Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Electrical Power and Energy Systems 113 (2019) 772–781

Contents lists available at ScienceDirect

Electrical Power and Energy Systems


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

Adjustable robust power dispatch with combined wind-storage system and T


carbon capture power plants under low-carbon economy

Rufeng Zhanga, Tao Jianga, , Linquan Baib, Guoqing Lia, Houhe Chena, Xue Lia, Fangxing Lic
a
Department of Electrical Engineering, Northeast Electric Power University, Jilin, JL 132012, China
b
Systems Engineering and Engineering Management, University of North Carolina at Charlotte, Charlotte, NC 28213, USA
c
Department of Electrical Engineering and Computer Science, The University of Tennessee, Knoxville, TN 37996, USA

A R T I C LE I N FO A B S T R A C T

Keywords: This paper proposes an adjustable robust power dispatch (ARPD) model under low-carbon economy to ac-
Robust power dispatch commodate the uncertainty of wind power with combined wind-storage system (CWSS) and carbon capture
Low carbon economy power plants (CCPPs). First, battery energy storage (BES), which is considered as adjustable under worst-case
Carbon capture power plants scenarios, is introduced to wind power plants to form a combined wind storage system structure to improve
Combined wind-storage system
operation flexibility. Then, a dispatching model of CCPPs is presented and analysed. The operating character-
istics of CCPPs make it possible to control net power outputs and carbon emissions independently and flexibly.
The integration of BES and carbon capture and storage (CCS) improves the operation flexibility of both wind
power plants and conventional power plants. The objective function of the proposed model under interval-
uncertain wind power is to minimize the total cost under the worst wind power output case and the context of
low-carbon economy. The proposed model is implemented on the modified PJM 5-bus system with one wind
farm and the IEEE-118 bus system with five wind farms. Numerical results demonstrate that the proposed model
is effective at reducing carbon emissions and wind power curtailment, and verify that as the number of CCPPs
increases, the overall costs come down.

1. Introduction resulting in net output reduction [8,9]. Carbon capture power plants
(CCPPs) control power output and carbon emissions separately, which
Global warming caused by CO2 has made it essential to develop low provides the operation flexibility [10]. The flexible operation me-
carbon economy [1]. For example, the carbon emissions reduction chanism of CCPPs has been modeled in [8] and [10]. Such operation
target scheme of the UK government is to achieve 80% reduction by flexibility has a significant impact on power system economic dispatch
2050 compared to the emissions level of 1990 [2]. The power industry under low carbon economy. Economic dispatch with CCPPs has been
is regarded as one of the largest producers of emissions, and carbon studied in [10–12], but none of them considers the operation within a
emissions should be taken into consideration in the process of power renewable energy context.
system dispatch to achieve emissions reduction goals [3]. Hence, re- Integration of large-scale volatile wind power can reduce fossil fuels
search on the operation of power systems under low carbon economy is consumption and carbon emissions by replacing part of the conven-
of great significance [3–5]. tional power output, but such integration will create challenges for
Promoting renewable energy integration and applying carbon cap- power system operation and dispatch. Such volatility (determined by
ture and storage (CCS) technology in conventional power plants can volatile wind speed) and inflexibility, which may result in wind power
contribute to reducing CO2 emissions [6]. CCS technology has been curtailment, are some of the disadvantages of wind power. Energy
regarded as one of the most promising options to reduce anthropogenic storage (ES) is an effective way to provide technical and economic
CO2 emissions from fossil fuel fired power plants, which is a major benefits to power systems, in which battery energy storage (BES) is a
source of CO2 emissions [6,7]. CCS technology incorporated in a fossil valid element in the design of wind power dispatch [13]. An approach
fuel fired power plant can capture CO2 and transport it to a storage for planning and operating in the electricity market from the view of
location with the consumption of a certain amount of energy. The combined wind storage systems (CWSS) is proposed in [14]. Compared
consumed power introduces an “efficiency penalty” effect to the plant, to common wind power plants, CWSS can provide more flexible power


Corresponding author.
E-mail address: tjiang@neepu.edu.cn (T. Jiang).

https://doi.org/10.1016/j.ijepes.2019.05.079
Received 13 December 2018; Received in revised form 10 April 2019; Accepted 29 May 2019
Available online 15 June 2019
0142-0615/ © 2019 Elsevier Ltd. All rights reserved.
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

Nomenclature PCG, i, t / PCN, i, t gross/net generation output (MW)


Pi,t base output of conventional units (MW)
ΔEc/ΔEd charging/discharging rates of BES PSOCmax maximum value of SOC (MWh)
Aid, t / Aiu, t downward/upward adjusted power (MW) PSOCmin minimum value of SOC (MWh)
cC cost coefficients of CO2 emissions ($/t) Rd reserve requirements (MW)
ci bidding prices of conventional unit i ($) Rdown/Rup up/down reserve requirements (MW)
cwind bidding prices of wind power units ($) S0,i,t initial volumes of solvent in the tanks
Eccpp,i,t captured CO2 emissions of unit i at period t (t) Sccpp,i,t dynamic volumes of solvent in the tanks
EG,i,t/E N,i,t gross/net CO2 emissions of unit i at period t (t) si,t participation factors
EP,i,t amount of CO2 emissions being treated (t) T number of dispatching periods
eG emission intensity per unit of gross power output (t.MW) U adjustable uncertainty sets
Hi/Gi the ith row of matrices H and G αC carbon capture rate
Liml limit for power flow of line l (MW) αC,max maximum value of carbon capture rate
Ni/Nj number of conventional/wind power units αj,t, βj,t dual variables
Pw,j,t dispatched wind power (MW) η auxiliary variable
Pc,j,t/Pd,j,t charging /discharging power of jth BES at period t (MW) ηc/ηd charging/discharging efficiency of BES
PCWSS,t dispatching power output of CWSS (MW) ηccpp operating energy penalty rate of CCPPs
PCEP, i, t power consumption of CCS (MW) μl-j,t, πl-j,t dual variables
,B EP , OP
PCEP, i, t / PC , i, t basic/operation power consumption of CCS (MW)
Pwf , j, t base-case dispatched wind power output (MW)

