Sie sind auf Seite 1von 11

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

Considering Carbon Emissions in Economic Dispatch Planning for Isolated Power


Systems-A Case Study of the Taiwan Power System

Ming-Tse Kuo Shiue-Der Lu Ming-Chang Tsou


Member, IEEE Member, IEEE
Department of Electrical Engineering Department of Electrical Engineering Leadtrend Technology Corporation.
National Taiwan University of Science Chung Yuan Christian University No.1, Taiyuan 2nd St., Zhubei City
and Technology Taoyuan City, Taiwan, R.O.C. Hsinchu County, Taiwan, R.O.C.
Taipei City, Taiwan, R.O.C. SDL@cycu.edu.tw ming@leadtrend.com.tw
mkuo@mail.ntust.edu.tw

Abstract -- By integrating simulated annealing and the increased. Fossil fuels are the primary source of both electric
interactive best-compromise approach, this study proposed a power and greenhouse gases in Taiwan. Hence, reducing
method for determining economic dispatch in power systems greenhouse gas emissions while satisfying the demand for
while accounting for carbon emissions. To satisfy load demands electric power and ensuring economic growth has become a
and operational constraints, traditional economic dispatch is vital concern.
determined through the minimization of total power generation
Previous studies on commonly used methods for solving
costs based on the power output of the generating units in a
system. In the proposed method, carbon emissions are taken into unit commitment problems have used numerical optimization
consideration. A CO2 equivalent model was created that techniques and stochastic search methods [4]-[6]. Numerical
accounted for the various fuel types (e.g., coal, oil, and natural optimization techniques, such as the priority list method [7]-
gas) used in power generation. The correlation between power [8], dynamic programming [9]-[10], and branch-and-cut [11]-
generation costs and carbon emissions was determined [13] offer fast convergence and easy direction to local
according to CO2 emissions tradeoff and incremental costCO2 optimums. However, they are unable to solve problems with
reduction curves. The proposed method was applied to the broad constraint conditions. Stochastic search methods, such
Taiwan Power Company system. The results indicated that the as artificial neural networks, particle swarm optimization, and
method was effective at determining the influence of CO2
emissions on power generation costs during off-peak, semi-peak,
simulated annealing (SA) can be applied to both
and peak hours, as well as daily load demands. The system differentiable and non-differentiable objective functions, as
enables the simultaneous consideration of economic and well as to convex and nonconvex sets. Stochastic search
environmental benefits. methods can be used to determine both successive and
discrete types of equality and inequality constraints.
Index Terms-- economic dispatch, simulated annealing, Additionally, stochastic search methods can trip local
carbon emission, bi-objective planning, isolated power system. optimums and approach global optimums.
SA was adopted in [14] as the main algorithm for
randomly generating a commitment solution. Subsequently,
I. INTRODUCTION local search was used to accelerate search efficiency and
The greenhouse effect is a global threat. The United convergence speed in determining neighborhood solutions.
Nations Framework Convention on Climate Change SA was used in [15] to solve multiarea unit commitment
(UNFCCC) was passed in 1992 in an attempt to reach a problems with optimal system security and reliability. The
consensus on this problem. The UNFCCC regulates the internal mechanisms of SA were modified in [16] and [17] to
density of greenhouse gases [1]. At the 18th United Nations increase constraints and reliability, incorporate dynamic
Climate Change Conference, held on December 8, 2012 in economic dispatch methods, and improve search ability. An
Doha, Qatar, an amendment to the Kyoto Protocol (the Doha absolutely stochastic SA method was proposed in [18] to
Amendment) was proposed. Planned in two stages (2008 solve unit commitment problems. Search speed and reliability
2012 and 20132020), the Doha Amendment asked were enhanced through fuzzy control and appropriate
signatories to promise to reduce greenhouse gas emissions by condition constraints. These studies have confirmed the
18% of 1990 levels. The UNFCCC and Kyoto Protocol were applicability of SA for solving thermal unit commitment
promulgated as internationally binding agreements to reduce problems. Therefore, this method was adopted to compute
CO2 emissions and ameliorate the greenhouse effect [2]. The unit commitment in this study.
Bureau of Energy, Ministry of Economic Affairs, Taiwan Because the Taiwan power system is isolated and not
aims to reduce carbon emission levels in 2020 and 2025 to supported by other interconnected systems, its power
their levels in 2005 and 2000, respectively [3]. Economic generation is largely based on fossil energy (e.g., fuels and
development in Taiwan has made electric power fuel oil) produced in thermal units, which generates large
indispensable, and annual demand has correspondingly amounts of CO2. The recent upsurge in environmental
consciousness and worldwide awareness regarding the

