Energy Conversion and Management 117 (2016) 454–465

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Energy, economy, and environment analysis and optimization

on manufacturing plant energy supply system
Lujia Feng a,⇑, Laine Mears a, Cleveland Beaufort b, Joerg Schulte b
a
Department of Automotive Engineering, Clemson University, 4 Research Drive, Greenville, SC 29615, USA
b
BMW Manufacturing Co., Greer, SC, USA

a r t i c l e i n f o a b s t r a c t

Article history: Increasing attention has recently been drawn to energy consumption in manufacturing plants. Facing the
Received 26 June 2015 challenges from reducing emissions coupled with rising raw material prices and energy costs, manufac-
Received in revised form 7 March 2016 turers are trying to balance the energy usage strategy among the total energy consumption, economy,
Accepted 11 March 2016
and environment, which can be self-conflicting at times. In this paper, energy systems in manufacturing
Available online 23 March 2016
environments are reviewed, and the current status of onsite energy system and renewable energy usage
are discussed. Single objective and multicriteria optimization approaches are effectively formulated for
Keywords:
making the best use of energy delivered to the production processes. Energy supply operation sugges-
Manufacturing plant
Energy supply
tions based on the optimization results are obtained. Finally, an example from an automotive assembly
Optimization manufacturer is described to demonstrate the energy usage in the current manufacturing plants and how
Energy the optimization approaches can be applied to satisfy the energy management objectives. According to
Economy the optimization results, in an energy oriented operation, it takes 35% more in monetary cost; while in
Environment an economy oriented operation, it takes 17% more in megawatt hour energy supply and tends to rely
more on the inexpensive renewable energy.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Worldwide energy demand has continued increasing in the past
decades. Every aspect of human activity – transportation, indus-
trial, residential and commercial activities – requires rising support
from energy. Fundamental to human development, energy could
now also be harmful and restraining to sustainability. High
expenses, unbalanced distribution and intensive demand leave
energy-sparse countries fewer opportunities to advance technolog-
ically. Additionally, the correlation between the energy consump-
tion and environmental degradation is well known – acid rain,
deforestation, greenhouse effect, particle matter pollution, and
similar negative environmental effects are examples. ‘‘Energy” is
discussed as a problem, and a large body of research (and coupled
enthusiasm) is directing expertise to solve this problem. This paper
focuses on the energy supply to manufacturing plants.
Among the above-mentioned four end sectors, industrial activ-
ity is believed to be the most energy intensive. It is also reported
that in recent years, more than one third of the energy produced
in the US was consumed by industrial activity [1]. Manufacturing
represents a significant contributor to this consumption. Compared
⇑ Corresponding author.
E-mail address: lujiaf@g.clemson.edu (L. Feng).
with the large cost in production materials and labor, energy used
to be considered in a minor role. However, manufacturers are fac-
ing a lot of pressure. Increasing energy price requires manufactur-
ers to spend less [2]. More restrictive policies require them to use
less energy per product [3]. Ascendant public environmental pro-
tection awareness demands plants to use more clean energy [4].
Manufacturers are now paying closing attention to their energy
use.
Generally, research in the manufacturing energy system can be
classified into two directions: energy demand and energy supply. A
great number of studies have been made in reducing the energy
demand in manufacturing systems, such as the improvement of
process scheduling and machine power level adaptation. As the
manufacturing systems get more complex, researches have be
done to study the energy in the manufacturing flexibility [5]. The
scheduling problem was specifically studied [6]. Machine power
level energy efficiency was also well developed [7]. On the other
hand, aside from recent attention to systemic intelligence such as
smart grid and smart metering approaches, research into improv-
ing energy supply systems, particularly at the local industrial level,
has seen less focus.
The plant level manufacturing system [8] is chosen as the study
objective reported in this paper. It is a relatively independent sys-
tem in energy operation, both from the supply/conversion side as

