This document discusses using statistical modeling to benchmark energy performance across textile plants. It presents a case study comparing the energy use of 3 plants based on production levels and counts. Key points:
- Production data and electricity use from the plants over 3 months is analyzed using statistical tools to model the relationship between energy use and factors like product type and count.
- The analysis identifies the largest impacts different factors have on each plant's energy use. It also calculates expected energy use and energy performance indexes based on the statistical models.
- The energy performance indexes change when the production mix or other factors are adjusted in the models, showing how normalization is important for accurate benchmarking. Deeper and more consistent data across more plants
2. 2
Contents
Need for Sectoral Energy Performance Benchmarking......................................... 3
Benchmarking types and factors:.........................................................................................................4
Model - based Benchmarking ..............................................................................................................4
Methodology:......................................................................................................................................5
Energy efficiency Benchmarking in Textile Units- A case Study............................ 6
Energy performance based on Statistical Energy Simulation ................................................................6
Data Preparation .............................................................................................................................6
Data Mining and exploring the relationship among features: ...........................................................7
Impact Assessment after statistical simulation:................................................................................7
Energy Performance Index Table: ........................................................................................................8
The traditional Energy Performance Index: ......................................................................................8
Energy Performance Index – Using Statistical Tool ...........................................................................8
Critical Analysis for Using Model Based Benchmarking:......................................................................10
Conclusion ........................................................................................................... 11
3. 3
Need for Sectoral Energy Performance
Benchmarking
Energy Efficiency market place is evolving and is moving more towards
market based mechanism, Cap & Trade markets and long term reduction
targets both at Country/Sector levels.
Evolution of the market place puts a certain risk on all stakeholders, as we
will see in the info graph below.
Market Based
Mechanisms
Cap & Trade
Long term
reduction
targets at
Country/Sector
level.
Stringent MRV
requirements.
Industries
•Exposure to
signifgicantly high
targets with respect to
improvement potential
/ sectoral
benchmarking
Goverments
•Market stability on
account of "partial"
benchmarking.
Banks/Financial
Institutions
•Potential and EE Project
effectiveness
dependent on EnPI
driven performance
evaluation.
ESCOs
•Perofrmance
contracts/investment
subject to
normalisation
factors/forecasting
models etc.
Benchmarking is a
positive, proactive
process to change
operations in a
structured fashion
to achieve
superior
performance. The
benefits of using
benchmarking are
that functions are
forced to
investigate
external industry
best practices and
incorporate those
practices into
their operations.
This leads to
profitable, high-
asset utilization
businesses that
meet customer
needs and have a
competitive
advantage
Comprehensive use of Data Analytics at different stages
i.e. Policy Design, Implementation and evaluation holds
the key to mitigating the risks for all stakeholders!
4. 4
Benchmarking types and factors:
Internal benchmarking is done to check and follow-up on the EnPIs1
of a plant or group of plants within
the same company. It mainly consists of quantitative data which can be both dynamic and static data.
Typical examples of factors considered for evaluation of EnPIs are Production capacity, count, TPI,
Spindle speed, Capacity Utilization, Product Mix etc.
External benchmarking is done to check and follow-up on the EnPIs of the plants at the cluster or the
country level. It mainly consists of Qualitative data with factor such as Technology, Labor Skills,
Weather, Age, Automation Level etc.
Model - based Benchmarking
Model-based benchmarking calculates benchmarks based on a simulation model of a facility’s
performance. The simulation models have the advantage that they can estimate the contribution of a
wide range of factors and parameters to facility energy performance in order to determine targets and
compare retrofit scenarios. The detailed simulation models require detailed facility information
(constructional, production and operational data as inputs).
Statistical analysis benchmarking (also known as regression model-based) is being deployed and used in
this case study. Here we try to make benchmarking process for energy performance indicators by means
of multiple regression analysis, where the relationship between electricity consumption and the
explanatory factors e.g. different product types and Count is developed.
1
Energy Performance Indicators
Benchmarking
Internal
Quantitative Data
Combination of Dynamic
and Static Data
External (Sectoral)
Combination of Qualitative
and Quantitative Data
Mostly static data.
Depth of Data (Type and Volume) Quality and consistencyof Data
5. 5
Methodology:
• Energy Intensive Industries can track their performance
of EnPIs on a cluster as well as global level. To improve
productivity,reliablity, and quality of the products.
Establish the Goal for
Benchmarking
•Need to identify the metrics requirement such as UKG,
specific energy consumption, utility efficiency, equipment
efficiency, capacity utilization, labour productivity. Also
the frequncy of the output metrics required i.e. daily ,
monthly, quarterly or yearly.