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

greenhouse effect has positioned carbon emission concerns at estimate load demand for off-peak, semi-peak, peak, and
the fore of global consciousness, incorporating these daily periods while accounting for the amount of CO2
problems into the realm of power dispatch. Energy policies emissions in these periods. The results for economic dispatch
must balance costs and environmental protection concerns suggested that the generators can be operated in a manner that
when satisfying power demands. In one unit-commitment yields environmental and economic benefits while accounting
study [19], the waste gas emission of units during operation for the effects of the amount of CO2 emissions on electricity
was recorded using a Lagrangian relaxation algorithm to generation costs. The analysis model and results of this study
ameliorate the waste gas problem. The constraints of CO 2 are expected to serve as a reference for Taipowers power
emissions were explored in [20]; carbon emission formulas grid dispatch and planning.
were used to determine unit commitment in the short term
and the optimums were obtained using SA. Generator II. ECONOMIC DISPATCH MODEL FOR THERMAL
maintenance scheduling was researched in [21] to enable the GENERATION UNITS
reduction of carbon emissions. Power Generation Cost Functions
Traditional economic dispatch methods cannot be used to Power generation data provided by the Taiwan Power
account for CO2 emissions, because CO2 emissions affect Company are typically expressed using heat-rate functions.
dispatch plans. To obtain optimums, multiobjective planning As shown in (1), the power generation function was the cubic
must be used to allow both costs and CO2 emissions to be expression of the heat-rate function, multiplied by the unit
simultaneously considered. Mathematical methods were energy cost (E).
employed in [22] to generate criteria weights and rank them (1)
by importance. This simple framework is applicable to
multiobjective decision-making. The optimal dispatch where
method for hydroelectric power generation was obtained in represents the fuel cost of the ith unit during interval t.
[23] by minimizing system security risks and power represents the power generation volume of the ith
generation costs, and by adopting a genetic algorithm to solve unit during interval t.
bi-objective optimums. In [24], profit-based unit commitment represents the unit energy cost .
under the emissions limit was determined using a When the number of thermal generation units was set as m,
multiobjective planning approach; weighting and scaling the power generation cost during interval t can be expressed
factors were employed to determine the optimums, and as (2):
tradeoff curves were adopted to assist in decision-making. (2)
Previous studies that have considered the cost of CO2
emissions in power generation have typically implemented CO2 Emission Model
planning approaches using weighted values, in which The power generation cost function of thermal power
objective functions are assigned different weights based on generation units, which is expressed as a cubic function, can
importance. The functions are multiplied by the assigned be obtained by converting the heat-rate function. The CO2
weights and summed into a single objective function before emission function is also expressed as a cubic function and
execution. However, power generation costs and CO2 acquired by converting the heat-rate function, as expressed in
emissions are expressed in different units, and one cannot be (3):
directly converted to the other, hindering the determination of
representative weight values. (3)
Therefore, this study proposed an interactive best- where
compromise method with SA that effectively allowed unit represents the CO2 emissions of the ith unit.
differences between objective functions to be neglected, represent the CO2 emission coefficients of
eliminating the need for weight determination. Furthermore, the ith unit.
this method is responsive to user demands. When solutions represents the power generation of the ith unit during
provided by the method are unsatisfactory, a user can modify interval t.
the objective functions until a satisfactory solution is yielded. Fuels have different power generation costs and result in
The interactive best-compromise method with SA proposed varying levels of CO2 emissions (Table I) [25]. To cut costs,
in this study incorporates user demands and constraints to coal is generally used for power generation; as shown in
identify optimal dispatch strategies that reflect power Table I, power generation using coal yielded the lowest cost,
generation costs and carbon emissions. only NT$218.7/Gcal, but the highest volume of carbon
In January 2017, the Taiwan government promulgated emissions, followed by oil, then gas, which yielded the lowest
amendments to the Electricity Act, which include listing levels of CO2 emissions. Therefore, using gas is superior for
renewable energy as a crucial role for development and cutting emissions.
defining and estimating carbon emissions coefficients.
Accordingly, this study used realistic data on the operation of
20 thermal generators from the Taiwan Power Company
(Taipower), the state-owned electric utility in Taiwan, to

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

TABLE I. conversion coefficients are used to calculate CO2 emissions.


POWER GENERATION COSTS AND CO2 EMISSIONS OF VARIOUS FUELS
The heat rate of thermal units can be calculated by
Cost Carbon emission
Fuel type multiplying the fuel usage volume by the heat value. Heat
(NT$/Gcal) (kg/Gcal)
Coal 218.7 388.2 rates describe power unit operation, and can be obtained from
Oil 490.9 320.7 power companies. Aside from heat rates, other conversion
Natural gas 701.5 233.7 coefficients were simplified as constants and defined as CO2
In economic dispatch planning, the characteristics of emission factors ( ) (6).
various fuel thermal units are considered to comprehensively (kg/Gcal) = 4.1868(J/cal) carbon emission coefficient (kg/GJ)
determine bi-objective (i.e., power generation costs and CO2 (44/12) (6)
emissions) power planning concerns, striking a balance Following (6), the CO2 emission formula can be derived
between economic and environmental concerns. from the heat rate function. The emission conversion
coefficient in the formula was simplified as the heat rate
Formulas to Calculate CO2 Emissions function multiplied by the CO2 emission factor. Accordingly,
Thermal power plants emit CO2. Assume that in the power the CO2 emission function was obtained. The CO2 emission
system, the CO2 emitted by the ith unit in a power system factor varied by fuel type. Thus, (7) can be obtained by
during interval t can be calculated using (4) [25]: substituting the heat rate and CO2 emission factor into (3) and
various CO2 emission mathematical models can be obtained
(4) by substituting various CO2 emission factors.
where
(7)
represents the CO2 emissions of the ith unit
where
represents the power generation volume of the ith
represents the CO2 emissions of the ith unit during
unit during interval t.
represents the fuel usage volume of the ith unit interval t.
(ton). represents the power generation volume of the ith
unit during interval t.
represents the heat value of the ith unit (kJ/kg).
represents the carbon emission coefficient of the ith represents the heat rate
unit (kg/GJ). function (Gcal/h).
represents the carbon oxidation rate of the ith unit. represents the CO2 emission factor (kg/Gcal).
Because carbon emission coefficients have not been The number of thermal units was set as m. The total CO2
researched or developed in Taiwan, coefficients published by emission amount of the system during interval t was
the Intergovernmental Panel on Climate Change (IPCC) expressed as (8).
regarding thermal units for different fuel types were used. (8)
The CO2 emission calculation formulas [26] provided by the
IPCC are expressed in (5). Economic Dispatch Planning Restriction Conditions
1) System Constraints
CO2 emission (kg/h) = fuel usage volume 4.1868 The balance between supply and demand, as well as the
carbon emission operating reserve, must account for the particularities of
coefficient (44/12) (5) power systems. The total power generation volume of power
where generation units must satisfy load demands. Two power
44/12 represents the weight ratio of CO2 to carbon system situations were devised: Situation 1, in which line
molecules. losses were not considered, and Situation 2, in which line
4.1868 represents the conversion of heat rate units to losses were considered. They were modeled as follows.
energy units (1 cal = 4.1868 J). Situation 1: Line losses are not considered.
TABLE II. Power generation units are near the load center, the
CARBON EMISSION CONVERSION COEFFICIENTS
Carbon emission Carbon transmission lines are short, and line losses during the
Fuel Heat value transmission process comprise a comparatively small
coefficient oxidation
type (kJ/kg)
(kg/GJ) rate proportion of the total power generation volume. This is true
Coal- of high-density urban power systems. In this situation, line
Coal 25.8 0.98 29308
fired
Heavy
losses are negligible and not considered in economic
21.1 Oil-fired 0.99 41031 dispatch. The supplydemand-balance constraint is expressed
oil
Light
20.2 Gas-fired 0.995 31736
as (9).
oil
Natural
15.3 (9)
gas
where
Table II shows the CO2 emission conversion coefficients, represents the summation of power volume generated
including carbon emission coefficients, carbon oxidation by units.
rates, and heat values. In (5), fuel usage volumes and other represents the total load demands.