output. is proposed in this work for the first time. The main motivation of
Another disadvantage of wind power is uncertainty. Previously, adapting adjustable robust optimization is: for adjustable robust opti-
robust optimization has been utilized to address uncertainties in power mization, adjustable variables can be adjusted to make the solutions
system operations [15–23], by seeking a solution for immunizing feasible under uncertain scenarios, and BES considered in this paper is
against all possible uncertain scenarios. In [15], a robust scheduling also used to provide adjustable contributions with preset participation
scheme for energy storage systems (ESSs) is presented to facilitate high factors to maintain power balance under worst-case uncertain sce-
penetration of renewable energy sources. Bertsimas et al. [16] for- narios. To analyze the impact of ranging participation factors on the
mulates the security constrained unit commitment (SCUC) problem as a dispatching of BES, adjustable robust power dispatch method is
two-stage adaptive robust optimization model and a Benders decom- adapted. BES is incorporated in wind farms to reduce wind power
position type algorithm is employed to solve the model. Robust interval curtailment (carbon emissions can also be reduced) and to provide
wind power dispatch approaches for look-ahead dispatch are developed adjustable resources under uncertain scenarios. CCPPs are considered
in [17,18], in which the decision variables include not only the dis- to reduce carbon emissions under low-carbon economy. A low carbon
patched wind power, but also the allowable wind power output inter- robust dispatch model is formulated to minimize total operation costs,
vals. Adjustable robust optimization is a type of robust optimization, in with consideration of carbon emission costs raised by low carbon
which decision variables include here-and-now variables (‘non-ad- economy. Operation constraints of BES and CWSS are imposed, and
justable variables’) and wait-and-see variables (‘adjustable variables’) wind power curtailments and carbon emissions can be reduced. Quick
[19]. Adjustable variables can be adjusted to make the solutions fea- regulated conventional units are selected as adjustable units and BES is
sible under uncertain scenarios. An adjustable robust optimization ap- also used to provide adjustable contributions with preset participation
proach has been successfully utilized for transmission network expan- factors to maintain power balance under worst-case uncertain sce-
sion planning [20], optimal power flow [21,22] and economic dispatch narios. Compared with the existing works, the major contribution of
[23]. However, low carbon economy is not considered in the above this paper can be summarized as follows.
works.
To the authors’ best knowledge, an adjustable robust power dis- (1) An adjustable robust power dispatch model with CWSS and CCPPs
patch (ARPD) framework from the perspective of low carbon economy is proposed to account for uncertainties under low-carbon

Carbon
CWSS Conventional emission
units

Wind power
generator Net
carbon
Power
emission
grid

Operation
Net power
power consumption
BES output CCS

CCPP

Fig. 1. Schematic of the CWSS and CCPPs system.

773
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

economy, in which CO2 emissions costs arisen from the low-carbon the power mismatch caused by wind power uncertainties.
policy are considered and the allowable wind power output inter-
vals are also calculated. As compared to robust optimization models BES obviously cannot be operated to charge and discharge in the
in [15–22], the proposed robust dispatch model is carried out under same period. The power that a CWSS is dispatched at period t equals to
low-carbon economy, and the robust dispatch model is transformed the sum of the dispatched wind power and the charging/discharging
into a single level LP problem, which can be effectively solved with power of BES and cannot exceed the forecasting output under base case,
CPLEX. which can be expressed as:
(2) BES and CCS are considered in this paper to improve the operation
flexibility of power outputs on the supply side. On the one hand, as 0 ⩽ PCWSS, t = Pw, j, t - PcBES BES
, j, t + Pd, j, t ⩽ Pw, j, t (1)
compared to the low carbon dispatch models in [3,11,12], based on
the charging/discharging capacities of BES, the joint outputs of
CWSS in this paper are more flexible than those of wind farms (b) Operation model of CCPPs
without BES. Moreover, BES is considered to provide regulating
capacities under uncertain scenarios to ensure power balance. On Post-combustion, pre-combustion and oxy-fuel combustion CCPPs
the other hand, for conventional power plants with CCS, the power are the three major types of CCPPs, which differ in the integration
consumption of CCS can be flexibly regulated to adjust the net section between the generation unit and CCS [11]. CCPPs with post-
output of CCPPs. combustion have been utilized in practice and hence are selected in this
(3) In the proposed robust dispatch model, carbon emissions from paper. The structure of benchmark post-combustion based CCPPs are
power plants are considered as controllable variables to minimize presented in detail in [8]. The existing studies show that post-com-
total carbon emissions. The benefits of integration of CWSS and bustion CCPPs could be operated flexibly to change net output by
CCPPs are analyzed. The impact of CWSS and CCPPs on carbon controlling captured carbon emissions independently. CCS operation is
emissions and wind power curtailment reduction is quantified and powered by the plant where it is located. However, the power con-
compared. sumption of CCS is not negligible, which will lower the net output of
(4) Uncertainty of wind power is represented as forecasting wind CCPPs. Such a decrease in output would result in reduced operation
power intervals and the allowable wind power generation interval profits and energy penalties to the plant.
is calculated to provide additional operation information to in- The power output of the plant injected into the grid decreases due to
dependent system operators (ISO). the integration of CCS, which is regarded as an efficiency drop in
generation. Net generation output PCN, i, t (the electricity delivered to the
power grid) equals gross generation output PCG, i, t minus the energy
2. Flexible operation of CWSS and CCPPs
consumed PCEP, i, t by carbon capture systems. The net output is as shown

