Professional Documents
Culture Documents
Carbon capture power plants: Decoupled emission and generation outputs MARK
for economic dispatch
Abdirahman Mohamed Abdilahi , Mohd Wazir Mustafa
Universiti Teknologi Malaysia, 81310 Johor Bahru, Johor, Malaysia
A R T I C L E I N F O A B S T R A C T
Keywords: Integrating exibly-operated carbon capture and storage (CCS) into the existing power plants has operational
Carbon capture and storage benets for the future low carbon power systems. This paper proposes an improved formulation for exible
Economic dispatch operation of carbon capture power plants (CCPPs) within the conventional economic dispatch (ED) problem. The
Low carbon power systems main contribution of this work is the simplication and the practicality of the variables used for the exible
Metaheuristic optimization algorithms
operation control of the facility. The optimal ED problem of thermal power generation portfolio with CCPPs
Power system operations
Carbon pricing
within the mix are computed using a chaos-enhanced Cuckoo Search optimization algorithm. To test the
proposed formulations, an IEEE 30 bus test system was used. The impact of varying carbon prices on the system
dispatch was investigated. The results reveal the potentiality of decoupling the generation and emission outputs
of the thermal power plants.
Abbreviations: CCPP, carbon capture power plants; CCS, carbon capture and storage; CP, carbon price; CS, cuckoo search; CT, carbon Tax; ED, Economic dispatch; EU, European Union;
IEA, International Energy Agency; NP, non-capture plant
Corresponding author.
E-mail addresses: abdirahmaan11@gmail.com, abdirahman@ieee.org (A.M. Abdilahi).
http://dx.doi.org/10.1016/j.ijggc.2017.05.001
Received 11 January 2017; Received in revised form 19 April 2017; Accepted 7 May 2017
1750-5836/ 2017 Elsevier Ltd. All rights reserved.
A.M. Abdilahi, M.W. Mustafa International Journal of Greenhouse Gas Control 63 (2017) 1219
formulations (e.g. the work in Hetzer et al., 2008). decision variables and one for the coordination between net emission,
The consideration of the operational exibility of CCPPs within net power output and the gross power output. For large-scale systems in
dispatch formulation has been, so far, addressed in few literatures (Ji particular, this approach presents a considerable complexity when
et al., 2013; Chen et al., 2010; Chen et al., 2012; Lu et al., 2013). First, using advanced metaheuristic optimization techniques.
an innovative work from Chen et al. (2010) and their follow-up work Capitalising the existing formulations and aiming to address the
(Chen et al., 2012) have provided initial grounds for formulations of the existing weaknesses, the main objective of this work is to develop a
exible operation of the CCPPs. However, their works incorporate the modied model formulation of the exible operation of CCPP units
generation eciency thread within the formulations. However, suitable for the conventional static ED problem. Based on the proposed
existing ED formulations are based on output power thread. Second, CCPP model, the conventional ED is reformulated taking into account
the approach presented in (Lu et al., 2013), which was used in other the generation mix diversity. Then the impact of CCPP within the
studies (Lou et al., 2015; Jiaming et al., 2015), does not formulate generation mix on dierent aspects of the power system such as
emission as an independent controllable variable. Instead, the pre- generation mix, generation cost, CO2 cost, emission intensity and others
sented formulation is a multi-decision-oriented procedure that is aimed are studied.
to determine the captured emission and net power output of the CCPP. As far as our current literature is concerned, this paper presents the
Thirdly, while the work of Ji et al. (2013) formulates emission as a rst attempt to integrate the CCPP within a set of thermal generators for
dispatchable resource, the work has two inherent mathematical com- the static ED problem. This is important to provide understanding of the
plexity. First, the four auxiliary decision variables, employed to help the optimal operating characteristic of the facility while using snapshots of
optimization routine to select an operating point, signicantly increase plant operation. The presented model may be useful in incorporating
the problem complexity by multiplying four extra decision variables for within the future releases of the professional power system operations
each considered CCPP unit within the system. From the mathematical software available to system planners and operators. For example,
optimization point of view, this is called the curse of dimensionality authors in van der Wijk et al. (2014) unavoidably developed a separate
problem i.e. the exponential rise in the time and space required to Excel-based module for CCS operation and then integrated it with
compute an approximate solution to a problem as the dimension (i.e. PLEXOS, a professional power system operations tool.