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

Situation 2: Line losses are considered represents the maximal power generation volume of
Power generation units are far from the load center, the ith unit.
transmission lines are long, and line losses during the
transmission process comprise a comparatively large 3) CO2 Emission Constraints
proportion of the total power generation volume. Therefore, The CO2 emissions of power units were calculated and
these line losses cannot be considered negligible and must be emission restrictions were planned. The optimal economic
calculated. The supplydemand-balance constraint is dispatch planning results were subsequently obtained. The
expressed as (10). CO2 emissions constraint is expressed as (14):
(10)
(14)
where
represents the line losses of the system
Line losses are commonly calculated using B-coefficients where
[27]. The output power of power generation units can be represents all CO2 emissions of all units during
expressed using a quadratic function (11). interval t.
(11) represents allowed CO2 emissions during interval t.

where III. ECONOMIC DISPATCH PLANNING METHODS


represents the output vector of the power generation Two types of curves that provide users with information
units. regarding power generation costs and CO2 emissions were
represents the transposition vector of P. used. The costCO2 emissions tradeoff curve describes the
represents the loss coefficient matrix. correlation between power generation costs and CO2
represents the loss coefficient vector. emissions, revealing their upper and lower limits in particular
, , and represent loss coefficients, also known economic dispatch plans, as well as the tradeoff correlation
as B coefficients. between power generation costs and CO2 emissions [28].
In economic dispatch, the operating reserve is considered These curves also allow volume reduction indicators to be
the power supply capacity reserved for coping with determined.
emergencies (e.g., load prediction errors, power generation The other curve describes the incremental cost of CO2
unit malfunctions, and system frequency adjustment). reduction (ICCR). ICCR is defined as power generation costs
Operating reserves (i.e., the power generation volume that after reductions in CO2 emissions. ICCR reflects the ratio of
can be dispatched daily) are expressed as daily peak loads. power generation costs to CO2 emissions and thus reveals the
The operating reserve of the Taiwan Power Company tradeoff relationship between power generation costs and
generating units was set at 15% (12). carbon emissions. In other words, ICCR allows the added
costs of carbon emissions reduction in power generation to be
(12) delineated. Subsequently, performance indicators can be
determined.
where The curve most appropriate to a particular economic
represents the load demand during interval t. dispatch decision must be selected. The two curves are
represents the on-or-off status of the ith unit during detailed as follows to enable users to make optimal decisions
interval t (on = 1 and off = 0). regarding economic dispatches that consider carbon
represents the maximal power generation volume emissions.
of the ith unit. 1. CostCO2 Emissions Tradeoff Curve
The costCO2 emissions tradeoff curve delineates the
2) Unit Constraints upper and lower limits of power generation costs and CO2
The maximal and minimal output power of the thermal emissions. The tradeoff curve also describes the tradeoff
units were considered. The maximal output power was relationship between power generation costs and CO2
constrained by the turbines maximal temperature operating emissions. In this study, the power generation costs of
limits. A minimal output power was established to prevent various amounts of carbon emissions were determined; these
thermal unit extinction caused by insufficient fuel costs were then used to form a tradeoff curve.
combustion. The restriction conditions of thermal unit 2. Incremental Cost of CO2 Reduction
inequality can be expressed as (13). ICCR reveals the increase in power generation costs after
(13) CO2 emissions reduction. ICCR was calculated as in (15).
where (15)
represents the power generation volume of the ith unit.
where
represents the minimal power generation volume of
the ith unit.