In this paper, CCS and BES are introduced to improve the rate of in (2):
wind power utilization and reduce carbon emissions. The schematic of
PCN, i, t = PCG, i, t − PCEP
, i, t (2)
the CWSS and CCPPs system is shown in Fig. 1. BES can reduce wind
power curtailment directly and carbon emissions indirectly, while CCS
In flexible operation mode, the net output of CCPPs can be adjusted
can reduce carbon emissions directly and wind power curtailment in-
by regulating the operation power consumption PCEP , i, t , which is ap-
directly. The power outputs of CWSS and CCPPs can be operated flex-
proximately proportional to the carbon capture rate. The power con-
ibly to reduce wind power curtailment, and carbon emissions of CCPPs
sumption PCEP , i, t of CCS consists of two parts: basic power consumption
can be controlled flexibly by capturing various amounts of CO2. ,B
PCEP
, i, t , which is supposed to be constant, and operation power con-
, OP , OP
sumption PCEP , i, t as shown in (3). PCEP
, i, t is formulated as (4).
(a) Operation of combined wind-storage system
PCEP EP , B EP , OP
, i, t = PC , i, t + PC , i, t (3)
Wind power output is volatile because it is determined by wind
speed and the transformation characteristic of wind turbines. Such
volatility means that wind power is not completely controllable or PCEP , OP
, i, t = ηccpp × αC × eG × PCG, i, t (4)
dispatchable. In our previous work [24], the structure of a CWSS is
presented, consisting of wind generators and BES connected to the same In this paper, CCS is considered to be operated at full-load or partial-
bus in a power system. With the advantage of having flexible charging/ load. The net output of CCPPs is related to the operation condition by
discharging ability, BES is effective at smoothing out fluctuating wind adjusting αC in the range of 0 – αC,max (80–95%), as shown in (5) and
power and making CWSS power output more flexible. BES can also (6):
reduce wind power curtailment. In this paper, the further goals of im-
plementing CWSS are: PCN, i = PCG, i − PCEP
, i, t

= (1 − ηccpp × αC × eG ) × PCG, i − PCEP ,B


∀ i ∈ GCCPP (5)
(1) To provide more flexible and dispatchable power output to the grid, , i, t

reducing not only wind power curtailment, but also carbon emis-
sions. BES charges/discharges power to reallocate dispatchable 0 ⩽ αC ⩽ αC max (6)
wind power across the dispatching time horizon, contributing to
peak load regulation. If the BES cannot store some power due to the Certain amount of CO2 can be captured and stored by CCS. The
limitation of the maximum charging capacity or charging rate, such operator can adjust the amount of captured CO2 to determine operation
, OP
extra power will be discarded (i.e. wind power curtailment). Hence, power consumption PCEP , i, t . By adjusting the carbon capture rate (the
both wind power curtailment and power output from fossil fuel operation power consumption is also adjusted), the CO2 emission in-
fired power units can be reduced, which will help to reduce carbon tensity of CCPPs can be sufficiently low or even negative. As shown in
emissions indirectly under low carbon economy. (7), net CO2 emission EN,i,t equals to gross CO2 emission EG,i,t minus
(2) BES can be operated under uncertain scenarios to accommodate captured CO2 emission Eccpp,i,t. It’s obviously unreasonable to capture
wind power uncertainty and ensure power balance within its op- CO2 in the atmosphere, so the net CO2 emission is limited to be zero or
eration constraints [21]. The charging (or discharging) ability of positive. In this paper, we assume the tanks can store CO2 with max-
, OP
BES can be treated as a type of adjustable resource responding to imum PCEP, i, t during certain periods in total.

774
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

G
⎧ EN , i, t = EG, i, t − Eccpp, i, t = eG × PC, i, t − αC × EP, i, t participation factors s as shown in (13). It is important to note that
⎪ EN , i, t ⩾ 0 participation factors s can be selected or optimized [21,22], however,
⎨ Eccpp, i, t + Sccpp, i, t − 1 = Sccpp, i, t that is not the focus of this paper.
⎪ max Δx = s (y s − y ) (13)
⎩ 0 ⩽ Sccpp, i, t + S0, i, t ⩽ Si ∀ i ∈ GCCPP (7)

In practical operation, it’s obviously unrealistic to incorporate tanks


4. Adjustable robust power dispatch model with CWSS and CCPPs
with volumes that allow long-term storage. It’s necessary to ensure that
under low-carbon economy
the amount of CO2 stored in solvent does not exceed the capacity of the
tanks.
In this section, we introduce the ARPD model with CWSS and CCPPs
under low carbon economy.
3. Adjustable robust interval power dispatch
(a) Decision variables
In this section, the abstract adjustable robust power dispatch
(ARPD) formulation with interval wind power is proposed.
Decision variables of the proposed ARPD model include:

(a) Brief formulation and solution


(1) The base power output of conventional units Pi,t (including normal
conventional units, adjustable units and CCPPs) and wind power
A robust power dispatch model with adjustable uncertainty inter-
Pw,j,t;
vals of wind power can be formulated as:
(2) The base power consumption of CCS, charging and discharging
power of BES;
⎧ x , min
y, y , y
f (x , y, y , y )
⎪ (3) The allowable wind power generation interval [ y , y ], the adjusted
⎪ x power of adjustable units and BES.
s. t . [H , G] ⎡ y ⎤ ⩽ C , ∀ y ∈ U (y , y )
⎨ ⎣ ⎦ (b) Objective function
⎪x ⩽ x ⩽ x, y ⩽ y ⩽ y
N N N
⎪[ y , y ] ⊂ [ y , y ] T i j i
⎧∑i = 1 (ci Pi, t ) + ∑ j c wind Pw, j, t + ∑i = 1 (cC EN , i, t ) + ⎫
⎩ (8) min ∑⎨ N ⎬
t ∑ j j c wind [(Pw, j, t - Pw, j, t ) + ( Pw, j, t − Pw, j, t )]
The constraints should be held for any uncertain wind power ⎩ ⎭
output, and the optimized uncertainty interval [ y , y ] should be within (14)
the forecasting wind power interval [ y , y ].
The objective function of the low carbon economic dispatch model
In robust optimization, to guarantee the security for any wind
is to significantly reduce total operation costs, including power bidding
power scenario, even for the worst-case scenario, the model is trans-
costs (conventional units with and without CCS and wind power), CO2
formed into a two-layer model as follows [17]:
emission costs (arising from low-carbon policy) and wind power cur-
tailments costs by minimizing the difference between wind power
⎧ Hi x + max(
y
Gi y ) ⩽ Ci ∀i
forecasting intervals and allowed intervals as shown in (14).
⎨ s. t . y ⩽ y ⩽ y
⎩ (9)
(c) Constraints for all wind power scenarios
By introducing an auxiliary variable, (9) can be transformed as:
(1) Power balance constraints in base case: Total generation (conven-
⎧ Hi x + ymax (Gi ( y + η (y − y )) ⩽ Ci tional units and output of CWSS) and load (load and power con-
, y, y
⎨ s. t . 0 ⩽ η ⩽ 1 sumed by CCS) are balanced at each time period t:
⎩ ∀i (10)
Ni Nj Nc

