Draft EPI Results - The Portland Cement Association

Methodology and Results of the
Energy Performance Indicator for
Cement Manufacturing
Presented to the
Portland Cement Association
Manufacturing Technical Committee
September 27th, 2004 - Chicago Illinois
By Gale A. Boyd, PhD
Argonne National Laboratory
Office of Science
U.S. Department of Energy
A U.S. Department of Energy
Office of Science Laboratory
Operated by The University of Chicago
Estimating the “Energy Efficiency Gap”
• Engineering models may represent best practice, while
statistical models are typically based on average practice.
- Measures of energy intensity based on average practice are of
limited use in managing energy use or for corporate goal setting.
- A more useful measure represents where a company or plant
lies within a distribution of performance.
“Is performance close (or far) from the industry best practice?”
• We modify the typical statistical approach to develop Industrial
Energy Performance Indicators (EPI) to measure “best
practice” and the “efficiency gap” for ENERGY STAR
- Method: Stochastic frontier regression analysis
- Data:
Plant level data on energy and production
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
The EPI Statistically Identifies “Best Practice”
•
“Best” practice is defined for a specific application, with observable
economic and structural differences accounted for
- Variation in observed practice can exist for a number of reasons;
- Economic decisions
- Energy prices
- Utilization rates
- Structural differences
- Production processes
- Materials choice
- Those differences are not part of the “efficiency gap”
•
Statistical models are well suited to account for these differences
- Statistical models are commonly based on aggregate data
- No explicit treatment of “best” and “average” practice.
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Stochastic Frontier is a Modified Regression
• Linear regression computes the “typical” performance given
•
•
exogenous effects
- Explains the data by finding the best fit line which “goes through
the middle” of the data
- Any deviation is “statistical noise” which is assumed to be
normally distributed, i.e. can be positive or negative.
The frontier computes the best-performing given those same
effects
- Explains the data by finding the best fit line which “envelopes
the frontier” of the data
- Some deviation is “noise”, but deviations may also be
inefficiency which is assumed to follow a one-sided distribution.
Estimating the statistical distribution allows us to compute a
normalized percentile score
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Plant Level Data as a Key Component
• Analysis uses confidential plant level data from two sources
- Center for Economic Studies (CES), U.S. Bureau of the Census
- Data provided to ANL by PCA
• Data from CES includes the non-public, plant-level data which is
the basis of the government statistics on manufacturing.
• Title 13 of the U.S. Code protects this data,
- CES allows researchers with Special Sworn Status to access
these confidential micro-data at a Research Data Center
(RDC).
- Confidentiality prevents the disclosure of any information that
would allow for the identification of a specific plant or firm’s
activities.
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Scope of the Current EPI Analysis
• Uses 1992 and 1997 Census of Manufacturing (CM) Data
•
•
•
- Those years have the most detail on production.
- Some plants were dropped for missing data
Manufacturing Energy Consumption Survey (MECS) provides
plant energy mix.
- All other plants were excluded from the analysis.
- Since MECS is a statistical sample, this is a “good” representation
of the industry.
PCA Plant Information Summary Report provides
- Kiln capacity and utilization.
- Cross check the MECS fuel and CM production
PCA Labor and Energy Survey data was provided to ANL under a
non-disclosure agreement
- This data is being used to check data used in the current EPI
- Additional analysis will be done as warranted
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Basic Inputs to the Cement EPI
• Uses Plant Total Primary Energy (TPE)
•
- Total BTU’s of coal, waste derived fuel, etc
- kWh of electricity purchased from the grid is converted to
BTUs at power plant thermal efficiency (average 10,236
BTU/kWh)
Includes these Factors
- Total Capacity of all Kilns in the plant
- Number of Kilns (average capacity)
- Capacity utilization – percentage
- Production Mix
- ASTM 1, 2 or 5
- ASTM 3
- ASTM 4
- Masonry
- Other Cement
- Clinker Shipped as a Separate Product
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Basic Steps in Frontier Statistical Analysis
• Specify the systematic relationship (linear) between variables
•
•
like production, plant size, output mix, etc.
Run linear regression on plant level dataset
- Data includes 137 observations
- About 69 plants for two years
Adjust the parameters using Maximum Likelihood methods to
improve the fit based on frontier assumption that inefficiency
has an exponential distribution
- Shift the linear regression intercept so that all data are on one
side of the line
- Adjust linear regression to represent simultaneous best fit of
slopes, error (noise) variance σ2, and Exponential distribution
parameter θ
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
The Statistical Frontier Model for Cement
• Stochastic frontier regression separates energy intensity into
- Systematic effects,
- Statistical (random) error
- Inefficiency
6
3
3
ln( Et ,i )    i ln(Yi )    i ln( X i )    i Z i  ui  vi
i 1
y
x
i 1
Z
i 1
- E is energy use, Y includes outputs, X is the vector of systematic
economic variables, Z is the vector of systematic external factors, β is
the vector of parameters to be estimated,
- v is the typical random error term
- u is distributed according to some one-sided error distribution, for
example exponential
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Parameter Estimates for the Current EPI
Parameter Estimates for the Total Primary Energy Frontier in the Cement Industry
Coefficient
Standard Error
z
Constant
1.928
0.97
1.97
ASTM125
-0.08295
0.027
-3.08
ASTM3
-0.1184
0.0058
-2.05
ASTM4
0.01190
0.0059
2.03
Masonry
0.00654
0.0054
1.21
Other Cement
0.01033
0.0085
1.21
Clinker
0.00142
0.0086
0.17
Kiln Capacity
0.9358
0.086
10.85
Kiln number
0.1391
0.054
2.54
labor
0.00815
0.096
0.08
Capacity Utilization (CU)
1.584
0.54
2.92
CU2
-0.4051
0.26
-1.57
Wet=1, Dry=0
0.1066
0.067
1.58
v
0.19389
0.023
.
u
0.33117
0.041
.