http://dx.doi.org/10.1016/j.enconman.2016.03.031
0196-8904/Ó 2016 Elsevier Ltd. All rights reserved.
 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465
Nomenclature
E energy
g efficiency
CE energy cost
CM maintenance cost
EF emission factor
Ele abbreviation for electricity
NG abbreviation for natural gas
LFG abbreviation for landfill gas
J matrix of ones
ED energy demand
S supplied energy
well as demand. Considering the great amount of energy consump-
tion in manufacturing plants, the decisions regarding supply strat-
egy are of great importance. For example, the high peak electricity
consumption of the plant is an utmost test to the local grid on its
power reconfiguration capacity. On the other hand, the influences
from both higher level (e.g., strategy setting, policy in reducing the
total carbon emission, or demand variations in an agile energy
market) and lower level (e.g., production schedule and weather
effects) can also impact the decision made by the manufacturers.
Energy supply is the focus of this paper. Recently, manufactur-
ing plants tend to have their own energy conversion and transmis-
sion systems to supply the energy demand for the main production
lines. How to operate onsite energy conversion and transmission
systems, how to cooperate the onsite system with the primary
energy delivery from the utility companies, and how to achieve
the best results in terms of energy, cost and emissions – these
are the questions pursued in this paper by analyzing and optimiz-
ing the energy supply system in manufacturing plants. A case
study from an automotive assembly plant is used to demonstrate
the approaches, and show how the operation can be affected by
both higher and lower level variability and decision makers’ prior-
ities. Compared with previous works, this work exam the energy
conservation opportunities from the supply instead of consump-
tion side, and clarify the savings from the energy, monetary cost,
and environmental emission aspects.
2. Background
Pertinent background related to energy usage within the cur-
rent manufacturing plant, especially the research on supply sys-
tems, renewable energy application, metering status,
optimization approaches and U.S. energy prices are reviewed and
discussed in this section.
2.1. Energy conversion and transmission
Purchasing all demand energy directly from the utility company
requires only a small capital investment; but it is neither cost rea-
sonable, nor pragmatic in the long term. To face the variable pro-
duction conditions and changeable energy prices, plants are
typically equipped with an onsite (decentralized) energy conversion
and transmission system. While energy transmission is only to
deliver the same forms and amount to the production lines, con-
version involves changes in the energy forms and quantities. Typ-
ical energy conversion forms include combustion, electricity
generation, air compression and thermal energy exchange. Repre-
sentative examples of the energy conversion is given here. Com-
bustible energy (such as coal, oil and natural gas) are burned in
the combustion chamber to generate steam which rotates the
455
P purchased energy
Subscript
in input
out output
i indicator of energy
j indicator of equipment
k indicator of objective
h indicator of energy demand
turbine connected with an electrical generator. In this way, the
chemical energy from the primary energy is converted to electric-
ity. Traditional fired generation systems release the exhaust gas to
the atmosphere; however a co-generation or tri-generation system
will recover part of the thermal energy through heat exchange to
create hot water or steam for later use. In this case, the thermal
energy was also captured [9]. Other examples for chemical and
thermal energy conversion, and electricity and thermal energy
conversion are from the burners and chillers respectively. Usually,
a burner/boiler will be onsite to supplement hot water/steam for
production or building heating [10]. Chillers use electricity to gen-
erate chilled water, used for equipment and building cooling [11].
In the case of air compression, air compressors use electricity as
energy input to compress air to a higher pressure for carrying
energy to the shop floor [12]. Detailed energy modeling of these
traditional energy conversion and transmission systems is rela-
tively straightforward and well-studied [13].
2.2. Renewable energy
Apart from the geothermal and biomass energy, which have
high requirements on the techniques and particular to location,
solar and wind generation are two relative popular renewable
energy sources for the manufacturing plants. However, compared
with traditional energy supplies, solar and wind are relatively
unstable.
Solar energy is used to provide high temperature as a process
heat source, which has been increased use recently [14]. The elec-
trical power from the PV solar panels depends on temperature and
temperature. Researches use the MPPT (maximum power point
tracker) to calculate the maximum power they can obtain from
the sun:
PS ðG; DT Þ ¼ k1  AS  G  ð1  kT DT Þ ð1Þ
where P S is the power from sun, AS is the total area of the PV model
(m2), DT ¼ T c  T cref is the temperature difference between the cell
temperature T c and the reference cell temperature T cref (°C), kT is
the temperature coefficient, and k1 is the PV module generation effi-
ciency [15]. Solar irradiation G is often described in stochastic mod-
els to solve the problem of unstable availability of the solar input.
Wind power can be captured through coupled wind towers and
turbines. The available wind generator power Pout can be expressed
as a function of wind speed V wind :
8
> ðV kwind V kin Þ
< Prated  ðV kin V krated Þ if V in 6 V wind 6 V rated
>
Pout ðV wind Þ ¼ ð2Þ
>
> Prated if V rated 6 V Wind 6 V out
:
0 Otherwise:
456 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465
where P rated is the rated power of turbine which is design specifics
generally given by the turbine manufacturers, V in is the cut-in wind
speed, V rated is the rated wind speed, V out is the cut-out wind speed,
k is the Weibull shape parameter. Like solar irradiation, wind speed
is also commonly described by a random variable distribution func-
tion [15].
Landfill gas is another renewable energy used to replace the
consumption of natural gas. Compared with natural gas, landfill
gas has lower methane content and relatively low quality. How-
ever, compared with other renewables, landfill gas is highly reli-
able and constant. As long as the manufacturing plant can find
suppliers with landfill gas, they can have a long-term contract
and by small modifications to their current equipment (usually
burners), landfill gas can be used directly.
2.3. Plant energy supply optimization
Multiple criteria need to be taken into consideration when mak-
ing decisions about sustainability in energy management. Wang
and his colleagues reviewed the work done in energy decision
making [16]. According to their paper, the criteria can come from
techniques, economy, environment and society. And they also
pointed out that the decision of criteria selection could be difficult,
and they came up with the principles to follow and elementary
methods to apply when choosing the major criteria.
The weighting method is one of the most popular approaches
when dealing with multicriteria optimization. Generally, the deci-
sion maker will assign preferential weights to different normalized
criteria and force the multi-objective problem to be a single cost
function. Equal weights method without prudent knowledge gives
the objectives the same priority and treated equally, while rank-
order weighting drives the ranking of each objective hierarchically
to determine the priority in optimization. This method does not
necessarily encompass deeper knowledge of the problem (such
as the lexicographic optimization), but instead calls for subjective
opinions [17].
Apart from converting to single objectives, multicriteria pro-
gramming allows for solving the problem with non-dominated
points called efficient or Pareto optima [17]. The Pareto optimal
solution is a state where it is impossible to improve one objective
without sacrificing at least one of the others. In planning dis-
tributed energy resources, application of the Pareto optima
approach are seen in [18–20]. Rodriguez and his group organized
a review on the multi-objective planning of distributed energy
resources; they concluded that this area is promising and will pro-
vide guidance to the future development of distributed energy
sources [18].
Buoro and his team studied an industrial area where different
economic sectors (e.g., food, plastics, furniture manufacturers,
and so on) clustered to share an energy facility – a district heating
Fig. 1. United States industrial electricity average retail price [21].
network, small CHP systems, large centralized solar plant and a
thermal storage [19]. In their paper, mixed integer linear program-
ming was used to consider the influence from energy cost and car-
bon dioxide emission caused by the operation. The relative weights
of the energy and emission minimization objectives were varied to
identify the Pareto front solutions. This optimization was devel-
oped under the condition of steady state operation without consid-
ering the fluctuation caused by demand variation in different
scenarios.
Lazzaretto and Toffolo took an example to discuss the energy
objective in terms of exergetic efficiency in a cogeneration plant,
economy in the total cost of fuel and environment effects through
the conversion of pollution damage cost of multiple emission pol-
lutants [20]. Their research considered the primary zone combus-
tion temperature, combustor inlet pressure and pressure drop in
the combustion chamber of cogeneration system to calculate the
emission of nitrogen oxides and carbon monoxide. Single objective
on each of the cost functions and multicriteria optimization with a
Pareto surface was given in the paper to illustrate the tradeoff of
the optimal solutions. This paper concentrated on the thermal sys-
tems design, but neglected the multiple energy demands of current
manufacturing systems.
It is easy to neglect and underestimate the great potential in
saving energy and improving sustainability by optimizing the
energy supply strategies in manufacturing plant, as these are often
seen in conflict with one another. Despite the traditional straight-
forward energy supply, modern manufacturers tend to have their
own onsite distributed energy generation and conversion systems
to fulfill the variety energy carriers’ demands. Huge saving possi-
bilities are hiding among the complex system. Consideration of
the proper strategy, importance of objectives, and optimization
method can be implemented to unveil the conservation secrets
behind everyday operations.
2.4. Energy price
According to the U.S. Energy Information Administration, the
price of electricity has continued increasing over the past 15 years.
Fig. 1 is the plot of United States average electricity retail price
to industry sector. This figure shows that the price in the past
15 years is increasing, with seasonal factors – higher price in the
summer time and lower price during the winter time.
Fig. 2 is the plot of United States natural gas average retail price
to the industrial sectors; the price of natural gas is expressed as US
Dollars per thousand cubic feet. This figure shows the natural gas
price peaked in year 2006 and 2008/9, but has tended to be rela-
tively stable in recent years.
The price of landfill gas is highly dependent on the techniques
used in the suppliers, the quality of the landfill gas and also the
market price for natural gas and electricity.
 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465 457

Fig. 2. United States industrial natural gas average retail price [22].