The dual problem of Eq. (10) is: ∑ Pi,t + ∑ (Pw,j,t − PcBES BES
, j, t + Pd, j, t ) = Dt + ∑ PCEP,i,t
i=1 j=1 i (15)
⎧ x ,min Hi x + Gi y + 1T βi ∀i
⎪ y, y (2) Generation output constraints of power units: Power outputs of con-
⎨ s. t . βi ⩾ Gi (y − y ) ventional units (gross output for CCPPs) are within their technical
⎪ βi ⩾ 0
⎩ (11) limits shown in (16) and dispatched wind power cannot exceed the
allowable interval shown in (17). Gcon and Gwind are the sets of
From duality theory and according to [17], the two-layer problem
conventional and wind power units.
can be expressed as:
Pi min ⩽ Pi, t ⩽ Pi max ∀ i ∈ Gcon (16)
min f (x , y , y )
x, y , y
Pw, j, t ⩽ Pw, j, t ⩽ Pw, j, t ⩽ Pw, j, t ⩽ Pw, j, t ∀ j ∈ Gwind (17)
s. t . Hi x + Gi y + 1T βi ⩽ C ∀i
βi ⩾ Gi (y − y )
(3) Ramping constraints:
βi ⩾ 0
x ⩽ x ⩽ x, y ⩽ y ⩽ y − Rampiu Δt ⩽ Pi, t - Pi, t − 1 ⩽ Rampiu Δt ∀ i ∈ Gcon (18)
y ⩽ y, y ⩽ y (12) − Rampid Δt ⩽ Pi, t - Pi, t − 1 ⩽ Rampid Δt ∀ i ∈ Gcon (19)

(b) Adjustable strategy and variables (4) Spinning reserve constraints: Upper and lower spinning reserves
should be provided by conventional units to guarantee the opera-
In this paper, several quick regulated conventional units are selected tion reliability of the system in base case.
as adjustable units [20,21] and BES is also used to provide adjustable
N
contributions to maintain power balance under worst-case scenarios.
∑ [min(Pi max − Pi,t , Rampiu Δt )] ⩾ Rd
The power mismatches under worst-case scenarios are adjusted by ad- i=1 (20)
justable conventional units and BES according to predefined

775
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

N u d u
⎧ Pi, t + Ai, t − Pi, t − 1 + Ai, t - 1 ⩽ Rampi Δt
∑ [min(Pi,t − Pi min, Rampid Δt )]⩾Rd ∀ i ∈ Gadj
i=1 (21) ⎨ Pi, t − 1 + Aiu, t − 1 − Pi, t + Aid, t ⩽ Ramp d Δt
⎩ i (27)

(5) Network constraints: The transmission limits are shown in (22). Pi, t + Aiu, t ⩽ Pi max
∀ i ∈ Gadj
Generation shift factor GSFl-i in DC power flow model is introduced. Pi, t − Aid, t ⩾ Pi min (28)
N N
Liml ⩽ ∑i =i 1 Gl − i × Pi, t − ∑b =b 1 Gl − b × Db, t +
N (2) For BES, the power mismatch can be adjusted by reserved charging/
∑ j =j 1 Gl − j × (Pw, j, t − PcBES BES
, j, t + Pd, j, t ) ⩽ Liml discharging abilities according to the participation factors sj,t as
∀ i ∈ Gcon ∀ j ∈ Gwind (22) shown in (29). The BES should be operated within the limits as
shown in (30) and (31).
(6) Rated capacity constraints of BES: The state of charge (SOC) of BES is N
B jd, t = sj, t ∑ j =j 1 (Pw, j, t − Pw, j, t ) ∀ j ∈ Gwind
calculated by (23) considering charging/discharging efficiency. The
N
SOC of BES should be limited to its rated upper and lower capacity B jc, t = sj, t ∑ j =j 1 (Pw, j, t − Pw, j, t ) ∀ j ∈ Gwind (29)
in (24). The total output of CWSS should be active as shown in (1).
PSOC , j (t ) = PSOC , j (t - 1) + ηc Pc, j, t − Pd, j, t /ηd ∀ j ∈ Gwind PSOC, j min ⩽ PSOC, j, t − 1 + ηc (Pc, j, t + B jc, t )
(23)
−(Pd, j, t + B jd, t )/ηd ⩽ PSOC, jmax ∀ j ∈ Gwind (30)
PSOC, j min ⩽ PSOC, j (t ) ⩽ PSOC, jmax ∀ j ∈ Gwind (24)
c d
⎧ ηc (Pc, j, t + B j, t ) − (Pd, j, t + B j, t )/ηd ⩽ ΔEc Δt
(7) Maximum charging/discharging constraints of BES constraints: The ∀ j ∈ Gwind
⎨ (Pd, j, t + B jd, t )/ηd − ηc (Pc, j, t + B jc, t ) ⩽ ΔEd Δt
charging/discharging power in time period t is limited by its ⎩ (31)
charging/discharging rates.
(3) The total power mismatch under worst-case scenarios is adjusted by
⎧ 0 ⩽ Pc, j, t ⩽ ΔEc Δt ∀ j ∈ Gwind adjustable units and BES in (32). The sum of participation factors
⎩ ⩽ d, j , t ⩽
⎨ 0 P ΔEd Δt (25) should be 1 as shown in (33).
i N
u j d N N
(8) CCPPs constraints: Refer to (2)–(7). ⎧∑i = 1 Ai, t + ∑ j = 1 B j, t = ∑ j =j 1 (Pw, j, t − Pw, j, t )
(d) Constraints under worst-case scenarios ⎨ ∑Ni Aid, t + ∑Nj B jc, t = ∑Nj (Pw, j, t − Pw, j, t )
⎩ i=1 j=1 j=1