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

represents the variation in power generation Step 2 Power Generation Costs, Accounting for Carbon
cost among solutions. Emissions
represents variation in CO2 emissions CO2 emission volumes can also be directly set in this step
among solutions. to obtain optimal economic dispatch strategies by
Comprehensive Problem Description determining CO2 volumes at various levels of carbon
Unit data from the Taiwan Power Company were analyzed emissions and acquire corresponding power generation costs
to incorporate carbon emissions into economic dispatch. within the obtained upper and lower limit ranges of power
Three types of dispatch planning models were adopted to generation costs and carbon emissions obtained in Step 1.
explore the power generation costs and CO2 emissions of Step 3 Drawing the Tradeoff and ICCR Curves
thermal units in various situations; these models are described The tradeoff and ICCR curves were drawn according to the
as follows. The results can serve as a reference for users. results obtained from Step 1 and 2 to determine the
1) Power Generation Cost Optimization model (16) correlation between power generation costs and CO2
Objective function emissions. This step can be modified to formulate dispatch
strategies based on user demands.
(16) Step 4 Bi-objective Interactive Best-Compromise
where Planning Method
represents the sum of power generation costs. There is no solution that fully satisfies both objective
functions (i.e., F(S) = Fideal and C(S) = Cideal) while solving
Restriction conditions: system and unit constraints.
optimization problems. An optimal compromise solution
within the upper and lower limit range of power generation
2) Carbon Emission Optimization Model (17)
costs and CO2 emissions was obtained using the bi-objective
Objective function
planning model. This solution may serve as a dispatch
(17) reference for users. Therefore, Fnonideal and Cnonideal are set as
the search boundaries, Fideal and Cideal are set as the optimal
where directions (i.e., optimal power generation costs and carbon
represents the sum of CO2 emissions. emissions) and a solution that relatively approaches both
Restriction conditions: system and unit constraints. Fideal and Cideal is used as the optimal compromise solution.
Assuming that a solution (i.e., Si ) is obtained from (18),
3) Bi-objective Optimization Model (18) the following operations are performed:

Si, Fi, and Ci are expressed in the form of interactive best-


(18) compromise solution space (Fig. 1).
Restriction conditions :
the system and unit constraints of bi-objective functions. C(S)

Problem Simulation Steps


Cnonideal
Data from 20 thermal units of the Taiwan Power Company SF
were analyzed. The bi-objective planning method proposed in
this study was used to determine economic dispatches by II I
considering carbon emissions. The interactive best-
compromise method and SA [29] were used to determine an Si
Ci
optimal compromise solution that balanced power generation III IV SC
costs and carbon emissions. The planning procedures were Cideal
divided into four steps, explained as follows.
F(S)
Step 1 Calculation of Dispatchable Intervals
By using the power generation cost and carbon emission Fideal Fi Fnonideal
optimization models, the upper and lower limits of power
Fig. 1. Interactive best-compromise solution space
generation costs and carbon emissions were obtained to
identify the dispatchable intervals of power generation costs
The four quadrants shown in Fig. 1 are obtained by using
and carbon emissions. The two single-objective planning
solutions Si, Fi, and Ci, performing relevant splitting.
models are listed as follows.
Assuming that Si is the optimal value of (18), no combined
when system and unit restriction conditions
solution (i.e., S) for F(S) and C(S) is to be found in Quadrants
were met.
I and III; only Quadrants II and IV can improve the solution
when system and unit restriction search direction. If the solution Si is not satisfactory, an
conditions were met. CO2 emissions were also constrained. objective function can be sacrificed to improve another

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

objective function and the target solution. The calculation particular user needs. Thus, the bi-objective optimal
process is as follows: compromise solutions can be improved to meet power
To improve the F(S), Fi obtained is set as the new Fnonideal, generation cost and CO2 emission conditions.
whereas Ci is set as the new Cideal, as shown in (19). The Fig. 4 shows the flow diagram for power system economic
solution space subsequently changes, as shown in Fig. 2 (this dispatch planning when accounting for carbon emissions.
is the interactive best-compromise solution space by Start
improving F(S)): Provide users with a
complete decision- Step 1
(19) making range
Calculation of dispatchable intervals by using
1.Optimization cost model of power generation
2.Optimization model of carbon emission
C(S)
Step 2 Calculate the cost of electricity for each carbon emission
1. The CO2 emission limit model is used to calculate the
cost of electricity generation for each carbon emission in
Cnonideal the dispatchable intervals
SF 2. The CO2 emission limit model is used to directly
specify the CO2 emissions for the power dispatching
planning
II I
Step 3 (Draw the Tradeoff and ICCR Curves)
According to the information of Step 1
Ci Cideal Si
and Step 2 to draw the curv
III IV
SC
Step 4
(Bi-objective planning)
1.Utilizing interactive best-compromise & SA method to determine the optimal
F(S) compromise solution to combine generation costs and CO2 emissions for users
making decision
Fideal Fnonideal 2.According to the users needs, aim at the objective function which needs to
be improved for bi-objective interactive best-compromise planning
Fi
Fig. 2. Interactive best-compromise solution space by improving F(S) No
Satisfied?

Yes
Similarly, to improve C(S), Ci obtained is set as the new
Obtain an optimal power
Cnonideal, whereas Fi is set as the new Fideal, as shown in (20). dispatching solution based on
The solution space subsequently changes, as shown in Fig. 3. the influence of carbon

Repeating Step 2 produces an improved solution.


End
(20)
Fig. 4. Flow diagram for power system economic dispatch accounting for
carbon emissions
C(S)
IV. ANALYSIS AND DISCUSSION ON THE SIMULATION
RESULTS
SF Tradeoff and ICCR Curves for Different Load
Conditions
II I Data from 20 Taiwan Power Company units were analyzed
in this study. Carbon emissions considerations were added to
Si traditional economic dispatch considerations. SA and the
Ci Cnonideal
IV
interactive best-compromise method were adopted to
III SC
Cideal simulate power generation during off-peak, semi-peak, and
peak hours. Table III shows power generation costs and
F(S) carbon emissions obtained using the three types of planning
Fideal Fnonideal models and different loads. Dispatch intervals were obtained
Fi using the power generation cost optimization model and the
Fig. 3. Interactive best-compromise solution space by improving C(S)
carbon emission optimization model. An optimal solution
was acquired in the solution space using the bi-objective
To handle the improved F(S) and C(S), users can decide optimization model proposed in this study.
the solution direction. Users only need to consider the Several power generation costs accounting for various
objective function to be improved. No weight-related amounts of carbon emissions were selected in the dispatch
problems exist and best-compromise solutions can be interval by using the carbon emission restriction model.
identified when all constraints are satisfied. Moreover, the Changes in power generation costs were observed when
data derived from the tradeoff and ICCR curves drawn in carbon emissions varied. The tradeoff and ICCR curves were
Step 3 may be used to set the items to be compromised and used to observe and program dispatches (Figs. 5-10).
the directions to be improved, yielding solutions tailored to