3. Model establishment maintenance cost is mostly periodic according to the scheduling

[23].
Although the initial investment and construction is critical, this X X
m n
 
research focuses on the operational stage of the energy usage and z2 ¼ CEi þ kj  CM j ð4Þ
its associated effects. The main assumptions of the below- i¼1 j¼1

described approaches are:

where CEi is the cost of ith purchased energy; j is the index of equip-
ment in the energy system; n is the number of pieces of equipment;
 The supply system is already onsite, and there is no need for
kj is the number of maintenance resources deployed during the
further capital investment.
modeling period to the jth equipment; CM j is the maintenance cost
 In order to achieve the optimal energy supplies, there is no need
for equipment upgrade. of jth equipment. Express in the format of matrix
 The energy inputs from the suppliers can satisfy any peak load z2 ¼ J 1m  CE þ J1n  CM ð5Þ
scenarios.
where the ones matrix J is used for summation. CE and CM are the
matrices with elements of CEi and CF i .
3.1. Objectives

3.1.3. Emission – CO2, NOX and SO2

It is unlikely the manufacturers can rely on the renewable
Emission related to energy usage can be quantified through the
energy completely, purchasing energy from suppliers is most of
emission index (also known as the environmental coefficient) with
the cases. Meanwhile the relation between environmental impact
units of kilogram per megawatt hour. For example, the three major
and the energy consumption is well known. While the objectives
emissions – sulfur dioxide, nitrogen oxides and carbon dioxide
of energy cost in terms of megawatt hours, US Dollars or emissions
from electricity in South Carolina, USA can be found in [21]. The
do not always lead to consistent energy management strategy. It is
pollutant effect of the emission can be used as one environment
of importance to understand the analysis and optimization
objective. Sulfur dioxide is the major component in the formation
objectives.
of acid rain. Nitrogen oxides can contribute to acidification and
eutrophication of waters and soils; and when it exits the atmo-
3.1.1. Purchased energy in MWh
sphere, could be the reason of particle matter and ground-level
Energy consumption per unit production is one of the key
ozone formation. Both sulfur dioxide and nitrogen oxide can cause
parameters to evaluate the overall efficiency of the energy usage
health problems in the respiration system. Carbon dioxide is well
from the manufacturing plant. In 1992, U.S. Environmental Protec-
known for its greenhouse gas effect, and series of impacts from glo-
tion Agency (EPA) launched a voluntary program (ENERGY STAR)
bal warming. The objective in this case could be formulated as
that was intended to assist the public to save money and protect
the environment. In this program, fifteen industrial foci compared z3 ¼ EF  E ð6Þ
and published the energy use in the same areas to encourage the
where EF is the emission vector for each of the emission. For exam-
best practice. Therefore, the amount of energy purchased by the
ple, in the latter case where the plant purchases landfill gas, natural
plant is one of the objectives. In this paper, the amount of pur-
gas and electricity. Emission objective z3 can be constructed as
chased energy in the unit of MWh is calculated and the first objec-
2 kg CO2 kg SO2
3 2 3
tive/criterion function can be expressed as: kg NOX
MWh Ele
6 MWh Ele MWh Ele MWh Ele
7 6 7
X
m
z3 ¼ EF  E ¼ 6 kg CO2 kg NOX kg SO2 7
 4 MWh NG 5 ð7Þ
z1 ¼ Ei ¼ J 1m  E ð3Þ 4 MWh NG MWh NG MWh NG 5