Identify Output Metrics Needed to
Supportthe Benchmarking Goal
•Data inputs required to generate the desired output
metrics such as production quantity, raw material
consumption, Electrcity consumption, gas, liuid fuel,
steam consumption, motor parameters. Qualitative data
such as Technology, Productivuty, weather, labor skills,
Automation.
Identify the Data Input
Requirement
• Select a tool to display the output metrics whether
spread sheets or web based platforms. The tools should
not be too difficult to use to the users. it should take care
the usablity and comapatiblity of the users.
Select a Benchmarking Tool
•Need to specify the collection techniques such as data
entry direct into the platform or in the spreadsheet. Also
the frequency of the data entry. Also the ways of
collection and aggregation of data should be the
specified.
Determine a Collection Method
•Performing onsite or offsite verification to check the
metrics and magnitude of the variables by comparaing
with the cluster or national level trend.
Consider the Data Verification
Process
•Compare with energy simulation /energy modelling to
compare the perormance of the plant with potential
performance. To go into detailed anlysis identifying low
performing plants within a group or cluster.
Evaluate Analysis Techniques
•Communicate the idea and the value to ensure
inolvement of all stakeholders.
Communicate the Plan and
Formalize the Process
6. 6
Energy efficiency Benchmarking in Textile Units- A case Study
Energy performance based on Statistical Energy Simulation
Here we are considering datasets for 3 months of three textile plants where we have the independent
variables: Product types i.e. P1, P2 and P3 in (Tons) and counts of production such as C1, C2 and C3 and
the dependent variable is Electricity consumption (kWh).
Data Preparation
a) Missing Value Treatment:
We can detect all the missing values and remove them to make the data more compact so that we can
build a standard model with more robust features. After detecting the missing values we will remove all
the missing values to make the data compact for further processing.
b) Outlier Detection:
We can use different graphical techniques for the detection of outliers. Graphical techniques such as
Histogram, density plot and box plot can be used for the detection of outliers. With use of box plot it is
clearly visible that the data does not contain any outliers.
After plotting Box plots of variable for Plant 1 above we can statistically detect that factors P2, P3 and
C3 have outliers. Similarly we can find outliers in datasets of Plant 2 and Plant 3.
7. 7
Data Mining and exploring the relationship among features:
We can apply data mining techniques to get measures of central tendency such as mean, median,
quartiles, min and max values for the numerical variables. Also we can use scatter plot and box plot to
see the relationship between the features.
In the figures the three types of products P1, P2 and P3 are represented by labels 1, 2 and 3 for the
three Plant 1, Plant 2 and Plant 3 respectively. We can observe that Plant 2 having the widest range in
the individual quantity for types of production. The production level for plant 3 is showing more
scattered in the production of three types.
Plant 1 Plant 2 Plant 3
Impact Assessment after statistical simulation:
We can visualize the individual impact of coefficient of independent variables (Production and Count of
different types) for all the three plants on electricity consumption. The impact of factor P1 is highest
among all the KPIs in Plant 1 and Plant 3 i.e. 40% for Plant 1 and 42% for Plant 3 . Considering C1 factor
we can interpret that Plant 1 has least impact of 4% but Plant 2 getting highest impact of 34% for
variable C1.
40% 22%
31%
4%
3% 20%
22%
25% 15%
34%
3%
19%
42%
17%
29%
11%
1%
8%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
P1 P2 P3 C1 C2 C3
Plant 1 Plant 2 Plant 3
8. 8
Energy Performance Index Table:
The traditional Energy Performance Index:
Plant Average Production
(Tons)
Average Electricity
Consumption
( kWh)
UKG Energy Performance
Index
Plant 1 45.00 101537 2.26 1.01
Plant 2 131.00 355777 2.72 1.21
Plant 3 25 56094 2.24 1.00
Energy Performance Index – Using Statistical Tool
Case 1
Plant Production % Product Type Average Production(Tons): 45 Counts Calculated
Plant 1 60% Pro1 27.00 26 KWh 116476
6% Pro2 2.70 27 UKG 2.59
34% Pro3 15.30 23
Production % Product Type Average Production(Tons):131 Counts Calculated
Plant 2 60% Pro1 78.60 26 KWh 354638
6% Pro2 7.86 27 UKG 2.71
34% Pro3 44.54 23
Production % Product Type Average Production(Tons): 25 Counts Calculated
Plant 3 60% Pro1 15.00 26 KWh 58959
6% Pro2 1.50 27 UKG 2.36
34% Pro3 8.50 23
Energy performance Index Table
Plant Average Production
(Tons)
Expected Average Electricity
Consumption
( KWh)
UKG Energy
Performance Index
Plant 1 45 116476 2.59 1.10
Plant 2 131 354638 2.71 1.15
Plant 3 25 58959 2.36 1.00
In case 1 we have normalized the production quantity product mix and count of the three types of
production. We can observe the change in the magnitude of the UKG and Energy performance index of
the three plants but there is no change in ranking on the basis of Index.