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

TABLE III.
POWER GENERATION COSTS AND CARBON EMISSIONS OBTAINED USING
VARIOUS MODELS AND DIFFERENT LOADS Tradeoff Curve (Semi-peak Load)
Off-peak load Semi-peak load Peak load 700

Generation cost (NT$/h)


x 10000
Power Carbon Power Carbon Power Carbon 600 The bi-objective
generati emissio generati emissio generati emissio
500
optimal
on cost n on cost n on cost n
(NT$/h) (ton/h) (NT$/h) (ton/h) (NT$/h) (ton/h) 400

300
Power
generatio 1,937,3 3,048,3 7,216,8 200
n cost 3,213 4,634 7,004
28 70 00 100
optimizati
0
on model
3430 3560 3690 3820 3950 4080 4210 4340 4470 4630
Bi- CO2 emission (ton/h)
objective 3,015,6 4,130,9 7,599,8
2,593 3,909 6,842
optimizati 59 73 85
on model Fig. 7. Tradeoff curve and bi-objective optimal compromise solution (semi-
peak load)
Carbon
emission 4,637,2 5,850,5 8,081,5
2,192 3,428 6,694
optimizati 05 50 56
ICCR Curve (Semi-peak Load)
on model
4500
4000
The bi-objective

ICCR(NT$/ton)
3500
Tradeoff Curve (Off-peak Load) 3000 optimal
2500
500
Generation cost (NT$/h)
x 10000

2000
450
400 The bi-objective 1500
1000
350 optimal 500
300
250 compromise 0
200 solutions 3430 3560 3690 3820 3950 4080 4210 4340 4470 4630
150
CO2 emission (ton/h)
100
50
0 Fig. 8. ICCR curve and bi-objective optimal compromise solution (semi-
2190 2300 2410 2520 2630 2740 2850 2960 3130 3210 peak load)
CO2 emission (ton/h)
Figs. 7 and 8 respectively show the Tradeoff and ICCR
Fig. 5. Tradeoff curve and bi-objective optimal compromise solution (Off- curves under semi-peak load. In Fig. 7, the generation cost
peak load) ranges from 3,050,000 to 5,850,000 NT$/h and the amount of
CO2 emissions from 3,430 to 4,630 ton/h. In Fig. 8, the ICCR
ICCR Curve (Off-peak Load) ranges from 938 to 4,077 NT$/ton and the amount of CO2
emissions from 3,430 to 4,470 ton/h. The optimal
5000
4500 The bi-objective compromise solution for ICCR is 2,900 NT$/ton.
4000 optimal
ICCR(NT$/ton)

3500
3000 Tradeoff Curve (Peak Load)
2500
Generation cost (NT$/h)

2000 820
x 10000

1500 800 The bi-objective


1000
500
780 optimal
0 760
2190 2300 2410 2520 2630 2740 2850 2960 3130 3210 740

CO2 emission (ton/h) 720


700

Fig. 6. ICCR curve and bi-objective optimal compromise solution (Off- 680

peak load) 660


6690 6720 6750 6780 6810 6840 6870 6900 6930 7000

Figs. 5 and 6 respectively show the Tradeoff and ICCR CO2 emission (ton/h)
curve under off-peak load. In Fig. 5, the generation cost
ranges from 1,940,000 to 4,640,000 NT$/h. In Fig. 6, the Fig. 9. Tradeoff curve and bi-objective optimal compromise solution (peak
ICCR ranges from 2,750 to 4,247 NT$/ton and the amount of load)
CO2 emissions from 2,190 to 3,130 ton/h. The optimal
compromise solution for ICCR is 3,300 NT$/ton.

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

TABLE IV.
ICCR Curve (Peak Load) SRS OF THREE TYPES OF DISPATCH PLANNING MODELS UNDER DIFFERENT
LOAD CONDITIONS
4500 Power
4000
The bi-objective generation
Carbon
optimal Bi-objective emission
ICCR(NT$/ton)

3500 cost
optimization model optimization
3000 optimization
model
2500 model
2000 Load SRF SRC SRF SRC SRF SRC
1500 condition (%) (%) (%) (%) (%) (%)
1000 Off-peak 60.03% 60.80% 100 % 0% 0% 100 %
500 Semi-peak 60.46% 60.06% 100 % 0% 0% 100 %
0 Peak 55.7% 52.16% 100 % 0% 0% 100 %
6690 6720 6750 6780 6810 6840 6870 6900 6930 7000
CO2 emission (ton/h)

Fig. 10. ICCR curve and bi-objective optimal compromise solution (peak
load)