i¼1
kg CO2 kg NOX kg SO2 MWh LFG
MWh LFG MWh LFG MWh LFG

where Ei is the amount of purchased energy in MWh and m is the
number of types of energy inputs to the plant.
3.1.2. Cost – operation cost, purchased energy cost
The cost of the energy operation comes from two major
parts – facility maintenance cost [23], and operation source energy
consumption. While the operation source energy consumption is
continuously proportional to the use of primary energy input, the
While considering several pollutants together, it is difficult to
compare the impact each of them could have on the environment.
Simply put, the harmfulness from one gram of carbon dioxide is
not equal to the harmfulness of one gram of nitrogen oxide. Thus,
straightforward summation of these parameters is not an ideal
way to set one objective. In paper [24,25], the concept of pollution
damage cost was introduced. They took the emission from a dis-
trict heat network and calculate the spending on heat pumps,
458 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465
cogeneration and/or gas furnace conversion. The monetary cost per
kilogram of nitrogen oxides and carbon dioxide were calculate
based on their system specifics. The revised third objective func-
tion can be expressed as:
z3 ¼ CEM  EF  E
where CEM is the pollution damage cost matrix [$/kg].
Once the environmental impact is expressed in terms of mone-
tary cost, the third object is combined with the second objective to
formulate the criterion of combined operational and environmen-
tal cost.
z23 ¼ J 1m  CE þ J 1n  CM þ CEM  EF  E
However, the absolute values of pollution damage cost are difficult
to stipulate. The reasons are complex, including but not limited to
the lack of data on pollution damages, and inability to precisely
measure the emissions [26].
3.1.4. Multicriteria objective function
As long as the objectives in the optimization problems are more
than one, the problem is called multicriteria objective optimiza-
tion. In this case, three objectives are:
minðz1 ; z2 ; z3 Þ
There are many ways to deal with the different objective. The
simplest one is to assign equal weights to each of the normalized
objective and sum them to be one objective. In this way, the out-
come will treat each of the objective as having the same
preference.
Z ¼ a1 z1 þ a2 z2 þ a3 z3
Programmers can also change the weights based on decision
makers’ preference on the objectives, or give multiple options to
rank different priorities to the problem.
3.2. Constraint
The constraints of the optimization problem come from three
aspects – equipment capacity, utility supply and production
demand.
3.2.1. Equipment
The existing facility will have a constant number of equipment
available, and the capacity of each piece of equipment is fixed in
certain range. During the optimization, equipment capacity needs
to be set to limit the feasible solutions.
3.2.2. Supply
Energy suppliers and the stability of renewable energy
resources need to be carefully considered in energy system. An
optimal solution outside of the supply capacity is infeasible.
3.2.3. Demand
The energy demand from the manufacturing production line is
one of the most important constraints needing to be satisfied. The
energy demand is not constant; it depends on many factors, such
as the production schedule, productivity, weather conditions, pro-
cess line maintenance, and mixture ratios of the products. For
example, in extreme days like very cold winter days, the energy
demand on the hot water and electricity will be tremendously
high. Optimization to satisfy these extremes is crucial in guaran-
teeing the throughput. It will be beneficial to have an energy
demand forecasting model in order to have a single day or single
week prediction on the coming energy demand. In that case, a
dynamic optimization can be developed in everyday operations
to reach the objectives.
With the onsite energy generation, conversion and transmission
system, the energy demand can be calculated by the amount of
purchased energy through the equipment specifics.
Ei ¼ xi;h  EDh ð12Þ
ð8Þ
where h is the index of energy from production line demand; EDh is
amount of hth the energy demand; xi;h is the facility operational
parameter to convert energy demand from manufacturing pro-
cesses to the plant purchased energy.
4. Automotive assembly manufacturing energy case study
ð9Þ
In this section, the authors use the programming methods
developed in Section 3 to study an automotive assembly manufac-
turing system and illustrate how the approach is applied.
4.1. Case study introduction
The automotive assembly manufacturing plant production lines
can be separated into three main departments – body shop, paint
shop and final assembly shop. The body shop is mainly responsible
for the vehicle body welding. Stamped panels and parts produced
ð10Þ
on site, or from an external supplier will be welded together to a
vehicle body-in-white. In the body shop, energy is used to move
the parts from one location to another, and electricity and com-
pressed air will be used in the welding process. The vehicle
body-in-white from the body shop will be transported to the paint
shop.
The paint shop is reported to be the most energy intensive
ð11Þ
department in the plant [27]. The painting and sealing process will
be deployed in this department to make the vehicle corrosion
resistant and protected. The vehicle body will go through several
painting and sealing process followed by oven curing. A pretreat-
ment tank with warm phosphate solution, booth with controlled
temperature and humidity, and oven with controlled air flow tem-
perature will call for large amounts of energy. Hot water, chilled
water, natural gas, and electricity are typically required in this
department to support the processes.
Final assembly is the department which assembles the vehicle
components and powertrain to the painted body. This department
also needs energy carriers such as electricity and compressed air.
Besides the energy used on the process lines, energy demand in
the plant is also used on building services, mainly – lighting, heat-
ing, ventilation and air conditioning [28].
In summary, the energy carriers’ demand includes: electricity,
natural gas, hot water chilled water and compressed air. The stud-
ied manufacturing plant purchases three energy carries from the
suppliers – electricity, natural gas and landfill gas. Thus, the onsite
energy conversion and transmission should be modeled as a three-
input–five-output system as represented in Fig. 3.
The cluster of onsite energy conversion and transmission sys-
tems is referred to in this paper as the Energy Center, and repre-
sented in Fig. 4.
This figure is a simplification diagram of the case study, which
illustrates a typical multiple input multiple output (MIMO) energy
system. Electricity, natural gas and landfill gas are used as three
energy source inputs to the Energy Center. Electricity, natural
gas, hot water, chilled water and compressed air are the five out-
puts of the Energy Center.
The cogeneration system, which converts burnable fuel to both
electricity and heat, is believed to have an average payback period
of 2–5 years [29,30]. In general, the combined heat and power
(CHP) cogeneration system improves the energy efficiency over
separate systems from traditionally 30% to an encouraging 70%.
The cogeneration system can use many different energy sources,
 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465 459

Fig. 3. Onsite energy conversion and transmission system schema.

Fig. 4. Energy center input – output schema.

such as combustion gas, gasoline, coal, or biofuel, and depends on 4.2. Model implementation
the equipment specifics; in this case the energy source is landfill
gas, and it generates two forms of energy – electricity and hot The Energy Center processes energy from suppliers and delivers
water. Hot water can also be produced from boilers to supplement the desired forms and amount of energy to the production line. The
the combustion gas chemical energy to thermal energy. In this optimization discussed in this section focuses on the energy pro-
case, landfill gas and/or natural gas are used in boilers. The hot cessed in the Energy Center.
water could also be circulated back from energy outputs to the Three energy inputs from the suppliers need to be purchased.
absorption chiller for chilled water production. The introduction Indexes were assigned to each of the three inputs (see Table 1):
of the absorption chiller to cogeneration, is making the whole sys- In this case m = number of indexes = 3. The energy purchase
tem even more efficient [31,32]. In some publications, the incorpo- vector is:
rate of the absorption chiller to the CHP is called tri-generation 2 3 2 3
E1 MWh Purchased Electricity
[32]. 6 7 6 7
Air compressors and centrifugal chillers transform the electric- E ¼ 4 E2 5 ¼ 4 MWh Purchased Natural Gas 5 ð13Þ
ity into compressed air and chilled water respectively. From here, E3 MWh Purchased Landfill Gas
it can be defined that the energy conversion as a process of chang-
Thus Eq. (3) in the Model Establishment section can be written
ing energy forms and qualities; energy pass-through, on the other
as:
hand is defined as a process of delivery energy in the same form 2 3
and quality. Case study energy system includes both energy con- E1
6 7
version and pass-through. z1 ¼ J 13  E ¼ ½ 1 1 1 4 E2 5 ð14Þ
In the processes of energy conversion and pass-through, auxil- E3
iary power is unavoidable. For example, landfill gas from supplier
needs to be pretreated before send to the combustion chamber of
the cogeneration equipment. And during the process of pretreat-
ment, such as gas filtration to eliminate the particle matter, elec- Table 1
Purchased energy indicator assignment.
tricity is used. In this model, the auxiliary procedures like the gas
pretreatment of the cogeneration system are not discussed as an Energy inputs Index i
individual process. Instead, it is taken as a part of the cogeneration. Electricity: 1
And the electricity usage caused by the auxiliary processes is calcu- Natural gas: 2
lated as the conversion loss/inefficiency of the Energy Center. Landfill gas: 3
460 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465

