Documente Academic
Documente Profesional
Documente Cultură
pubs.acs.org/IECR
1. INTRODUCTION
Steam power plants are the main energy supplier for running
chemical processing. Typically, a steam power plant consists of
various units including boilers, gas turbines, steam turbines,
electric motors, steam headers, etc. In the plant, steam is
converted into two types of energy, specically, electricity and
mechanical power. Electricity demands are from the power
required to function process devices. Mechanical power demands
are from the requirement to drive process units. Steam demands
are from heat duties for the heat exchange network or heat
sources for the reaction process.
The design of a steam power plant is a large and complex
problem, where the layout of all types of units and the operating
conditions must be optimized for ecient operation. The steam
distribution network (SDN) is an essential element in devising
the energy management system of a steam power plant. A large
volume of related studies have already been published in the
literature. Basically, two distinct approaches were adopted in
these works: (a) the heuristics-based thermodynamic design
method1,2 and (b) the model-based optimization method.35
The former networks were synthesized with thermodynamic
targets for getting the maximum allowable overall thermal
eciency, while the latter were designed with mixed-integer
linear/nonlinear programs for attaining the minimum total
annualized cost (TAC).
The above-mentioned works were developed to address the
design of an SDN assuming that all units operate at full load to
satisfy a single set of demands and conditions. However, in many
existing chemical processes the common operational feature is
varying demands. This may be due to changing feed/product
specications or changes of heat loss with seasonal variation in
the continuous operation plants, or changes in operations for
batch plants. For example, energy demands in peak season are
higher than in o peak season or steam power plants need more
heat demands in winter since the heat loss is higher.
Because of the limitations of these types of studies, capable
methodologies for the period-varying demands were developed.69
However, the research was only addressing operational problems
for existing plants or design problems without simultaneously
r 2011 American Chemical Society
2. PROBLEM STATEMENT
The design problem addressed in this paper is stated as
follows: Given are a set of steam demands or a set of hot/cold
Received: October 10, 2010
Accepted: May 6, 2011
Revised:
April 21, 2011
Published: May 06, 2011
8097
ARTICLE
3. SUPERSTRUCTURE
A superstructure of SDN is constructed to incorporate all
possible ow connections, as presented in Figure 1. More
Figure 2. Superstructures of SDN units: (a) boiler, (b) HRSG, (c) gas
turbine, (d) single-stage steam turbine, (e) multistage steam turbine,
(f) deaerator, and (g) steam header.
ARTICLE
pressure levels are selected at saturated steam pressure corresponding to the minimum temperature requirements. Specically, values not less than their saturated temperatures are
suitable for choices of steam headers. The operating temperatures of headers are treated as decision variables to be optimized.
For the simultaneous design of SDN and HEN, the level and the
amount of steam demands are not given. Both pressures and
temperatures of headers are variables to be determined by the
optimization approach.
There are three types of power-generating devices in the
superstructure for satisfying power demands, i.e., gas turbines,
steam turbines, and electric motors. Turbines can generate
electricity or mechanical power. A gas turbine includes a compressor, a combustor, and a turbine, where the hot air exiting the
turbine can further be used in HRSG to enhance the steam
production. Both back-pressure steam turbines and condensing
turbines are considered. For the back-pressure turbine, the
pressure of exhaust steam is higher than atmospheric pressure.
If the condensing turbine is chosen in the design, then a
condenser has to be selected for its operation. Electric motors
are also used to meet the required shaft power demands.
A deaerator is installed to remove dissolved gases to provide
feed water to the boilers to meet the process water demand.
Demineralized water is added to compensate for plant losses.
Pumps are included for the supply of boiler feed water and
cooling water and for the return of condensate.
The stagewise superstructure12 is adopted for the design of
HEN, as shown in Figure 3. Nonisothermal mixing of streams
is considered for the exible design. Each utility can be treated
as a process stream with unknown loads. By this superstructure one can determine the loads of utilities for the giving
process streams.