In a robust optimization framework, solutions should be feasible ∀ i ∈ Gadj ∀ j ∈ Gwind (32)


under worst-case wind power scenarios. Problem (9) is a two-layer Ni Nj
robust interval optimization model, and the constraints are linear. Ac- ∑ si,t + ∑ sj,t = 1
cording to the method in Section ‘Adjustable Robust Interval Power i i (33)
Dispatch’ and [17], the worst-case scenario constraints are as follows:
(4) According to the method in Section ‘Adjustable Robust Interval
(1) For adjustable conventional units, the upward adjusted power Aiu, t Power Dispatch’, the positive/negative reserve constraints and po-
and downward adjusted power Aid, t are distributed according to the sitive/negative transmission limits can be transformed into
participation factors si,t as in (26). The adjustable units under worst- (34)–(35) and (36)–(37) by introducing dual variables αj,t, βj,t, μl-j,t,
case scenarios should satisfy the ramping and generation con- πl-j,t.
straints as in (27) and (28). N N N
⎧ ∑i =i 1 Pi, t − ∑i c PCEP j BES BES
, i, t + ∑ j = 1 (Pd, j, t − Pc, j, t )
u N ⎪
⎧ Ai, t = si, t ∑ j =j 1 (Pw, j, t − Pw, j, t ) ∀ i ∈ Gadj N N N
+ ∑i =i 1 Aiu, t + ∑ j =j 1 Pw, j, t − ∑ j =j 1 αj, t ⩾ Dt
⎨ Aid, t = si, t ∑Nj (Pw, j, t − Pw, j, t ) ⎨
∀ i ∈ Gadj ⎪ − αj, t ⩽ Pw, j, t − Pw, j, t ∀ i ∈ Gadj ∀ j ∈ Gwind
⎩ j=1 (26) (34)

E D
$10 $35
600MW 400MW
CCPP Brighton Sundance

$15
200MW
$30
$14 A B C 420MW
220MW
Park City Solitude

Fig. 2. PJM 5-bus system.

776
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

N N
∑i =i 1 Pi, t − ∑i c PCEP j BES N
BES There are two conventional units at Bus A (Park City). A CWSS in-
⎧ , i, t + ∑ j = 1 (Pd, j, t − Pc, j, t )
⎪ N N N
cluding a wind farm with 450 MW capacity and BES with 300 MWh
− ∑i =i 1 Aid, t + ∑ j =j 1 Pw, j, t + ∑ j =j 1 βj, t ⩽ Dt capacity is installed at Bus E, and its bidding cost is set to $5. The

⎪ βj, t ⩾ Pw, j, t − Pw, j, t ∀ i ∈ Gadj ∀ j ∈ Gwind participation factor of BES is 0.2. ηc and ηd are 0.9. The hourly charged/
⎩ (35)
discharged limits are all 30% of the capacity and initial SOC is 100 MW.
N i c N EP b N The load is distributed on three load buses. The lower generation limits
⎧ ∑i = 1 Gl − i × Pi, t - ∑i Gl − b × PC , i, t − ∑b = 1 Gl − b × Db, t +
⎪ of the conventional units are set to be 30% of their capacities. The
N Nj
∑ j =j 1 Gl − j × ( Pw, j, t − PcBES BES
, j, t + Pd, j, t ) + ∑ j = 1 μl − j, t ⩽ Liml generation unit at Bus E is assumed to be a CCPP, and units at Bus C and

⎪ μl − j, t ⩾ Gl − j × (Pw, j, t − Pw, j, t ) ∀ i ∈ Gadj ∀ j ∈ Gwind D are assumed to be adjustable units with the same participation factor
⎩ (36)
s = 0.4. The maximum capture rate is set as 90%. ηccpp is set as 0.269
N i c N EP b N MWh/t CO2. The CCPP can capture CO2 at a maximum rate for 4 h. The
⎧ ∑i = 1 Gl − i × Pi, t − ∑i Gl − b × PC , i, t − ∑b = 1 Gl − b × Db, t +
⎪ upward and downward system spinning reserve rates are both set as
N Nj
∑ j =j 1 Gl − j × ( Pw, j, t − PcBES BES
, j, t + Pd, j, t ) − ∑ j = 1 πl − j, t ⩾ Liml 10%. The carbon tax is 20 $/t CO2. System loads are shown in Fig. 3.

⎪ πl − j, t ⩾ −Gl − j × (Pw, j, t − Pw, j, t ) ∀ i ∈ Gadj ∀ j ∈ Gwind
⎩ (37)
(1) Analysis of wind power dispatch
The ARPD model can be represented as (38), in which dual decision
variables αj,t, βj,t, μl-j,t, πl-j,t are added. The two-layer robust optimization Wind power forecasting errors are assumed to be 20% in this paper.
models are transformed into a single level LP problem, which can be Robust interval optimization results of wind power are presented in
solved with CPLEX. Fig. 4. In time periods 1, 2, 4–8, 10 and 16, the upper allowable wind
power bounds are lower than the upper forecasting wind power bounds.
⎧ min Eq (14) The reason is that in those periods, if the forecasting wind power out-
⎪ s. t . Eq (1) − (7) puts equal to the upper forecasting values, wind power curtailment
⎨ Eq (15) − (37) would occur. The upper allowable wind power outputs are determined
⎪α , β , μ , π
j, t j , t l − j , t l − j, t ⩾ 0 ∀ j ∈ Gwind (38) according to load values. To maintain robust for worst-scenarios, the

upper allowable wind power outputs are less than the upper forecasting
wind power bounds in these periods. In all time periods, the lower al-
5. Case studies lowable wind power bounds are the same as the lower forecasting wind
power bound. The obtained allowable wind power interval is within the
In this section, the proposed adjustable robust optimization frame- forecasting interval.
work under low carbon economy with CWSS and CCPPs is performed
on the modified PJM-5 bus system with one wind farm and the IEEE- (2) Benefits of CCPPs and BES
118 bus system with several wind farms. The ARPD model is solved
using CPLEX [25] with Matlab on a PC with Intel Core i7 3.00 GHz CPU To show the impact of CCPPs and BES on dispatch results, the fol-
and 8 GB RAM. lowing four cases are conducted. BES is not adjustable in these cases.