Figs. 9 and 10 respectively depict the Tradeoff and ICCR


curve under peak load. In Fig. 9, the generation cost ranges
from 7,220,000 to 8,080,000 NT$/h and the amount of CO2
emissions from 6,690 to 7,000 ton/h. In Fig. 10, the ICCR
ranges from 2,000 to 4,000 NT$/ton and the amount of CO2
emissions from 6,690 to 6,930 ton/h. The optimal
compromise solution for ICCR is 2,900 NT$/ton.
Therefore, the proposed method determines economic
dispatch in power systems while accounting for the amount
of carbon emissions. This enables the derivation of a
definitive bi-objective optimal compromise solution to the
generation cost and the amount of CO2 emissions to include
in economic dispatch plans and the adjustment of the solution
in line with tradeoff and ICCR curves and individual needs.
Thus, if the Taipower decision maker finds the solution
unsatisfactory, the generation cost or the amount of CO2
emissions can be adjusted to improve the dispatch results.
Daily Load Test Data
Fig. 11. Daily load curve of the four seasons in 2011 [30]
The interactive best-compromise method proposed in this
study was integrated with SA to obtain an optimal solution Fig. 11 shows the daily load curve in 2011 by season.
that accounted for both power generation costs and carbon Because the load curve fluctuated substantially during
emissions. Furthermore, the user satisfaction rate (SR) for summer, the hourly load demand varied considerably. This
power generation costs and CO2 emissions in the bi-objective allowed power generation costs and CO2 emissions to be
interactive best-compromise model was compared with those more clearly compared. The daily load curve in summer was
of the power generation cost-optimization and carbon used as a benchmark for the 1-day daily load demand table.
emission optimization models. A high SR indicated a near- Data from 20 Taiwan Power Company units were used to
optimal solution within the limits of the range. The formula calculate power generation costs and CO2 emissions under
for SR is expressed as (21). the 1-day load. To consider both economic and
environmental factors, the interactive best-compromise
method was combined with SA to acquire the power
(21) generation costs and CO2 emissions at various hours.
Subsequently, the SRs of the power generation cost-
where optimization and carbon emission optimization models were
compared to infer the degree to which power generation costs
and CO2 emissions were improved.
Table V shows the daily load demands of 20 Taiwan Power
Table IV presents the SRs of three types of dispatch Company units during 1 day. According to the summer daily
planning models under the three load conditions. load curve, shown in Fig. 11, the off-peak occurred in the 7th
hour and the peak occurred in the 14th hour. Three types of
target planning models were used to satisfy load demands in

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

each hour. In addition, power generation costs and CO 2 generation cost and CO2 emission were NT$171,618,651 and
emissions were obtained. 132,597 tons, respectively.
Power generation cost and CO2 emission daily load
TABLE V.
demands for 1-day were acquired using the bi-objective best-
ONE-DAY LOAD DEMAND
HR DEMAND HR DEMAND compromise model. An optimal solution in compliance with
1 6854 13 7827 preset restrictions was obtained (Table VII). The total power
2 6648 14 8318 generation cost and CO2 emissions for 1-day were
3 6443 15 8164 NT$148,565,178 and 138,793 tons, respectively (Table VIII).
4 6238 16 8010
5 6032 17 7856
The results shown in Tables Vi to VIII indicate that the
6 5827 18 7702 models yielded varying power generation costs. The carbon
7 5622 19 7760 emission optimization model yielded the highest power
8 6088 20 7817 generation costs, followed by the bi-objective best-
9 6554 21 7636
10 7020 22 7456
compromise model, and the power generation cost-
11 7486 23 7275 optimization model. The models also yielded varying
12 7952 24 7095 amounts of CO2 emissions. The power generation cost-
optimization model yielded the greatest amount of CO2
First, power generation cost and CO2 emission daily load emissions, followed by the bi-objective best-compromise
demands for 1 day were obtained using the power generation model, and the carbon emission optimization model.
optimization model (Table VI). As Table VI shows, the Therefore, if the generation cost is minimized, then the
minimums and maximum power generation cost and CO2 amount of CO2 emissions will be maximized; if the amount
emissions occurred in the 7th and 14th hours, respectively. of CO2 emissions is minimized, then the generation cost will
This result was consistent with the summer load curve in Fig. be maximized. The proposed bi-objective best-compromise
11. Based on the result of each hour, the 1-day total power model strikes the optimal balance between them.
generation cost and CO2 emission were NT$133,950,508 and
TABLE VII.
147,106 tons, respectively. POWER GENERATION COSTS AND CARBON EMISSIONS IN THE CARBON
EMISSION OPTIMIZATION MODEL
TABLE VI.
POWER GENERATION COST AND CARBON EMISSIONS IN THE POWER Power Power
Carbon Carbon
GENERATION COST OPTIMIZATION MODEL HR generation cost HR generation cost
emission (kg) emissions (kg)
(NT$) (NT$)
Power Power
Carbon Carbon emission
HR generation cost HR generation cost
emission (kg) (kg) 1 5312397.4 6522235.1 13 6144237.7 7727638.8
(NT$) (NT$)
2 5065591.7 6512583.3 14 6543241.6 7978674.1
1 6023726.3 5062905.0 13 6535875.6 6511947.3
3 4785856.6 6855882.7 15 6416620.6 7889461.8
2 5862134.3 4917586.1 14 6876490.2 7024784.8
4 4630060.3 6768107.6 16 6291616.3 7810838.8
3 5702478.8 4616487.1 15 6768653.4 6863303.0
5 4479083.9 6588048.8 17 6167584.2 7740818.7
4 5541033.4 4476845.7 16 6661440.5 6702609.6
6 4332614.3 6374933.8 18 6044871.9 7666865.4
5 5378905.6 4347572.1 17 6556379.0 6542124.7
7 4187172.9 6157727.9 19 6027154.9 7622800.4
6 5328211.5 3962988.2 18 6447121.3 6382163.4
8 4519367.2 6646661.4 20 6073261.3 7651880.6
7 5193895.2 3757363.8 19 6488444.7 6442342.5
9 4942678.4 6944235.7 21 6138452.6 6892841.1
8 5423185.8 4374939.4 20 6531522.8 6341516.5
10 5379505.2 7190342.6 22 5729735.7 7418155.2
9 5790088.2 4692280.7 21 6406902.4 6152979.7
11 5878452.1 7573105.0 23 5756436.4 6621770.3
10 6152684.5 5200157.6 22 6254878.5 5932134.8
12 6244892.2 7784461.4 24 5506366.3 6678580.4
11 6443108.2 5932565.0 23 6146264.6 5678530.4
Total carbon emission: 132,597 (ton)
12 6592614.4 6609433.1 24 6000074.5 5424947.0
Total power generation cost: 171,618,651 (NT$)
Total carbon emission: 147,106 (ton)