Table 2 ity and natural gas, while the consumption of both energy inputs
2012 South Carolina electricity emission profile [21]. can cause environmental problem. In this case, landfill gas should
Emissions Value (lbs/MWh) be considered as a clean source which helps to prevent emissions
Sulfur dioxide 1.5 by using less electricity and natural gas. Thus, when dealing with
Nitrogen oxide 0.5 the landfill gas emissions, use negative values that represent the
Carbon dioxide 778 emission reduction by replacing the electricity and natural gas
(Table 4).
To simplify the problem, here only the cost from the energy If one of the three major pollutants is used as an objective, the
purchased is considered, excluding the maintenance and degrada- third single objective function can be written as:
tion fees. Eqs. (4) and (5) in the previous section can be written as:
z3 ¼ ðF 3 ÞT  E ð18Þ
X3
z2 ¼ i¼1
CEi ¼ J 1m  CE ð15Þ F 3 could be the vector with any of the emission pollutants. For
example, if choose carbon dioxide as the objective emission,
where the unit price of the three purchased energy used is repre- 2 3
778
sented in the price vector F 2 : 4
F3 ¼ 401 5 when landfill gas is only used in cogeneration sys-
2 3
60 447
6 7 tem, with unit of pound per megawatt hour. F 3 can also be any
F 2 ¼ 4 30 5 ð16Þ
combined emission factors to represent the degree of harm from
15
each energy to the environment. For example, in paper [20],
F 2 has the unit of US Dollars per megawatt hour. authors use the unification of damage cost to combine the factors
The elements in the vector CE – cost of energy, and the second from different emission pollutants. However, in this case study, the
objective (z2 ) function can be expressed as follows: emissions from different pollutant is not combined because of the
big gaps among the countries and years.
z2 ¼ ðF 2 ÞT  E ð17Þ

(Note: Superscript T represents matrix transpose.) 4.2.1. Efficiency

In terms of the environmental objective. The purchased electric-
ity and natural gas can be converted to environmental emissions
based on how the electricity was generated and natural gas com-
position respectively. According to the US Environmental Informa-
tion Administration, in year 2012, South Carolina has the electricity
emission profile as shown in Table 2.
Natural gas has a range for the emission profile according to the
boilers used and the quality of natural gas. These values are given
in Table 3.
The emission caused by the landfill gas used is worthy of dis-
cussion. The main components in the landfill gas are methane
and carbon dioxide. Additionally, there are many unstable compo-
sitions of both organic and inorganic compounds. The combustion
of landfill gas will release greenhouse gases and other emissions
into the atmosphere. However, the landfill gas is produced from
landfill, without centralized collection and pretreatment for the
later use, it would be discharged to the environment directly
[34]. On the other hand, due to the consumption of landfill gas,
the manufacturing plant does not need to purchase more electric-
Equipment efficiencies and energy conversion ratios are repre-
sented as an energy-equipment coefficient and all together
denoted in the 6  8 matrix Coeff as shown below.
 
Coeff ¼ ci;j ; i ¼ 1; 2; . . . ; 6; j ¼ 1; 2; . . . ; 8 ð19Þ
Consider ci;j to represent the coefficient of ith energy carriers with
jth equipment. For example, the hot water produced through CHP
has the efficiency of 40%. Here, i refers to the hot water generated,
j refers to the CHP system, and ci;j ¼ 0:4 in the matrix. For a specific
energy supply system, the coefficient matrix can be obtained from
the equipment manual or energy monitoring system.
4.2.2. Constraints
Energy supplied from the Energy Center to the production line
can be expressed as:
S ¼ TF  ½Coeff  X   B  X ð20Þ

Table 3
Natural gas emission profile [33].

Emissions Volumetric mass (lbs/106scf)a Power-normalized mass (lbs/MWh)

Sulfur dioxide 0.6b 0.0020
Nitrogen oxide 32–100c 0.11–0.33
Carbon dioxide 120,000d 401.39
a
The value is based on an average natural gas high heating value of 1020 Btu/scf.
b
Based on 100% conversion of fuel sulfur to SO2, with the natural gas sulfur content of 2000 grains/106 scf.
c
Based on small boilers with <100 MMBtu/h heat input. The value range is caused by the NOx control condition: 100 is uncontrolled,
32 is low NOx burner with flue gas recirculation.
d
Based on 100% carbon converted to CO2.

Table 4
Landfill gas emission profile.

Emissions Through cogeneratione Value (lbs/MWh) Through boilerf value (lbs/MWh)

Sulfur dioxide 0.45 0.0020
Nitrogen oxide (0.21) to (0.33) (0.11) to (0.33)
Carbon dioxide 447.47 401.39
e
Based on 30% efficiency in electricity, 40% efficiency in hot water, and 75% efficiency in boilers.
f
Based on the same efficiency in boilers for both gases.
 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465 461

where TF is 5  6 transfer matrix; Coeff – 6  8 Coefficient matrix; X

– 8  1 Equipment/Energy Center consumption vector (it is pro-
cessed as the form of energy, and in unit of MWh), and B is 5  8
inner loop matrix that represents the amount of energy cycling
inside of the Energy Center.
Demand (D) should be no more than the output of the Energy
Center (S), i.e. D 6 S, where D – 5  1 Energy demand vector.
Assume the optimization is developed on a daily basis. The
average energy demand distribution to the three departments is
shown in Fig. 5. In order to protect the confidentiality of opera-
tional data at the OEM, a nominal representative value of daily
energy demand ED from the major plant was chosen, and all energy
data can be normalized to this value.
Aside from major plant demand, constraints also come from
capacities. Constraints from equipment lower bounds and upper
bounds can be defined through the matrix X : LB 6 X 6 UB as
shown in Table 5. Fig. 5. Energy demand distribution.
The lower bound is assumed as the situation when the plant is
completely shut down and the only electricity consumption is to
where
make sure the plant and its facilities are protected from damage.
The upper bound is assumed as the equipment and/or supply
capacity. C ¼ F  T; C 2 R18
In addition, the transfer function T is used to transform the
energy consumed by equipment to energy purchased from suppli- 2 3
Coeff
ers, as
2 3 6 7
A¼4 I 5; A 2 R198 ; I is the identity matrix; I 2 R88
Co generation
6 Boiler ðLFGÞ 7 I
6 7 2 3
6 7 D0
2 3 2 36 Boiler ðNGÞ 7
Electricity    6 7 6 7
6 7 b ¼ 4 UB 5; b 2 R191
6 7 6 .. . . .. 76 Absorbtion Chiller 7
4 NG 5 ¼ 4. . . 56 7 LB
6 Centrepital Chiller 7
LFG  6 7
6 Air Compressor 7 Because of the complex dimensions in this linear optimization
6 7
6 7
4 Pass through Gas 5 problem, a traditional geographic method is difficult to apply to
generate the solution. However, once the linear optimization prob-
Pass through Electricity
2 3 lem is structured as in the above format, many solution methods
2 3 x1
t 11 . . . t 17 6 7 can be applied, such as the primal–dual relationships method
6 . .. . 7 6 x2 7 and simplex method by further changing the constraints to be
¼6
4 ..
76 7
. .. 56 .. 7 ð21Þ equalities. And there is also a large body of software available for
4 . 5
t 31    t 38 solving the problem, such as MATLAB, and AMPL. In this study,
x8 MATLAB was used to solve the optimization problem.