the number of control variables) increases. Second, the work also The work presented is organized as follows. The existing model
developed two additional equality constraints; one for the auxiliary formulations are rstly improved within the Section 1. It is then
13
A.M. Abdilahi, M.W. Mustafa International Journal of Greenhouse Gas Control 63 (2017) 1219
implemented with the ED formulations at Section 2. Section 3 develops The ability of the stripper/compressor units to treat rich solvent is
a metaheuristic-based optimization methodology to solve the estab- assumed to be larger than the ability of the scrubber (absorber) by 20%.
lished problem. It is then followed by the results and discussion of the This allows higher stripping rate because of the potential availability of
paper, presented at Section 4. The paper is concluded at Section 5. stored rich solvent. Therefore, the Cmax is set to 120%. According to
Cohen et al. (2011), this equipment sizing allows a maximum of four
2. Modelling the exible operation of CCPP hours of full storage mode, assuming a daily cycling operation of the
storage system.
Flexible operation of CCS units is the ability of these plants being With the above formulations, one can draw an operating region for
built such that part-load operation and partial CO2 capture capabilities the CCPP by considering the two output variables that are of interest for
are enabled. Flexibly operated CCS have the ability to signicantly the system operator i.e. PN and EN, as shown in Fig. 1.
reduce emissions while providing system support services within the From this operating region, the two inputs (i.e.PGand C) aect the
power grid. controllability of the net emission and net power output. For instance,
when the operating penalty of the CCS is at its minimum, the plant can
2.1. Mathematical formulations achieve its highest variation in both net emission and net power output.
This phenomena is indicated by line 1 as shown at Fig. 1. However,
There is an inherent energy penalty with the installation of CCS when the operating penalty of the CCS is high, any change to the gross
retrot into the existing fossil-red power plants. Taking this energy power output has minimum eect on net emission variation and the
penalty into account, the net power output (PN ) of the facility can be range of the net power output decreases considerably. This is indicated
described as: by line 2. In this way, the plants outputs are controlled as shown in the
block diagram in Fig. 2.
PN = PG PCP (1)
The capacity penalty can be broadly divided into a basic penalty 2.3. Comparison of proposed model with existing models
(PBP ) and an operating penalty(POP ) as shown below:
Table 2 summarizes the dierences between the proposed model
PCP = PBP + POP (2)
and those presented in the existing literature (Ji et al., 2013; Chen et al.,
The basic power penalty is incurred due to changing the plant 2012; Lu et al., 2013).
structure for CCS retrot such as diversion of heating systems. This is
normally a xed value that contributes a small percentage to the overall 3. Mathematical formulation
capacity loss of the basic plant. The second component is the power
used to operate the various modules of the CCS. Typically, it is To demonstrate the eect of the decoupled emission and generation
dependent to the amount of operation of the CCS. This power can be output within system dispatch, a system-wide optimization model is
associated with the amount of CO2 treated (EC ) as shown below: constructed. The model makes operating decisions for the plant in such
a way as to minimize the overall operating costs.
POP = wCC EC (3)
The mathematical model of the new ED problem is presented as
On the other hand, the total amount of emission generated from the follows:
power cycle (EG ) is presented as:
NNP
NCP
14
A.M. Abdilahi, M.W. Mustafa International Journal of Greenhouse Gas Control 63 (2017) 1219
Fig. 2. Block diagram showing the two control inputs and the resulting outputs.
4. Optimization methodology
NNP NCP
subject to Pd + PNk PD PL = 0 In this paper, the Cuckoo Search (CS) algorithm, an ecient global
d =1 k =1 (12) metaheuristic optimization algorithm developed in (Yang and Deb,
2009), is adopted as the solver. The algorithm couples Gaussian
PGmin
k
PGk PGmax
k (13) distribution-based random walk for local exploitation with Levy ight
for global search. With this balanced characteristic, the algorithm is
Pd min Pd Pd max (14) suitable for solving multimodal functions a characteristic present in
non-convex ED formulations. The complete and detailed owchart of
Eqs. (7) is the overall cost function of the optimization problem. the algorithm is shown in Fig. 3.
Eqs. (8) and (9) represent the fuel and emission cost functions of the non-capture
power plants, respectively.
power
Eqs. (10) and (11) represent the fuel and emission cost functions of the carbon capture
plants, respectively.
Eq.
4.1. Implementation of the optimization algorithm
(12) represents the power balance (equality) constraint of the problem.