Total power generation cost: 133,950,508 (NT$) To comprehensively differentiate these three models, the
SRs of power generation costs and CO2 emissions of the bi-
The load demands for power generation cost and CO2 objective interactive best-compromise model were compared
emissions for 1 day were determined using the carbon with those of the power generation cost optimization model
emission optimization model. The 1-day total power and the carbon emission optimization model. A high SR

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

indicated that the solution was a near-optimal solution within system economic dispatch. An optimized bi-objective power
the range. Table IX shows SRs for the three objective dispatch planning method was proposed.
planning models. Furthermore, the method was used to estimate economic
According to Table IX, compared with the carbon emission dispatch (the off-peak, semi-peak, peak load, and daily load
optimization model, the bi-objective best-compromise model demand with the hourly generation cost and CO2 emissions)
yielded a decrease in the CO2 emission SR from 100% to in Taipowers thermal generators. As the simulation results
57% and an increase in power generation cost SR from 0% to shown, the interactive best-compromise method integrated
61%. The results indicated that in the bi-objective best- with SA was used to solve bi-objective planning problems,
compromise model, power generation units satisfied 1-day thereby reducing generation costs and CO2 emissions.
load demands as well as considered costs and environment The study results also yielded the following findings.
protection. To further address the concern of CO2 emission, 1. This study proposed various objective planning models
power generation cost compromise items can be incorporated (i.e., a power generation cost-optimization model, a
into the combination of the interactive best-compromise carbon emission optimization model, and a bi-objective
method and SA to reduce CO2 emissions and enhance the SR best-compromise model), allowing user demands in
of carbon emissions. economic dispatch planning to be flexibly satisfied.
2. The influence of variations in CO2 emissions on power
TABLE VIII. generation costs in various load conditions were observed
POWER GENERATION COSTS AND CARBON EMISSIONS IN THE BI-OBJECTIVE
by drawing tradeoff and ICCR curves. Economic dispatch
BEST-COMPROMISE MODEL
strategies can be proposed on the basis of such curves.
Power Power 3. One-day load demands and unit commitments were
Carbon Carbon
HR generation cost HR generation cost
emission (kg) emissions (kg) determined. These results possess high reference value
(NT$) (NT$)
regarding CO2 emission reduction and power generation
1 5601485.9 5582964.0 13 6307307.9 6987584.5 cost control. The method proposed in this study can be
2 5303743.9 5804283.1 14 6700682.0 7443375.3
used to effectively reduce power generation costs.
4. To allow user demands to be satisfied, the objective
3 5151035.0 5465184.1 15 6584364.1 7302322.0 planning method proposed in this study integrated SA and
4 4923578.9 5520255.6 16 6492516.9 7092124.9
the interactive best-compromise method. The method
does not require weighting of any kind. This method can
5 4928349.1 4909912.9 17 6299393.1 7182478.0 be strategically used to consider carbon emissions in
6 4704977.1 4884203.6 18 6200068.0 6862086.8
economic dispatch planning.
In addition to considering the influence of CO2 emissions,
7 4642469.9 4492849.7 19 6256522.1 6876825.9 the optimized economic dispatch method proposed in this
8 4838961.0 5209458.9 20 6316644.0 6879159.3
study introduced various objective planning models: a power
generation cost-optimization model, a carbon emission
9 5321071.0 5559027.2 21 6217913.9 6505703.7 optimization model, and a bi-objective best-compromise
10 5743964.9 5847869.5 22 6040308.1 6461298.4
model. Tradeoff and ICCR curves were used to observe the
influence of power generation costs and CO2 emissions on
11 6096328.0 6418827.6 23 5906178.1 6157287.0 economic dispatch planning. The system allows power
generation costs and CO2 emissions to be determined by
12 6412428.1 7068829.4 24 5802590.0 6051266.9 users, allowing economic and environmental factors to be
Total carbon emission: 138,793 (ton)
considered simultaneously.

Total power generation cost: 148,565,178 (NT$)


REFERENCES
[1] Taiwan's Environmental Protection Administration. (2014.02.05).
TABLE IX UNFCCC information network. [Online]. Available:
SRS OF THE THREE OBJECTIVE PLANNING MODELS http://unfccc.epa.gov.tw/unfccc/chinese/index.html
[2] UNFCCC. (2012.12.08). Doha Amendment to the Kyoto Protocol.
SRF(%) SRC(%) [Online].Available: http://unfccc.int/kyoto_protocol/items/2830.php
[3] Taiwan's Bureau of Energy, Ministry of Economic Affairs. Set carbon
Power generation cost optimization model 100% 0% reduction targets. [Online]. Available:
Bi-objective best-compromise model 61% 57% http://www.moea.gov.tw/AD/Ad04/content/ContentDetail.aspx?menu_i
d=4621
Carbon emission optimization model 0% 100% [4] Y. W. Jeong, J. B. Park, S. H. Jang, and K. Y. Lee, "A New Quantum-
Inspired Binary PSO Application to Unit Commitment Problems for
Power Systems," Power Systems, IEEE Transactions on, vol. 25, pp.
1486-1495, Aug. 2010.
V. CONCLUSIONS [5] A. Y. Saber, T. Senjyu, T. Miyagi, N. Urasaki and T. Funabashi, "Unit
commitment by heuristics and absolutely stochastic simulated
SA and the interactive best-compromise method were annealing," IET Generation, Transmission & Distribution, vol. 1, pp.
integrated in this study to consider carbon emissions in power 234-243, 2007.