(a) PJM-5 bus system Case 1: ARPD without considering BES and CCPP.
Case 2: ARPD with CCPP.
The parameters of the PJM 5-bus system are from [26] (see Fig. 2).

1800

1700

1600

1500
System Load (MW)

1400

1300

1200

1100

1000

900

800
2 4 6 8 10 12 14 16 18 20 22 24
Time Period/h
Fig. 3. System load of PJM-5 bus system.

777
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

Upper Forecasting Wind Power


600 Lower Forecasting Wind Power
Dispatched Wind Power
500 Upper Allowable Wind Power
Lower Allowable Wind Power
Wind Power (MW)
400

300

200

100

2 4 6 8 10 12 14 16 18 20 22 24
Time (h)
Fig. 4. Dispatched wind power calculated from robust interval optimization.

Case 3: ARPD with BES. A comparison of the results of these four different cases is reported
Case 4: ARPD with BES and CCPP. in Table 1. Overall cost is defined as the total cost considered in the
objective function over the entire 24 h. The column of CO2 emissions
The dispatched wind power (or output of CWSS when BES is in- shows that CCPPs can help to reduce the emission of CO2 while BES
tegrated) and output of Unit 3 for different cases are shown in Fig. 5. makes less contribution. It can be seen from the column of total dis-
Unit 3 is one of the adjustable units, s of which equals to 0.5. Note that patched wind power that the integration of either CCPPs or BES can
dispatched wind power in Case 4 is the largest comparing to other cases contribute to the accommodation of wind power. When BES and CCPPs
in most periods. For Unit 3, different amounts of dispatched wind are combined to ARPD, the wind power can be dispatched at the highest
power correspond to different output results. The output of Unit 3 in rate. Because of the amount of the wind power that has been accom-
Case 4 is also the largest in most periods, because larger dispatched modated, the overall cost can be reduced for Cases 2, 3 and 4. Since
wind power decreases the upward adjusting requirement for adjustable ARPD with BES and CCPPs accommodate more wind power and CCPPs
Unit 3 (the difference between upper wind power interval and dis- yield less CO2 emissions, Case 4 results in the least CO2 emissions, most
patched value decreases). dispatched wind power, and lowest costs. Please also note that if the

400
Wind Power (MW)

300

200
Case 1
100 Case 2
Case 3
Case 4
0
2 4 6 8 10 12 14 16 18 20 22 24
Time (h)

500
Case 1
Output of Unit 3 (MW)

Case 2
400 Case 3
Case 4
300

200

100
2 4 6 8 10 12 14 16 18 20 22 24
Time (h)
Fig. 5. Wind power and Unit 3 under different cases.

778
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

Table 1
Comparisons of results under four cases.
Cases CO2 emissions Total dispatched Total power output Total power output Total power output Total power output Total power output Overall costs
(104 t) wind power of Unit 1 (103 MWh) of Unit 2 (103 MWh) of Unit 3 (103 MWh) of Unit 4 (103 MWh) of Unit 5 (103 MWh) (105 $)
(103 MWh)

1 2.10 5.25 3.47 3.60 4.49 3.20 12.49 9.28


2 1.95 5.59 3.49 3.61 4.55 3.21 12.72 9.02
3 2.10 5.47 3.52 3.59 4.43 3.22 12.47 9.26
4 1.95 5.79 3.53 3.60 4.49 3.23 12.72 9.01

wind power curtailment cost increases, the overall cost of Case 4 can be costs of Unit 3 and Unit 4 are high, more power has to be dispatched
further reduced by its wind power accommodation ability. These ob- from the units corresponding to larger participation factors to provide
servations suggest that the proposed model can help to develop low adjustable capacities under worst-case scenarios.
carbon economy.
To further verify the benefit of BES, the SOC results of BES with (b) IEEE 118-bus system
participation factors s as 0, 0.2, 0.4 and 0.6 is presented in Fig. 6. When
s is equal to 0, SOC has the largest variance, and the reason for that is The IEEE 118-bus system is applied to further demonstrate the ap-
that the BES is considered not adjustable. When s is positive, the cor- plicability of the proposed model for large systems. This system consists
responding BES is considered an adjustable resource, and a larger of 118 buses, 54 generators, and 186 branches. The generator bidding
participation factor means more power mismatches caused by wind data are similar to that in [27]. Five wind farms are connected at Busses
power uncertainty should be compensated by the BES before it reaches 16, 37, 48, 75, and 83 with the half forecasting output the same as that
its operation limit. It can be seen from Fig. 6 that the SOC variances in the PJM 5-bus system. Power units at Busses 12, 54, 59, 61, 66 and
decrease with the increase of s. Moreover, the total utilized wind power 69 are assumed to be CCPPs and units 5, 12, 21 and 40 are adjustable
corresponding to different s values are 5787.4 MWh, 5787.4 MWh, units with the same participation factors 0.2. The participation factors
5678.5 MWh and 5539.8 MWh. As the participation factor increases, of the five BES are all set to 0.04. The hourly load is then multiplied by
the SOC variances and the amount of utilized wind power decrease. The different factors to match the installed generation.
reason is that the larger its participation factor is, the more ’charging/ In this subsection, we compare the ARPD model to the deterministic
discharging reserve’ BES must leave to be adjusted under worst-case economic dispatch (DED) model. For the DED model, wind power is
scenarios. limited by its forecasting output and no uncertainties are considered.
Table 2 lists the dispatching results of various CCPPs without BES The objective function includes the first three items of (14), but the
for 24 h. Note that when the number of CCPPs in the system increases, constraints under worst-case scenarios are ignored. The total operation
the total CO2 emissions decrease, which verifies that the integration of cost of the ARPD model is $6.13 * 106, while the total operation cost of
CCPPs can directly contribute to CO2 emissions reduction. The overall the DED model is $6.11 * 106. The utilized wind power and carbon
costs during the dispatching horizon also decrease with the increase in emissions are almost the same for the two models. The slight increase in
the number of CCPPs (more captured CO2 and less emissions). It is total cost for the ARPD model is caused mainly by the different dis-
worth noting that dispatched wind power increases when CCPPs are patching results of adjustable units.
considered, but extra CCPPs (more than 2 in this case) cannot increase Fig. 8 shows a comparison of the dispatched power outputs of the
the amount of dispatched wind power anymore because the utilized four adjustable units of the two models. Note that the dispatched power
wind power has reached the upper limit. outputs of the corresponding units are the same in most time periods for
Participation factors can influence the dispatched power of ad- the two models, i.e. the dispatched outputs of different units based on
justable units. Unit 3 and 4 correspond to units at Bus C and D. Six ARPD is almost as economic as the results from DED. The differences
subcases, cases 4.1–4.6 with different participation factors as shown in between the results from ARPD and from DED (e.g. unit 5 in periods 15
Fig. 7a), are conducted with the assumption that BES is not adjustable. and 16, and unit 12 in periods 16 and 23, etc.) are due to the ability of
The average output of adjustable units 3 and 4 are shown in Fig. 7b). ARPD to tackle wind power uncertainties. In other words, the solution
The sum of participation factors of Unit 3 and Unit 4 equals to 1. Note of the ARPD model is conservative and the adjustable units must have
that when the participation factor of Unit 3 or Unit 4 increases, the enough reserves to ensure power balance under worst-case scenarios in
average output increases. In the PJM 5-bus system, since the bidding certain time periods.