9. 9
Case 2
Plant Production % Product Type Average Production(Tons): 45 Counts Calculated
Plant 1 20% Pro1 9.00 26 KWh 100724
42% Pro2 18.90 27 UKG 2.24
38% Pro3 17.10 23
Production % Product Type Average Production(Tons):131 Counts Calculated
Plant 2 35% Pro1 45.85 15 KWh 336247
30% Pro2 39.30 38 UKG 2.57
35% Pro3 45.85 25
Production % Product Type Average Production(Tons): 25 Counts Calculated
Plant3 60% Pro1 15.00 26 KWh 59237
6% Pro2 1.50 25 UKG 2.37
34% Pro3 8.50 24
Energy performance Index Table
Plant Average Production
(Tons)
Expected Average Electricity
Consumption
( KWh)
UKG Energy Performance
Index
Plant 1 45 100724 2.24 1.00
Plant 2 131 336247 2.57 1.15
Plant 3 25 59237 2.37 1.06
In Case 2 after change in production mix quantity, the change in the magnitude and the ranking of the
Plants on the basis of Energy Performance Index can be observed.
Energy Performance Index Table
Plant Traditional Way Case 1 Case 2
Plant 1 1.01 1.10 1.00
Plant 2 1.21 1.15 1.15
Plant 3 1.00 1.00 1.06
10. 10
So we can infer that energy market based mechanisms should take into account more qualitative and
quantitative features with more depth and consistency before providing any energy or emission
reduction targets which might be very lower or very higher .
Critical Analysis for Using Model Based Benchmarking:
The importance of Qualitative data in Benchmarking cannot be ignored. The qualitative factors can be
sub-categorized and provide a label to have better insight into plant performance. Examples of a few
factors have been listed below:-
I) Technology/Modernity: a) New, b) Medium and c) Old : The plant with new technology will consume
less energy and would have higher productivity.
II) Plant Automation: a) High, b) Medium and c) Low : Plant automation does not help in the lesser
energy consumption (in most cases) it only removes human interference in the process.
III) Weather: a) Dry, b)Moderately Humid and c) Humid : Plant in the dry area will be consuming more
energy as it will require higher humidification to maintain the humidity level for the products to be
manufactured and vice versa.
We see that post normalization the indices have changed; in this case we have
only considered 2 factors and 3 units, when we consider a cluster with almost
200 units and with over a dozen factors both the magnitude of the index and the
overall rankings can change impacting all stakeholders!
The magnitude and Energy performance index will become completely opposite of the general
trend or can remain the same which can only be told with higher volume and varied datasets!
11. 11
Conclusion
The objective is to make best benchmarking tool that caters all the factors significant in the driving the
energy consumption of the plant. With larger volume, greater depth and consistent data more insight
can be brought into the performance level, and also the hidden factors causing the improvement and
deterioration in energy performance can be captured and tracked.
Analysis and indexing to be done by
Algorithms, making the process system
driven (Faceless)
Integrationof
unit wise data
through EIMAS
or through data
loggingfacility.
Establish Central
EnergyEfficiency
Data Repository
Identify the
depth (Type and
Volume) of
Quantitative and
Qualitativedata
to be collected
12. 12
About E-Cube Energy
E-Cube Energy has emerged as “Thought Leader” around the subject of using Energy Analytics in
Industries/SMEs to accelerate energy efficiency.
Energy Analytics represents systematic, evidence and data led process in identifying and assessing
opportunities to advance energy savings in unit operations in industry. It's beginning to make inroads in
Indian industry as a management tool to establish realistic energy savings targets and as a means to
measure and verify progress.
Our company through its products and solutions helps industries/utilities gets started with their Energy
Data Strategy and transition to Faster, Better and Cheaper way of assessing, reporting and managing
Energy Performance.
To know more about us and how we can help you get started with your EDA strategy visit us at
www.eetpl.in or call us at +91 033 40052780