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Input Section of Cement EPI Spreadsheet
Energy Performance Indicator Tool
for Cement Plants
04/04/04
Back
Plant Characteristics
Baseline Year
SIC Code: 3241 (Cement)
Zip Code: 60601
Select Year:
Location: Chicago, IL
Current Year
Units
2002
Kiln Capacity
400,000
Total Clinker Produced
360,000
Capacity Utilization:
90%
ASTM 1, 2 and 5
400,000
400,000
360,000
90%
400,000
short tons
ASTM 3
short tons
ASTM 4
short tons
Masonry
short tons
Other Cement
short tons
Clinker Shipped as a Separate Product
short tons
Number of Kilns
2003
short tons
1
1
Energy Consumption
Electricity
Select Units
Baseline Year
(2002)
Current Year
(2003)
Data Entry Check
Pioneering
Science and
Technology
Annual Consumption
Annual Cost ($)*
Annual Consumption
Annual Cost ($)*
MWH
Gas
mMBtu
Distillate Oil
Gallons
Residual Oil
Gallons
Coal
mMBtu
78,720
30,882
1,400,000
3,699,840
124,490
1,692,320
78,720
30,882
1,400,000
3,699,840
124,490
1,692,320
Waste derived
mMBtu
* - user can overwrite with actual costs
Office of Science
U.S. Department
of Energy
Output Section of Cement EPI Spreadsheet
Results
Your Plant
Baseline (2002)
Your Plant
Current (2003)
Average Plant
(2003)
Efficient Plant
(2003)
36
36
50
75
$5,516,650
$5,516,650
$4,545,137
$3,611,264
Total Cement Shipped (Short Tons)
400,000
400,000
400,000
400,000
$ Energy/Total Clinker ($/ short Ton)
$15.32
$15.32
$12.63
$10.03
Total Primary Energy (mMBtu)
2,237,401
2,237,401
1,843,382
1,464,629
Total Site Energy (mMBtu)
1,699,475
1,699,475
1,400,188
1,112,496
6.22
6.22
5.12
4.07
3.43
201,000
3.43
201,000
3.43
165,603
3.43
131,577
EPI
Annual Energy Cost ($/year)
Energy / Output
(mMBtu/short Ton
of clinker)
Best Practice (mMBtu/short Ton)
CO2 (tons/year)
Baseline Year (2002)
Current Year (2003)
100%
90%
90%
EPI = 36
80%
60%
50%
40%
30%
EPI Percentile Score
70%
80%
70%
60%
50%
40%
30%
20%
20%
10%
10%
0%
0
2000
4000
6000
8000
10000
Total Primary Energy (Billion Btu)
Pioneering
Science and
Technology
12000
0%
0
2000
4000
6000
8000
10000
12000
Total Primary Energy (Billion Btu)
Office of Science
U.S. Department
of Energy
EPI Percentile Score
EPI = 36
100%
The “Average” Cement Plant
•
•
•
Has a kiln capacity of around 800,000 TPY
Consumed 3.5 trillion BTUs of Primary
Energy
- 2.7 trillion BTUs of Fuel
- 0.8 trillion BTUs of Electricity
- 5.1 million BTU TPE per ton of clinker
Shipped $50 million in products
…but the average doesn’t tell the whole
story…
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Some Stylized Results Include
• Lower plant utilization incurs an
•
•
implicit “penalty” in energy use
“Size” matters
- Larger plants are more efficient
- Larger kilns are more efficient
Cement types other than ASTM I, II, or
V involve amounts of additional energy
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
A Large Kiln is “Better” than Several Small Ones
100%
90%
EPI = 57
80%
70%
60%
50%
40%
Single 400,000 TPY Klin:
EPI = 41
Best Practice = 3.34
30%
EPI Percentile Score
Four 100,000 TPY Klins:
Best Practice = 4.05
20%
10%
0%
0
2000
4000
6000
8000
10000
12000
T otal Primary Energy (Billion Btu)
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Some Cement Types have Higher Energy Requirements
100%
90%
EPI = 59
80%
70%
60%
50%
40%
ASTM Standard: Best
Practice = 3.34
EPI = 41
30%
EPI Percentile Score
80% Standard, 10%
ASTM IV, 10% Masonry:
Best Practice = 4.14
20%
10%
0%
0
2000
4000
6000
8000
10000
12000
T otal Primary Energy (Billion Btu)
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Larger Kilns - Lower Observed Best Practice
100%
90%
EPI = 34
80%
70%
60%
50%
40%
400,000 TPY: Best
Practice = 3.34
30%
EPI = 41
EPI Percentile Score
800,000 TPY: Best
Practice = 3.01
20%
10%
0%
0
2000
4000
6000
8000
10000
12000
T otal Primary Energy (Billion Btu)
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Lower Utilization - Higher Observed Best Practice
100%
90%
EPI = 48
80%
70%
60%
50%
40%
100% Utilization: Best EPI = 41
Practice = 3.34
30%
EPI Percentile Score
75% Utilization: Best
Practice = 3.66
20%
10%
0%
0
2000
4000
6000
8000
10000
12000
T otal Primary Energy (Billion Btu)
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy
Analysis for the Cement EPI is Ongoing
•
Some additional adjustments using engineering
principles are under development
- Low Alkali cement may require kiln by-pass
- Other emissions control also require by-pass
•
Use of PCA data will enhance the analysis
because of (hopefully) better accounting of
- By-product based fuels
- Clinker production
•
Final version will be publicly available from EPA
and PCA
Pioneering
Science and
Technology
Office of Science
U.S. Department
of Energy