Eqs. (13) and (14) represent the power output limits for both capture and non-capture
power plants. 4.1.1. Initialization step
In this modication, the aim is to explore potential techniques for
Furthermore, heat-run tests of thermal generators indicate that ED initializing more intelligently. Initial locations of the search agents for
problems are more accurately modelled as non-convex, non-dierenti- population-based optimization techniques aect signicantly the abil-
able and multimodal problem. This is due to the ripple eect produced ity of achieving optimal solution. In this study, the Latin Hypercube
by valve-point loading as the generators output is increased (Walters Sampling (LHS) technique is proposed as a potential replacement for a
and Sheble, 1993). To represent this eect, the WalterSheble model Gaussian distribution-based random variable. Using this mechanism, a
(Walters and Sheble, 1993) is used, whereby a sinusoidal function is population of nests is initialized randomly, so that each decision
included in the thermal fuel cost function. To model the carbon price variable satises its operating limits as shown in (15):
(CP ) eects in the ED formulations, the model used in Fang et al. (2012)
xi, j = ximin + (ximax ximin ) lhsdesign (ND , Nx ) (15)
has been adopted as presented in Eq. (9).
Table 2
Comparison between the proposed model and the existing models.
REFs Features within the Existing Models Features within the Proposed Model
Chen et al.
(2012)
Formulates the problem using eciency thread. Formulates the problem using the power generation thread, which is more suitable for
dispatch computations.
Lu et al. (2013) Does not formulate emission as an output variable that
inuences the dispatch.
Formulates
dispatch.
both emission and generation outputs as variables that directly inuence the
i ND andj = 1,2,...Nx providing better randomness characteristics than the Gaussian distribu-
min max tion based variables, particularly in the metaheuristic optimization
where x and x are the upper and lower limits of each decision
techniques (Fister et al., 2015). In another study, we found out that the
variable within the considered set of online thermal generators. The
chaotic maps introduce an additional explorative power in the swarm
function lhsdesign is a MATLAB command used to statistically generate
intelligence-based search process that prevents the search agents from
random sample positions for each nest. Our preliminary experiments
trapping in local optima (Mustafa et al., 2016). For that reason, this
with this command revealed that the number of population required to
study adopts a chaotic sequence to improve the performance of the
achieve certain quality are smaller with the use of this command in
original CS algorithm. Table 3 shows the original and modication
replacement of classical rand operator.
proposed for the updating equations implemented at the discovery
stage of the CS.
4.1.2. Adjustment of Levy ight steps The current authors carried out an in-depth experimental investiga-
Within CS, the Levy ight is employed in order to carry out a global tion for eight dierent chaotic sequences at several dierent locations
exploration search by performing occasional long-distance jumps. within the CS. The chaotic sequences are obtained from Gandomi et al.
These sudden jumps increase the search eciency of the CS signi- (2013). Based on statistical robustness of each resultant variant as a
cantly. The updating formula for the Levy ight process is given as result of the chaotic sequence used, the Chebyshev chaotic sequence
follows: proofed to be highly potential for integrating CS. The Chebyshev
xinew = xi + randn Scale Levy ( ) (16) chaotic sequence can be described mathematically as:
where Scale is the length scale used to control the length of the steps y (t + 1) = cos(t cos1 y (t )) (18)
taken by the Levy ight and randnis a normally distributed stochastic
number. In order to transform the original algorithm to problem where y (t ) is the value of the chaotic map at each iteration t. In order to
specic type which makes it more ecient and less volatile in nding draw this map, an initial value of y0 = 0.1 is used and the resulting
the global optimal solution (Yang and Deb, 2014), this work proposes waveform is as shown in Fig. 4.
the integration of the decision variables limits as part of the length The detailed methodology and the appropriate simulations that
scale of the Levy ight equation. This is expected to increase the follow are explained in Mustafa et al. (2016). This paper, without loss of
awareness of the algorithm about the search-space characteristics. The generality, focuses on the low carbon economic dispatch formulations
length scale is dened as follows: with advanced metaheuristic optimization technique.
(ximax ximin )
Scale = Table 3
100 (17)
Illustration of modications implemented.
Equations Involved
4.1.3. Chaotic maps at the discovery stage
To perform local exploitation, the CS algorithm metaphorically Original
mimics the concept of discovery of the Cuckoo egg within the host nest. xinew = xi + rand H (Pa ) (xm xn )
Mathematically, this concept brings in a randomization feature and a Modied
local search. In the standard CS, the randomization is implemented xinew = xi + yt H (Pa ) (xm xn )
using a Gaussian distribution-controlled random walk. On the other
hand, recent research indicates the potentiality of chaotic maps in
16
A.M. Abdilahi, M.W. Mustafa International Journal of Greenhouse Gas Control 63 (2017) 1219
Table 5
Emission factors thermal Generating units.
5. Numerical illustrations
Table 4
Cost and emission coecients of the six unit system.
a b c (x103) d e f g h (x106)
Note: (1) The coecients of ai, bi, and ci are in $/MW2, $/MW, and $.