E T X
4.3. Results
T – 3  8 Transfer matrix (transfer equipment/Energy Center
consumption vector X to E). Table 6 summarizes the single optimization results.
Table 6 gives out the optimization results on three objectives
E – 3  1 Purchased Energy vector. and also the resultant energy supply, monetary cost, and carbon
dioxide emission vary while operating the Energy Center under
each objectives strategy. The minimum amount of energy is used
4.2.3. Summary when the objective is set to be MWh of energy; however, the cost
To write the linear optimization problem in canonical form: to purchase the minimum amount of energy is 35% more than the
min CX result of monetary cost oriented optimization, while the emission
s:t:; AX P b is about 45.38 times of the environment oriented optimization.
XP0 Likewise, the economy and environment oriented optimization
gives out results with the minimum US Dollar cost and kg of CO2
released, but has a higher (17% higher) results on the amount of
Table 5
Lower and upper bounds.
Table 6
(MWh) Lower bound Upper bound
Optimization results.
Cogeneration 0 0:70ED
Boiler (LFG) 0 0:70ED Result
Boiler (NG) 0 0:25ED Objective Supply (MWh) Cost ($) Emission (kg CO2)
Absorption chiller 0 0:02ED
Centrifugal chiller 0 0:04ED Energy E 1:35C 45:38EM
Air compressor 0 0:04ED Economy 1:17E C EM
Pass through Gas 0 Infinity Environment 1:17E C EM
Pass through electricity 0:03ED Infinity
E: energy in MWh; C: cost in US Dollars; EM: emission in kg CO2.
462 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465
purchased energy in terms of megawatt hours. It proves the con-
flict among different objectives, and quantifies the differences.
It is worth paying attention to the results of the emission objec-
tive. In this automotive assembly plant, landfill gas is used as a
renewable energy to generate hot water and electricity. Without
the consumption of landfill gas, more electricity and natural gas
will be used. In this consideration, the emission factor of landfill
gas was set to be negative. Thus extra constraints need to be set
to avoid the problem of misapplication the landfill gas to decrease
plant emission. In the results of environmental emission as the
objective, the negative result of emission will be achieved, since
the system will automatically use more landfill gas over the elec-
tricity directly from the grid.
This table shows three objectives leads to various energy supply
operation strategies, where the results of operation could lead to
significant different in the energy consumption, carbon dioxide
emission, and US dollar cost.
4.3.1. Energy demand
A good energy demand forecasting for the next few steps (usu-
ally in days) are critical in energy operation strategy on the supply
side. The traditional energy demand forecasting techniques are
based only on the historical records and cannot typically satisfy
the accuracy requirement. An inaccurate forecasting of the energy
demand can lead to complete waste in energy supply operations or
supply failures, which can result in tremendous monetary loss.
Energy demand of the manufacturing plant depends on many vari-
ables; some key inputs are – the production rate, production
schedule, working shifts, maintenance, and weather condition
[35]. Incorporating this extra knowledge into the traditional time
series model can make the forecasting more robust and reliable.
A time series model is proposed in [36] to predict the energy
demand for a given time horizon. By combining the energy
demand forecasting and energy supply optimization, manufactur-
ers can create a more informed strategy for the production
scheduling and save tremendous amount of energy, as well as
monetary cost and carbon dioxide emission.
The original data used for demand is the plant running in a 2-
shiftworking load (results shown in Table 6). It is common for
the manufacturer to reduce the shifts for holidays and have fewer
production planned days. However, the energy demand for half
production does not usually equate to half the expected work load.
Assume the energy used in one-shift working days use about 60%
of energy as a full production day, and the energy forms distribu-
tion keeps same breakdown as indicated in Fig. 5.
Table 7 shows the percentages of each energy form in terms of
daily energy demand ED (2 shifts) for normal production days. It
compares the energy supply in two working load scenarios.
When the objective is to minimize the amount of purchased
energy per produced vehicle, the optimization results will abandon
the cheap, clean landfill gas and choose to use electricity and nat-
ural gas directly. While the optimization objective is either econ-
omy or environment, except for natural gas, electricity and
landfill gas do not reduced proportionally as the demand. They
Table 7
Effects of Demand on Energy Supply.
(in ED ) Energy Economy Environment
Supply 2 Shifts 1 Shift 2 Shifts 1 Shift 2 Shifts
(%) (%) (%) (%) (%)
Electricity 44 26 23 5 23
Natural 61 37 30 18 30
gas
Landfill 0 0 70 70 70
gas
use the maximum capacity of the Energy Center to achieve the dif-
ferent goals. For example, when the optimization objective is the
environment protection, the result show the landfill gas purchase
amount is the same as the 2-shift working load which lead to the
small amount of electricity demand from the grid.
Except for the working load change, there are many other rea-
sons influencing the energy demand, such as the production rate
(number of vehicles produced per day), weather condition (sea-
sonal changes and extreme weather days), and implementation
of energy intensive or energy saving equipment. The model can
be easily applied to test energy operation strategies according to
the different reasons that cause the demand change, by changing
the demand matrix accordingly.
Except for the objective priorities, the energy demand also
affects the operation of onsite supply system.
4.3.2. Economy
Economy is crucial to the manufacturing plant. Low monetary
spending on the energy of the plant results in a more profitable
product. However, the cost of energy is affected by both higher
and lower level – purchase energy unit price from the suppliers
and energy demand from the production line.
The industrial average retail price for natural gas from January-
2001 to January-2015 has ranged from $3.02 to $13.06 per thou-
sand cubic feet; the heat content of NG is about 1030 BTU/ft3.
The industrial average retail price for electricity from January-
2001 to January-2015 has ranged from 4.71 to 7.72 cents/kWh
(Table 8).
Table 8 shows the energy unit price range for both natural gas
and electricity. Assume the landfill gas is the half price of natural
gas.
The effect of unit price of landfill gas on the operation strategies
is studied.
A continuous change in the landfill gas unit price show the
operation strategies change. From Fig. 6, From Fig. 6, the landfill
gas equation can be summarized into three parts:
8
< LFG Price < 18 MWh
> USD
USD
18 MWh 6 LFG Price 6 21 MWh
USD
>
:
USD
21 MWh 6 LFG Price
In the first partition, when the landfill gas is inexpensive, the
optimization results in running the full capacity of cogeneration
system, even when the produced hot water is greater than the
demand from the production line. In this way the total cost is still
minimized, because the cogeneration system produces the maxi-
mum amount of electricity.
In the second partition, even though the price of landfill gas
increases, it still shows a high running rate. This is because using
the extra hot water produced from the cogeneration system to
run the absorption chiller is still less expensive than the cost of
running centrifugal chiller by using grid power.
In the third partition, where the landfill gas is higher than
21 USD per MWh, the cogeneration only runs to give out enough
hot water for the production line and the corresponding electricity.
This is still less expensive than running the boiler and purchasing
electricity from the grid.
Further examination also indicates that only when the unit
price of landfill gas is larger than 33 USD/MWh, operators should
1 Shift
(%) Table 8
5 Energy unit price range.
18
USD/MWh Min Max
70 Natural gas 10.0 43.3
Electricity 47.1 77.2
 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465
Fig. 6. Effect of landfill gas unit price on purchased energy.
Fig. 7. Combined effect of electricity and natural gas unit price on purchased
energy.
Fig. 8. Combined effect of electricity and natural gas unit price on cogeneration
operation.
463
Fig. 9. Combined effect of electricity and natural gas unit price on overall energy
cost.
refrain from quit using the cogeneration system completely and
choose to directly purchase grid power and natural gas instead.
And it is worth noting that the optimization result is highly related
to the unit prices of the different purchased energies and also the
efficient in energy conversion and transmission. In some cases,
when the efficiencies of the equipment degrade as the time passes,
the operators need to verify if the previous operation strategy still
results as in the desired state.
To better understand the effect from the purchased energy unit
prices, analyses of electricity and landfill gas prices and how they
together will affect the optimal results are given below (Figs. 7–9).
Obviously the overall energy cost will increase along with the
unit price of natural gas and electricity. It worth noting that the
usage of electricity and landfill gas changed over the price. The rea-
sons of shift from one energy to another is the same as explained in
Fig. 6.
4.3.3. Environment
The environmental emission is measured as the weight of the
carbon dioxide in this section. The optimization results can be
achieved through the single objective optimization as demon-
strated in Sections 4.3.1 and 4.3.2. It worth to pay attention that
the emission lead optimization also shows a discrete operation
strategies as the adjustment of the emission parameters to each
of the purchased energy. In calculating different environmental
influences, changing the coefficient vector of F.
4.3.4. Multicriteria optimization (MOP)
The decision makers will have multiple objectives in the real
world energy management. Optimum operation strategies for the
minimum energy consumption in terms of MWh do not necessary
lead to the optimal result of energy cost. Multicriteria optimization
is introduced in this section to illustrate how different objectives
can be involved according to the priority of energy managers.
minðz1 ; z2 ; z3 Þ ð22Þ
Among many MOP techniques, weighted sum scalarization tech-
nique is one of the most preferred:
minðZÞ ð23Þ
where = a1 z1 þ a2 z2 þ a3 z3 , weights (a1 ; a2 ; a3 ) are assigned to each
objective as the priority in optimization.
Besides the weighted sum optimization, the e-constraint
method is another one commonly used.
464 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465