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2017.2771338, IEEE
Transactions on Industry Applications

[6] C. C. Kuo, "A Novel Coding Scheme for Practical Economic Dispatch by [26] Lixueyu, "Energy and Climate Registry," Climate Change Program
Modified Particle Swarm Approach," Power Systems, IEEE Innovation Center For Energy And Transportation.
Transactions on, vol. 23, pp. 1825-1835, Aug. 2008. [27] H. Saadat, Power System Analysis, NewYork: McGraw-Hill Primis
[7] M.T. Kuo, S. D. Lu and M. C. Tsou, "Economic dispatch planning based Custom Publishing, 2002.
on considerations of wind power generation and pumped storage [28] M. Muslu, "Economic dispatch with environmental considerations:
hydroelectric plants for isolated power systems," IEEE/IAS 51st tradeoff curves and emission reduction rates," Electric Power Systems
Industrial & Commercial Power Systems Technical Conference Research, vol. 71, pp. 153-158, 2004.
(I&CPS), pp. 1-10, 2015. [29] C. C. Kuo. A Multi-Objective Approach Based Decision software for
[8] W. Ge, "Ramp Rate Constrained Unit Commitment by Improved Priority Distribution Automation. Ph.D. dissertation, department of electrical
List and Enhanced Particle Swarm Optimization," International engineering, National Taiwan University of Science and Technology,
Conference on Computational Intelligence and Software Engineering, 1998.
pp1-8, 2010. [30] Gordoncheng's Blog. [Online] Available:
[9] J. J. Hargreaves and B. F. Hobbs, "Commitment and Dispatch With http://gordoncheng.wordpress.com/2010/09/06/
Uncertain Wind Generation by Dynamic Programming," Sustainable
Energy, IEEE Transactions on, vol. 3, pp. 724-734, 2012.
[10] W. J. Hobbs, G. Hermon, S. Warner and G. B. Shelbe, "An enhanced
dynamic programming approach for unit commitment," Power Systems,
IEEE Transactions on, vol. 3, pp. 1201-1205, Aug. 1988.
[11] M. A. Bragin, P. B. Luh, J. H. Yan and G. A. Stern, "Novel exploitation
of convex hull invariance for solving unit commitment by using
surrogate Lagrangian relaxation and branch-and-cut," IEEE Power &
Energy Society General Meeting, pp. 1-5, 2015.
[12] M. A. Bragin, P. B. Luh, J. H. Yan and G. A. Stern, " Surrogate
Lagrangian relaxation and branch-and-cut for unit commitment with
combined cycle units," IEEE PES General Meeting Conference &
Exposition, pp. 1-5, 2014.
[13] P. Wang, Y. Wang and Qing Xia," Fast bounding technique for branch-
and-cut algorithm based monthly SCUC," IEEE Power and Energy
Society General Meeting, pp. 1-8, 2012.
[14] G. K. Purushothama and L. Jenkins, "Simulated annealing with local
search-a hybrid algorithm for unit commitment," Power Systems, IEEE
Transactions on, vol. 18, pp. 273-278, 2003.
[15] K. Venkatesan and C. C. A. Rajan, "A simulated annealing method for
solving multi-area unit commitment problem," in Process Automation,
Control and Computing (PACC), 2011 International Conference on, pp.
1-7.
[16] D. N. Simopoulos, S. D. Kavatza, and C. D. Vournas, "Reliability
Constrained Unit Commitment Using Simulated Annealing," Power
Systems, IEEE Transactions on, vol. 21, pp. 1699-1706, 2006.
[17] D. N. Simopoulos, S. D. Kavatza, and C. D. Vournas, "Unit
Commitment by an Enhanced Simulated Annealing Algorithm," in
Power Systems Conference and Exposition, IEEE PES, pp. 193- 201,
2006.
[18] A. Y. Saber, T. Senjyu, T. Miyagi, N. Urasaki, and T. Funabashi, "Fuzzy
Unit Commitment Scheduling Using Absolutely Stochastic Simulated
Annealing," Power Systems, IEEE Transactions on, vol. 21, 2006.
[19] T. Gjengedal, "Emission Constrained Unit-Commitment (ECUC),"
Energy Conversion, IEEE Transactions on, vol. 11, pp. 132-138, 1996.
[20] D. N. Simopoulos, Y. S. Giannakopoulos, S. D. Kavatza, and C. D.
Vournas, "Effect of Emission Constraints on Short-Term Unit
Commitment," in Electrotechnical Conference, 2006. MELECON 2006.
IEEE Mediterranean, pp. 973-977.
[21] O. Taegon, C. Jaeseok, C. Junmin, B. Ungki, and K. Y. Lee, "Generator
maintenance scheduling considering minimization of CO2 emissions,"
in Power and Energy Society General Meeting, 2011 IEEE, pp. 1-6.
[22] R. Jiang, "A Mathematically Generated Criteria Weight Approach for
Multi-Criterion Decision," in Management and Service Science (MASS),
2010 International Conference on, 2010, pp. 1-4.
[23] P. H. Chen and C. C. Kuo, "Bi-Objective Hydroelectric Optimal
Dispatch under Electricity Deregulated Environment," in Transmission
and Distribution Conference and Exhibition: Asia and Pacific, 2005
IEEE/PES, 2005, pp. 1-5.
[24] J. P. S. Catalao, S. J. P. S. Mariano, V. M. F. Mendes, and L. A. F. M.
Ferreira, "Profit-Based Unit Commitment with Emission Limitations: A
Multiobjective Approach," in Power Tech, 2007 IEEE Lausanne, 2007,
pp. 1417-1422.
[25] D. N. Simopoulos, Y. S. Giannakopoulos, S. D. Kavatza, and C. D.
Vournas, "Effect of Emission Constraints on Short-Term Unit
Commitment," in Electrotechnical Conference, 2006. MELECON 2006.
IEEE Mediterranean, 2006, pp. 973-977.

0093-9994 (c) 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Das könnte Ihnen auch gefallen