350
s=0
s=0.2
300 s=0.4
s=0.6

250
SOC (MWh)

200

150

100

50
5 10 15 20 25
Time (h)

Fig. 6. Exchanged power and SOC of BES with ranging participation factors.

779
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

Table 2
Dispatch results with various CCPPs.
Numbers of CO2 emissions Total dispatched Total power output Total power output Total power output Total power output Total power output Overall costs
CCPPs (104 t) wind power of Unit 1 of Unit 2 of Unit 3 of Unit 4 of Unit 5 (105 $)
(103 MWh) (103 MWh) (103 MWh) (103 MWh) (103 MWh) (103 MWh)

0 2.10 5.25 3.47 3.60 4.49 3.20 12.49 9.28


1 1.95 5.59 3.49 3.61 4.55 3.21 12.72 9.02
2 1.88 5.59 3.57 3.63 4.57 3.20 13.00 8.93
3 1.80 5.59 3.64 3.72 4.58 3.21 13.27 8.82
4 1.77 5.59 3.65 3.76 4.59 3.21 13.40 8.78
5 1.72 5.59 3.67 3.80 4.60 3.22 13.57 8.72

Unit 3 Unit 4 dispatched by adjusting the adjustable units and BES according to the
participation factor

1 corresponding participation factors. A trial is defined as successful if


0.8 none of the generation limits, power flow limits, and BES operation
0.6 limits are violated. The APRD model shows a 100% success rate for the
10,000 Monte Carlo trials, verifying that the proposed model can im-
0.4
munize the solution against wind power uncertainty.
0.2
The results indicate that the proposed ARPD model makes the ad-
0
4.1 4.2 4.3 4.4 4.5 4.6 justable units operate to have more reserves under worst-case scenarios
Cases with only a slight cost increase, and immunizes the solution against the
a) allowable interval wind power uncertainty set.

Unit 3 Unit 4
Power output of Unit 3

Power output of Unit 4

195 140 6. Conclusion


138
190
136 A novel adjustable robust power dispatch model with CWSS and
(MW)

(MW)

185 134 CCPPs under low carbon economy is proposed in this paper. The
132 structure of CWSS is presented to improve the operation flexibility of
180
130
wind power, and the CCPPs are incorporated to reduce carbon emis-
175 128
4.1 4.2 4.3 4.4 4.5 4.6 sions. Quick regulated units and BES are assumed to be adjustable with
Cases predefined participation factors under worst-case scenarios to make the
b) solution feasible under wind power uncertainty. Aiming to provide a
dispatch solution immunized against wind power uncertainty under
Fig. 7. Participation factors and average output of adjustable units under dif- low carbon economy, the ARPD model is formulated and transformed
ferent cases.
into an LP formulation based on interval wind power uncertainty.
Allowable wind power intervals are also calculated and can provide
To further demonstrate the necessity of having reserves for ad- reference to practice operations. Numerical cases show that the in-
justable units and the effectiveness of the proposed ARPD, a Monte tegration of BES and CCPPs can help to reduce wind power curtailments
Carlo analysis with 10,000 trials is carried out. In each Monte Carlo and carbon emissions. When the participation factor of BES increases,
trial, a wind power output scenario is generated randomly based on the BES must leave more ‘charging/discharging reserve’ to be adjusted
calculated allowable wind power interval. Then the system is re- under worst-case scenarios. More CCPPs would reduce carbon

400 500
DED-unit 5 DED-unit 12
ARPD-unit 5 400 ARPD-unit 12
Power (MW)

Power (MW)

300
300
200
200

100 100
5 10 15 20 5 10 15 20
Time (h) Time (h)

400 800

300
Power (MW)

Power (MW)

600
200
400
100 DED-unit 21 DED-unit 40
ARPD-unit 21 ARPD-unit 40
0 200
5 10 15 20 5 10 15 20
Time (h) Time (h)
Fig. 8. Comparison of outputs of adjustable units for ARPD and DED.