minðzk Þ ð24Þ Table 9

Multi-objective optimization results.
Subject to a1 a2 a3 P1 P2 P3

zk 6 ek0
0 0 1 1 0 0 665 918.75 0
k ¼ 1; . . . ; p k – k
2 0 1 0 357.5 450 1012.5
3 0 0 1 357.5 450 1012.5
where e 2 Rp . In this method, the energy managers can optimize the 7 1/3 1/3 1/3 357.5 450 1012.5
target objective to be minimal and control the other parameters 8 8/10 1/10 1/10 383.75 450 937.5
low. For example, the e-constraint method can be used to minimize
the cost, while control the energy consumption and emission within
the certain thresholds. solutions for this bi-objective problem. To better understand the
A plot of the objectives in both decision and criteria space is relationship between the two objectives, the criteria outcome
given below. In the decision space, the objectives are plotted in space is constructed in Fig. 11.
the vertical axis and the constraint is in the horizontal axis; while In such a case with an infinite number of efficient points, the
in the criteria space, one objective is plot in the vertical axis and decision makers’ preference can be applied to choose a preferred
the other is plotted in the horizontal axis. This method gives a bet- solution.
ter understanding on the relation between the constraint and Normalizing the F’s and setting different weights to each of the
objectives, and between two different objectives. It is only feasible objectives, the result below is achieved:
when the objective is limited to two. Here, the objective of energy Table 9 gives the multi-objective optimization by giving sets of
in MWh and cost in USD are selected for demonstration. weights to each of the three objectives. It is interesting to see the
By subtracting the mean values of two objectives, the decision discrete change in the result of purchased energy. Energy man-
space multicriteria optimization problem is shown as Fig. 10. In agers can select different operation strategies based on their vari-
the constraint range of X1, all of solutions are Pareto efficient ous priorities of energy, cost, and environment.

5. Conclusion and discussion

In this paper, related studies on energy conversion, manage-

ment, simulation, and optimization as summarized. General mod-
eling approaches with centralized energy conversion and
transmission department were given. A case study from an auto-
motive assembly plant with a relatively complex three energy form
inputs and five energy form output, MIMO system was built to
study the energy supply system as a representative manufacturing
plant. Both single objective optimizations and multi-objective opti-
mizations were described in this paper. Optimization of energy,
economy, and environment were analyzed, and shown to be in
conflict – in an energy oriented operation, it takes 35% more in
monetary cost; while in an economy oriented operation, it takes
17% more in megawatt hour energy supply and tends to rely more
on the inexpensive renewable energy.
Energy optimization results were proved to be conflicted among
objectives (energy in MWh, US Dollars and carbon dioxide emis-
sion). The onsite energy generation system needs to adjust its
Fig. 10. Decision space.
operation strategies according to both higher level – energy mar-
ket, and lower level – production energy demand. Both the deci-
sion makers’ strategic priorities, equipment efficiency are also
critical in the optimal operation strategies.

Acknowledgement

The authors would like to thank BMW Manufacturing Co. for

their project sponsorship and access to the manufacturing facility.

References

[1] U.S. Energy Information Administration. Annual Energy Review.