4. MODEL FORMULATION
Having introduced the superstructures of SDN and HEN, one
can then formulate the synthesis problem as a MINLP. The
material and energy balance equations associated with every unit
are included as the constraints of the optimization problem. The
corresponding equipment models adopted for SDN are taken
from Aguilar et al.10
4.1. Steam Distribution Network. 4.1.1. Boilers. Equation 1
states the mass balance of a boiler, where the feed boiler water
fbip
fbip
"b B, p P
1
fbp
iI
bfw deaer
fbp
H
qbp
iI
fbip hbip
iI
X bd sat, l
fbip Hi
"b B, p P
iI
2
bd
fbip
jfbip
"b B, p P
"b B, i I , p P
zbi g zbip
zbi e
"b B, i I , p P
zbip
"b B, i I
6
7
pP
There are two types of boilers used in this work, which are
multifuel boilers (b MB) and HRSGs (b HB). The
equipment models of boilers are adopted from Aguilar et al.10
4.1.2. Gas Turbines. For convenience, a set of gas turbines
g G is defined. Figure 2c shows the superstructure of a gas
turbine. Since exhaust of gas turbine g is sent to the HRSG unit b
8099
ARTICLE
Tgp
Tgbp
"g G , p P
b HB
tT , pP
ii0 t zii0 tp e wii0 tp e ii0 t zii0 tp
jJ
25
12
zb
zgb
" b HB
zgb g zgbp
zgb e
"g G , b HB, p P
15
zgbp
16
"g G , b HB
17
"g G , p P
18
"i, i0 I , i < i0 , t T , p P
zii0 t g zii0 tp
X
zii0 t e
gG
26
Also, operation for multiperiod problems should be considered, which is like the boiler and gas turbine units discussed
previously. Equations 27and 28 are the constraints of steam
turbines.
b HB
"i, i0 I , i < i0 ,
t TS , j J , p P
"g G , b HB, p P 13
24
"g G , b HB, p P 11
"g G , j J , p P
23
"i, i0 I , i < i0 ,
tT , pP
jJ
"i, i0 I , i < i0 ,
zii0 tp
"i, i0 I , i < i0 , t T
27
28
pP
pP
zgp
zgbp
"i, i0 I , i < i0 , t T , j J , p P
30
b HB
zgj g zgjp
zgj e
"g GS , j J , p P
zgjp
"g GS , j J
20
21
zii0 tj e
zgp
zii0 tjp
"i, i0 I , i < i0 , t T , j J
31
pP
pP
i, i0 I
i < i0
zgjp
"g GS , p P
22
jJ
ARTICLE
"i, i0 I , i < i0 , t MT
33
"i0 I , t MT
zii0 t e 1
4.1.4. Deaerator. Figure 2f shows a schematic representation for a deaerator device. In this unit the inlet streams
may come from low pressure steam, condensate return from
process, or treated water makeup. After water is treated and
its dissolved gas is removed, the feed water is sent to the boiler
or to the let-down station. Equation 41 describes the mass
flow rate balance. Equation 42 is an energy balance to
guarantee that enough steam is injected into the deaerator
so that the feed water leaving this unit is at saturated liquid
conditions.
X
X
X
bfw
fip fpc
fbp
fipld "p P 41
fpw
iI
fpw Hpw
34
iI
i < i0
fii0 tp fii0 tp
fii0 i00 tp
"i, i0 I , i < i0 , t MT , p P
i00 I
i < i0 < i00
35
X
fii00 tp
fii0 i00 tp
bB
fipld
i0 I
tT
X
i0 I
i0 > i
ps
i0 < i
fii0 tp
tT
fbip hbip
bB
pd
i0 I
i0 > i
X X
i0 I
tT
i0 < i
i I
i < i0 < i00
"i, i I , i < i , t MT , p P
i0 > i
00
i I
i < i0 < i00
ps
ps
i0 I
B
B
B
B X
B
B
B
B i0 I
@
38
43
i0 < i
zii0 i00 tp e 1
42
"i I , p P
"p P
i0 I
i <i
zii0 i00 tp e 1
H deaer
fi0 itp
iI
X
iI
fbip
00
36
i<i <i
bfw
fbp
bB
i0 I
0
iI
bB
X
tT
fii0 tp
X
i0 I
i0 > i
C
C
C
C
pd
fii0 p fip fipvent fip C
Chip
C
C
A
"i I , p P
44
39
00
00
t MT , p P
40
ARTICLE
t TS
i, i I
"h H , c C , k K , p P
mM
51
i <i
dem, s
wjp
"h H , c C , k K , p P
45
52
X
g GE
"j J , p P
wgp
X
t TE
i, i0 I
X
mM
,e
wii0 tp wimp
p
X wmjp
,e
wexp
p
m
jJ
"p P
46
heat exchange for all process streams. The constraints specify that
the overall heat of each hot process stream is removed with cold
process streams or cold utilities. Similar constraints also apply for
all cold streams, as stated in eqs 47 and 48:
X X
kK cC
qhckp qcu
hp
in
t h, 1, p
Thp
out
in
Tcp
Fcp
Tcp
kK hH
qhckp qhu
cp
"h H , p P
"c C , p P
qhckp
49
tckp tc, k1, p Fcp
qhckp
"c C , k K , p P
55
"c C , p P
56
"h H , k K , p P
57
"c C , k K , p P
58
"h H , p P
59
out
Tcp
g tc, 1, p
"c C , p P
60
"h H , k K , p P
cC
"h H , p P
out
e th, K1, p
Thp
48
54
47
X
"c C , k K , p P
in
out
Thp
Fhp
Thp
fchkp Fcp
hH
i < i0
,e
wdem
p
cC
"h H , p P
61
"c C , p P
62
hH
qhckp zhckp e 0
50
"h H , c C , k K , p P
63
4.2.3. Heat Balance for Each Unit. For each local exchange
unit, heat balances are needed, where fhckp and fhckp are split heat
cu
qcu
hp zhp e 0
8102
"h H , p P
64
ARTICLE
"c C , p P
65
4.2.8. Calculation of Approach Temperatures. For determining the area requirement of the heat exchanger, approach
temperatures are used to calculate the log mean temperature
difference (LMTD) approximated by using the Chen equation.13
The area requirement of each match will be incorporated in the
objective function. The constraints 6669 are expressed as
inequalities since the cost of the exchanger decreases with higher
values for the approach temperatures.