17
A.M. Abdilahi, M.W. Mustafa International Journal of Greenhouse Gas Control 63 (2017) 1219
Table 6
Optimal generation settings (in MW) at dierent CP (in $/tCO2).
Fig. 5. The impact of a varying carbon price on the three dierent categories of system costs: fuel, emission and overall operating cost.
Fig. 6. The optimal settings of gross generation (MW) and the normalized rate of the stripper/compression units for dierent carbon prices.
an optimal dispatch point of view. Consequently, the G3 becomes the plants output by 22% with an eective net power output of only
slack unit that is dispatched in order to balance the load demand. 463.67 MW (shown in Fig. 6). With this power output setting, the plant
With a CP range between 6.5 $/tCO2 and 7.5 $/tCO2, a high has a fuel cost of $11,393.96 $/h. As a result of the penalty power of the
operation of the CCS is economically justiable, leading to a decreased CCS, the plant incurs an inclusive fuel cost of up to 2,577.14 $/h. On
net output of the CCPP by a 17% (as shown in Fig. 6). At this point, the the other hand, despite the expensive operation of the CCS, the plant
penalty power consumed by the CCS is 123 MW leading to a penalty has a capturing capability of 443 tCO2/h.
cost of 2,340.12 $/h based on the resulting cost per unit of the fuel cost Beyond this price level, any CP increase leads to higher operating
of the plant. This penalty cost becomes justiable than an emission cost expenses from system-wide perspectives as shown in Fig. 5. The carbon
caused by the vented CO2 of the facility (as shown in Fig. 6). As a result price does not aect the operation of the CCPP unit and only forces the
of this, the emission cost of the whole system is lowered down (as heavy-polluting non-capture units to operate at part-load operation,
shown in Fig. 5). similar to the observation presented in Fang et al. (2012).
With a CP range higher than 8 $/tCO2, the optimal setting of the It is worth-noting that the optimization involved in this work
CCPP unit is to fully operate its CCS unit. Consequently the highest focuses the perspectives of the system operator. However, other studies
penalty power (135.53 MW) is incurred leading to a degradation of the might have other perspectives such as maximizing the prots of
18
A.M. Abdilahi, M.W. Mustafa International Journal of Greenhouse Gas Control 63 (2017) 1219
operating the CCPP as an individual generator. combustion CO2 capture in response to volatile electricity prices. Energy Procedia 4,
26042611.
Fig. 6 shows that despite the changing CP value, the control of the Cohen, S.M., Rochelle, G.T., Webber, M.E., 2012. Optimizing post-combustion CO2
net emission variation only aects the net power output supplied to the capture in response to volatile electricity prices. Int. J. Greenh. Gas Control 8,
grid. However, it does not aect the gross power output of the facility 180195.
Fan, J.-Y., Zhang, L., 1998. Real-time economic dispatch with line ow and emission
throughout the whole CP spectrum (as shown in Fig. 6). The connection constraints using quadratic programming. IEEE Trans. Power Syst. 13, 320325.
between the net power output and the net emission is the amount of Fang, Y., Zhao Yang, D., Ke, M., Zhao, X., Iu, H.H., Kit Po, W., 2012. Quantum-inspired
penalty power incurred as a result of operation adjustment of the CCS. particle swarm optimization for power system operations considering wind power
uncertainty and carbon tax in Australia. IEEE Trans. Ind. Inf. 8, 880888.
Fister, I., Perc, M., Kamal, S.M., 2015. A review of chaos-based rey algorithms:
6. Conclusion perspectives and research challenges. Appl. Math. Comput. 252, 155165.
Global CCS Institute, 2015. The Costs of CCS and Other Low-carbon Technologies 2
Global CCS Institute (July).
Optimal dispatch of emission and generation output has been a
Hetzer, J., Yu, D.C., Bhattarai, K., 2008. An economic dispatch model incorporating wind
classical problem for utilities. Earlier formulations solved the problem power. IEEE Trans. Energy Convers. 23, 603611.
while emission and generation outputs are dependent to each other. IEAGHG, 2015. Integrated Carbon Capture and Storage Project at SaskPower's Boundary
However, with the genuine adoption of the CCS technology, the Dam Power Station. (August).
International Energy Agency, 2014. CO2 Emissions from Fuel Combustion-Highlights,
resulting thermal power plants present a new opportunity for the 2014 edition. IEA, Paris.
system operators. Operating the CCPPs exibly can lead to the Ji, Z., Kang, C., Chen, Q., Xia, Q., Jiang, C., Chen, Z., et al., 2013. Low-carbon power
independent control of the emission and generation outputs of the system dispatch incorporating carbon capture power plants. IEEE Trans. Power Syst.