[2] Introna V, Cesarotti V, Benedetti M, Biagiotti S, Rotunno R. Energy management
maturity model: an organizational tool to foster the continuous reduction of
energy consumption in companies. J Clean Prod 2014;83:108–17.
[3] Sanders DR, Irwin SH. Energy futures prices and commodity index investment:
new evidence from firm-level position data. Energy Econ 2014;46(1):57–68.
[4] Dixon RK, McGowan E, Onysko G, Scheer RM. US energy conservation and
efficiency policies: challenges and opportunities. Energy Policy 2010;38
(11):6398–408.
[5] Bruzzone AAG, Anghinolfi D, Paolucci M, Tonelli F. Energy-aware scheduling
for improving manufacturing process sustainability: a mathematical model for
Fig. 11. Criteria space. flexible flow shops. CIRP Ann Manuf Technol 2012;61(1):459–62.
 L. Feng et al. / Energy Conversion and Management 117 (2016) 454–465
[6] Pach C, Berger T, Sallez Y, Bonte T, Adam E, Trentesaux D. Reactive and energy-
aware scheduling of flexible manufacturing systems using potential fields.
Comput Ind 2014;65(3):434–48.
[7] Li L, Sun Z, Yang H, Gu F. Simulation-based energy efficiency improvement for
sustainable manufacturing systems. In: ASME 2012 international
manufacturing science and engineering conference. Notre Dame (IN); 2012.
p. 1–7.
[8] Duflou JR, Sutherland JW, Dornfeld D, Herrmann C, Jeswiet J, Kara S, Hauschild
M, Kellens K. Towards energy and resource efficient manufacturing: a
processes and systems approach. CIRP Ann Manuf Technol 2012;61
(2):587–609.
[9] Ghadimi P, Kara S, Kornfeld B. The optimal selection of on-site CHP systems
through integrated sizing and operational strategy. Appl Energy
2014;126:38–46.
[10] Saidur R, Ahamed JU, Masjuki HH. Energy, exergy and economic analysis of
industrial boilers. Energy Policy 2010;38(5):2188–97.
[11] Yu FW, Chan KT. Improved energy performance of air cooled centrifugal
chillers with variable chilled water flow. Energy Convers Manage 2008;49
(6):1595–611.
[12] Zahlan J, Asfour S. A multi-objective approach for determining optimal air
compressor location in a manufacturing facility. J Manuf Syst 2015;35:176–90.
[13] Vepa R. Dynamic modeling, simulation and control of energy
generation. London (UK): Springer; 2013.
[14] Carnevale EA, Ferrari L, Paganelli S. Investigation on the feasibility of
integration of high temperature solar energy in a textile factory. Renew
Energy 2011;36(12):3517–29.
[15] Li J, Wei W, Xiang J. A simple sizing algorithm for stand-alone PV, wind, battery
hybrid microgrids. Energies 2012;5:5307–23.
[16] Wang J, Jing Y, Zhang C, Zhao J. Review on multi-criteria decision analysis aid
in sustainable energy decision-making. Renew Sustain Energy Rev 2009;13
(9):2263–78.
[17] Ehrgott M. Multicriteria optimization. Berlin (Germany): Springer; 2005.
[18] Alarcon-Rodriguez A, Ault G, Galloway S. Multi-objective planning of
distributed energy resources: a review of the state-of-the-art. Renew Sustain
Energy Rev 2010;14(5):1353–66.
[19] Buoro D, Casisi M, De Nardi A, Pinamonti P, Reini M. Multicriteria optimization
of a distributed energy supply system for an industrial area. Energy
2013;58:128–37.
[20] Lazzaretto A, Toffolo A. Energy, economy and environment as objectives in
multi-criterion optimization of thermal systems design. Energy 2004;29
(8):1139–57.
[21] U.S. Energy Information Administration. South Carolina electricity profile;
2012.
465
[22] U.S. Energy Information Administration. United States natural gas industrial
price.
[23] Owayedh MS. Energy cost based technique for maintenance scheduling of
generating systems. In: Presented at electrotechnical conference proceedings
of the 7th Mediterranean; 1994.
[24] Curti V, von Spakovsky MR, Favrat D. An environomic approach for the
modeling and optimization of a district heating network based on centralized
and decentralized heat pumps, cogeneration and/or gas furnace, part I:
methodology. Int J Therm Sci 2000;39(7):721–30.
[25] Curti V, Favrat D, von Spakovsky MR. An environomic approach for the
modeling and optimization of a district heating network based on centralized
and decentralized heat pumps, cogeneration and/or gas furnace, part II:
application. Int J Therm Sci 2000;39(7):731–41.
[26] U.S. Environmental Protection Agency. National Center for Environmental
Economics – Pollution Charges, Fees, and Taxes.
[27] Rivera JL, Reyes-Carrillo T. A framework for environmental and energy analysis
of the automobile painting process. Proc CIRP 2014;15:171–5.
[28] Liu Hao, Zhao Qianchuan, Huang Ningjian, Zhao Xiang. A simulation-based tool
for energy efficient building design for a class of manufacturing plants. IEEE
Trans Autom Sci Eng 2013;10(1):117–23.
[29] Zarnikau J, Reilley B. The evolution of the cogeneration market in Texas.
Energy Policy 1996;24(1):67–79.
[30] Sun Z. Energy efficiency and economic feasibility analysis of cogeneration
system driven by gas engine. Energy Build 2008;40(2):126–30.
[31] Marques RP, Hacon D, Tessarollo A, Parise JAR. Thermodynamic analysis of tri-
generation systems taking into account refrigeration, heating and electricity
load demands. Energy Build 2010;42(12):2323–30.
[32] Tassou SA, Chaer I, Sugiartha N, Ge Y-, Marriott D. Application of tri-generation
systems to the food retail industry. Energy Convers Manage 2007;48
(11):2988–95.
[33] U.S. Environmental Protection Agency. Emissions factors chapter 1: external
combustion sources.
[34] Rosli NAM, Kamarrudin N, Ku Hamid KH, Akhbar S, Yusof NM. Greenhouse gas
emission of MSW landfill site. In: Presented at IEEE business engineering and
industrial applications colloquium (BEIAC); 2013.
[35] Cheng-Wen Yan, Jian Yao. Application of ANN for the prediction of building
energy consumption at different climate zones with HDD and CDD. In:
Presented at 2nd international conference on future computer and
communication (ICFCC); 2010.
[36] Feng L, Mears L. Time series analysis and forecasting of manufacturing energy
demand. In: Proceedings of the 2015 industrial and systems engineering
research conference. Nashville (TN); 2015.