780
R. Zhang, et al. Electrical Power and Energy Systems 113 (2019) 772–781

emissions significantly and reduce overall costs. Numerical results on a [8] Heuberger C, Staffell I, Shah N, Dowell NM. Quantifying the value of CCS for the
larger system also verify that the proposed model can immunize the future electricity system. Energy Environ Sci 2016;9(8):2497–510.
[9] Zhou W, Zhu B, Szolgayová FJ, Obersteiner M, Fei W. Uncertainty modeling of CCS
solution against the allowable interval wind power uncertainty set. investment strategy in China’s power sector. Appl Energy 2010;87:2392–400.
In practice, the participation factors of adjustable units and BES [10] Chen Q, Kang C, Xia Q. Modeling flexible operation mechanism of capture power
should be allocated according to different operation conditions and plant and its effects on power-system operation. IEEE Trans Energy Convers
2010;25(3):853–61.
price signals. Therefore, the future work of this paper involves con- [11] Chen Q, Kang C, Xia Q, Kirschen DS. Optimal flexible operation of a CO2 capture
structing a more reasonable participation factors allocation method power plant in a combined energy and carbon emission market. IEEE Trans Power
considering time varying operation conditions and price signals. Syst 2012;27(3):1602–9.
[12] Lu S, Lou S, Wu Y, Yin X. Power system economic dispatch under low-carbon
economy with carbon capture plants considered. IET Gener Trans Distrib
Acknowledgements 2013;7(9):991–1001.
[13] Li Q, Choi SS, Yuan Y, Yao DL. On the determination of battery energy storage
capacity and short-term power dispatch of a wind farm. IEEE Trans Sustain Energy
This paper was supported in part by National Natural Science
2011;2(2):148–58.
Foundation of China (Grant No. 51877033, 51677022), National Key [14] Dicorato M, Forte G, Pisani M, Trovato M. Planning and operating combined wind-
Research and Development Program of China (2017YFB0903400), storage system in electricity market. IEEE Trans Sustain Energy 2012;3(2):209–17.
Integrated Energy System Innovation Team of Jilin Province [15] Yi J, Lyons PF, Davison PJ, Wang P, Taylor PC. Robust scheduling scheme for en-
ergy storage to facilitate high penetration of renewables. IEEE Trans Sustain Energy
(20180519015JH), and International Clean Energy Talent Programme 2016;7(2):797–807.
(iCET) of China Scholarship Council. [16] Bertsimas D, Litvinov E, Sun X, Zhao J, Zheng T. Adaptive robust optimization for
the security constrained unit commitment problem. IEEE Trans Power Syst
2013;28(1):52–63.
Appendix A. Supplementary material [17] Wu W, Chen J, Zhang B, Sun H. A robust wind power optimization method for look-
ahead power dispatch. IEEE Trans Sustain Energy 2014;5(2):507–15.
Supplementary data to this article can be found online at https:// [18] Li Z, Wu W, Zhang B, Wang B. Robust look-ahead power dispatch with adjustable
conservativeness accommodating significant wind power integration. IEEE Trans
doi.org/10.1016/j.ijepes.2019.05.079. Sustain Energy 2015;6(3):781–90.
[19] Ben-Tal A, Goryashko A, Guslitzer E, Nemirovski A. Adjustable robust solutions of
References uncertain linear programs. Math Program 2004;99(2):351–76.
[20] Moreira A, Street A, Arroyo JM. An adjustable robust optimization approach for
contingency-constrained transmission expansion planning. IEEE Trans Power Syst
[1] Bruninx K, Van Den Bergh K, Delarue E, D’haeseleer W. Optimization and allocation 2015;30(4):2013–22.
of spinning reserves in a low carbon framework. IEEE Trans Power Syst [21] Jabr RA, Karaki S, Korbane JA. Robust multi-period OPF with storage and renew-
2015;31(2):872–82. ables. IEEE Trans Power Syst 2015;30(5):2790–9.
[2] IPCC. Special report on carbon dioxide capture and storage. Cambridge (UK): [22] Jabr R. Adjustable robust OPF with renewable energy sources. IEEE Trans Power
Cambridge Univ. Press; 2005. Syst 2013;28(4):4742–51.
[3] Li X, Zhang R, Bai L, Li G, Jiang T, Chen H. Stochastic low-carbon scheduling with [23] Yang M, Wang MQ, Cheng FL, Lee WJ. Robust economic dispatch considering au-
carbon capture power plants and coupon-based demand response. Appl Energy tomatic generation control with affine recourse process. Int J Electr Power Energy
2018;210:1219–28. Syst 2016;81:289–98.
[4] Zhang R, Chen H, Li X, Jiang T, Li G, Ning R, et al. Low-carbon economic dispatch [24] Chen H, Zhang R, Li G, Bai L, Li F. Economic dispatch of wind integrated power
model with combined wind-storage system and carbon capture power plants. Proc systems with energy storage considering composite operating costs. IET Gener
2010 IEEE power & energy soc general meeting, July 2017. 2017. p. 1–5. Trans Distrib 2016;10(5):1294–303.
[5] Li Han, Eseye Abinet Tesfaye, Zhang Jianhua, Zheng Dehua. Optimal energy man- [25] ILOG CPLEX, ILOG CPLEX Homepage; 2009 [Online]. Available: http://www.ilog.
agement for industrial microgrids with high-penetration renewables. Protect com.
Control Modern Power Syst 2017;2(2):122–35. [26] Li F, Bo R. Congestion and price prediction under load variation. IEEE Trans Power
[6] Chen Q, Kang C, Xia Q, Zhong J. Power generation expansion planning model to- Syst 2009;24(2):911–22.
wards low-carbon economy and its application in China. IEEE Trans Power Syst [27] Fang X, Hu Q, Li F, Wang B. Coupon-based demand response considering wind
2010;25(2):1117–25. power uncertainty: a strategic bidding model for load serving entities. IEEE Trans
[7] Lu Z, Lu C, Feng T, Zhao H. Carbon dioxide capture and storage planning con- Power Syst 2015;31(2):1025–37.
sidering emission trading system for a generation corporation under the emission
reduction policy in China. IET Gener Trans Distrib 2015;9(1):43–52.

781

You might also like