28, 46154623.
facilities. This paper has proposed and investigated the concept of Jiaming, L., Jinyu, W., Xingning, H., 2015. Low-carbon unit commitment with intensive
decoupling emission and generation outputs as a result of exible wind power generation and carbon capture power plant. J. Mod Power Syst. Clean
operation of CCPPs. When the exible CCPP is considered within the Energy 3, 6371.
Kossoy, A., Grzegorz, P., Oppermann Klaus, Prytz, N., Alyssa, G., Klein Noemie, Lam
power generation mix, the system or plant operators are able to Long, Wong, L., 2015. Carbon Pricing Watch. World Bank, Washington, DC, pp. 2015.
capitalize from the existing market and/or regulatory situation. In line Lou, S., Lu, S., Wu, Y., Kirschen, D.S., 2015. Optimizing spinning reserve requirement of
with that, the operators dispatch routines should be altered in a way power system with carbon capture plants. IEEE Trans. Power Syst. 30, 10561063.
Lu, S., Lou, S., Wu, Y., Yin, X., 2013. Power system economic dispatch under low-carbon
that utilizes the resources available to them in respect to the opportu-
economy with carbon capture plants considered. IET Gener. Transm. Distrib. 7,
nities within the regulation/market. 9911001.
Mustafa, M., Abdilahi, A.M., Mustapha, M., 2016. Chaos-enhanced cuckoo search for
economic dispatch with valve point eects. TELKOMNIKA (Telecommunication
Acknowledgement
Computing Electronics and Control) 14.
Oates, D.L., Versteeg, P., Hittinger, E., Jaramillo, P., 2014. Protability of CCS with ue
The authors would like to acknowledge the Ministry of Higher gas bypass and solvent storage. Int. J. Greenh. Gas Control 27, 279288.
Education, Malaysia and the Universiti of Teknologi Malaysia for Secui, D.C., 2015. A new modied articial bee colony algorithm for the economic
dispatch problem. Energy Convers. Manage. 89, 4362 (1/1/2015).
providing nancial support under the Malaysian International Tao, D., Rui, B., Fangxing, L., Hongbin, S., 2015. A Bi-level branch and bound method for
Scholarship. economic dispatch with disjoint prohibited zones considering network losses. IEEE
Trans. Power Syst. 30, 28412855.
Walters, D.C., Sheble, G.B., 1993. Genetic algorithm solution of economic dispatch with
References valve point loading. IEEE Trans. Power Syst. 8, 13251332.
Yang, X.-S., Deb, S., 2009. Cuckoo Search via Lvy Flights, in World Congress on
Barisal, A.K., Prusty, R.C., 2015. Large scale economic dispatch of power systems using Nature & Biologically Inspired Computing. NaBIC 2009pp. 210214.
oppositional invasive weed optimization. Appl. Soft Comput. 29, 122137. Yang, X.-S., Deb, S., 2014. Cuckoo search: recent advances and applications. Neural
Brouwer, A.S., van den Broek, M., Seebregts, A., Faaij, A., 2015. Operational exibility Comput. Appl. 24, 169174.
and economics of power plants in future low-carbon power systems. Appl. Energy Zhan, J., Wu, Q., Guo, C., Zhou, X., 2015. Economic dispatch with non-smooth
156, 107128. objectivespart I: local minimum analysis. IEEE Trans. Power Syst. 30, 710721.
Chen, Q., Kang, C., Xia, Q., 2010. Modeling exible operation mechanism of capture van der Wijk, P.C., Brouwer, A.S., van den Broek, M., Slot, T., Stienstra, G., van der Veen,
power plant and its eects on power-system operation. IEEE Trans. Energy Convers. W., et al., 2014. Benets of coal-red power generation with exible CCS in a future
25, 853861. northwest European power system with large scale wind power. Int. J. Greenh. Gas
Chen, Q., Kang, C., Xia, Q., Kirschen, D.S., 2012. Optimal exible operation of a CO2 Control 28, 216233.
capture power plant in a combined energy and carbon emission market. IEEE Trans. Gandomi, A.H., Yun, G.J., Yang, X.-S., Talatahari, S., 2013. Chaos-enhanced accelerated
Power Syst. 27, 16021609. particle swarm optimization. Commun. Nonlin. Sci. Numer. Simulat. 18, 327340.
Cohen, S.M., Rochelle, G.T., Webber, M.E., 2011. Optimal operation of exible post-
19