%
48.78%
49.94%
51.95%
49.33% LA
83,987,702
6,640,614
91.77%
91.74%
92.06%
92.20%
92.67%
92.09% MI
99,263,860
4,050,238
96.01%
96.16%
96.21%
96.10%
96.08%
96.11% MN
67,870,999
848,368
98.73%
98.84%
98.75%
98.77%
98.76% MO
41,332,442
42,545,955
49.67%
49.47%
50.33%
49.52%
49.28%
49.65% MS
22,014,990
27,393,641
45.89%
45.24%
44.78%
44.73%
44.56%
45.04% MT
803,729
13,298,663
5.36%
5.40%
5.55%
5.57%
5.70%
5.51% ND
10,935,451
7,304,281
71.49%
66.16%
65.40%
59.95%
66.71%
ND+MT
11,739,180
20,602,944
37.35%
37.90%
36.76%
37.46%
36.30%
37.15% SD
3,120,871
9,233,855
26.87%
26.07%
26.02%
25.32%
25.26%
25.91% TX
21,815,593
367,854,227
5.66%
5.46%
5.99%
5.74%
5.60%
5.69% WI
69,494,755
100.0%

67. Historical Load Shares (LRZ)
MISO LRZ
State
State Level MISO Load Fraction
Average
2010
2011
2012
2013
2014 1 IA
1.77%
1.76%
1.73%
1.78%
1.83% IL
0.0002% MI
0.14%
0.13% MN
96.81%
96.73%
96.76%
96.93%
96.89%
ND+MT
37.15%
37.35%
37.90%
36.76%
37.46%
36.30% SD
24.10%
24.97%
24.28%
24.24%
23.51% WI
16.77%
16.59%
16.94%
16.23%
17.02%
17.05% 2
5.09%
5.22%
5.28%
4.89%
4.94%
5.14%
83.23%
83.41%
83.06%
83.77%
82.98%
82.95% 3
91.25%
91.14%
91.28%
91.48%
91.15%
91.22%
1.42%
1.45%
1.40%
1.95%
2.00%
1.97%
1.91%
1.86%
2.01%
1.81%
1.90%
1.79%
1.80%
1.75% 4
33.11%
33.12%
33.35%
32.49%
33.17%
33.44% 5 MO
49.41%
49.22%
50.09%
49.28%
49.06% 6
IN+KY
49.29%
47.49%
48.49%
48.60%
49.94%
51.95% 7
90.88%
90.65%
90.75%
91.19%
91.02%
90.80% 8 AR
70.83%
70.57%
70.39%
70.52%
70.45%
72.23%
0.24%
0.26%
0.25%
0.23%
0.22% TX
0.01% 9 LA
92.09%
91.77%
91.74%
92.06%
92.20%
92.67%
5.68%
5.65%
5.46%
5.98%
5.73%
5.59% 10 MS
45.04%
45.89%
45.24%
44.78%
44.73%
44.56%

68. 2016 Allocation Factors
With the exception of IN+KY and MT+ND, all allocation factors were determined from the average of the 2010 – 2014 load shares
Chicago MSA has a similar growth as Illinois for GSP, so no adjustment
St. Louis MSA has lower population but higher non-manufacturing employment, which tend to cancel each other out, so no adjustment

69. IN+KY
The closure of the Paducah Gaseous Diffusion Plant in non-MISO Kentucky (mid-year 2013) caused a shift in the LRZ 6 load share for IN+KY
We use the 2014 load share, since it is the first year where the plant is shut down for the entire year

70. MT+ND
Growth and contraction in the Bakken region, which results in a spike in statewide consumption in 2014, occurs primarily outside the MISO region
This volatility should not be reflected in the LRZ 1 forecast
We use the average of 2010 – 2013 load factors to determine the allocation factor

71. MISO Allocation Factors - AR

72. MISO Allocation Factors - IA

73. MISO Allocation Factors - IL

74. MISO Allocation Factors – IN+KY

75. MISO Allocation Factors – LA

76. MISO Allocation Factors – MI

77. MISO Allocation Factors – MN

78. MISO Allocation Factors – MO

79. MISO Allocation Factors – MS

80. MISO Allocation Factors – ND+MT

81. MISO Allocation Factors – SD

82. MISO Allocation Factors – TX

83. MISO Allocation Factors – WI

84. Next Steps

85. September Workshop We will present updated energy to peak conversion
We will cover the draft forecast results at LRZ and MISO level
Please send your comments and suggestions for improvements