8600 h/year
45 000 kJ/kg
50 244 kJ/kg
electric prices
0.07 $/kWh
0.19 $/kg
0.22 $/kg
0.05 $/ton
"h H , c C ,
kK, pP
dthc, k1, p e thc, k1, p tc, k1, p 1 zhckp
"h H ,
c C, k K , p P
out, cu
1 zcu
dtcu
hp e th, K1, p T
hp
period
66
67
"h H , p P
68
out, hu
tc, 1, p 1 zhu
dthu
cp e tcp
cp "c C , p P 69
min J1
x 1 1
X
pP
Cu fbup
bB uU
bB
20
55
16
66
22
60
10
45
75
82
84
60
electricity demands
4.5
7.2
2.8
3.5
1.2
2.0
1.3
1.8
1.5
1.0
1.1
0.9
0.7
0.6
0.5
0.8
x1
X
dD
pd
imp;e
exp;e
max
max
wii0 tjp ; wmax
; wp ; zb ; zbp ; zbi ; zbip >
g ; wt ; wmjp ; wm ; wp
>
>
>
>
>
>
z
;
z
;
z
;
z
;
z
;
z
;
z
;
z
;
z
;
z
d g gp gb gbp gj gjp m t ii0 t ; zii0 tp ; zii0 tj ; zii0 tjp ; zii0 i00 tp >
>
>
>
>
>
>
0
00
>
>
"b B; d D ; g G ; i; i ; i I
>
>
>
>
>
>
>
>
;
:
j J ;m M ;p P;t T ;u U
pP
cw
imp, e imp, e
, e wexp, e
Cwp fpw Ccw
wp
Cexp
p f p Cp
p
p
X X
Cu fbup
bB uU
bB
g
var
zg Cfix
g Cg G g
tT
zb Cfix
b
gG uU
b
Cvar
b Gb
t
Cvar
t Gt
d
Cvar
d Gd
X X X
hH
70
cC
8103
hrs
Cu fgup qcu
hp tp
g
var
zg Cfix
g Cg Gg
mM
var m
zm Cfix
m Cm Gm
var
d
zd Cfix
d Cd Gd
hH cC kK
var m
zm Cfix
m Cm G m
X X
gG
zt Cfix
t
dD
zd Cfix
d
ps
fii0 i00 tp ; fii00 tp ; fpw ; fip ; fpc ; fipld ; fii0 p ; fip ; fipvent ; fip ; fdmax
hbip ; hii0 tp ; hip ; hcp ; Tgp ; Tgbp ; qbp ; qbup ; qgp ; wgp ; wgjp ; wii0 tp
var
t
zt Cfix
t Ct G t
9
>
>
>
>
>
>
>
>
>
>
>
>
=
bfw
bd max
fbp ; fbip ; fbip
; fb ; fbup ; fgp ; fgbp ; fii0 tp ; fii0 tp
min J2
x2 2
Cu fgup tphrs
8
>
>
>
>
>
>
>
>
>
>
>
>
<
gG
mM
gG uU
var
b
zb Cfix
b Cb Gb
tT
X X
cw
imp, e imp, e
,e
Cwp fpw Ccw
wp
Cpexp, e wexp
p fp C p
p
X X
var
hck
zhck Cfix
hck Chck Ghck
fix
var
h
zcu
h Ch Ch Gh
fix
var c
zhu
c Cc C c G c
71
ARTICLE
x2
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
bfw
hu
cu
bd max
dtcp
; dthckp ; dthp
; fbp ; fbip ; fbip
; fb ; fbup ; fchkp ; fgp
0
0
fgbp ; fhckp ; fii0 tp ; fii0 tp ; fii0 i00 tp ; fii00 tp ; fpw ; fip ; fpc ; fipld ; fii0 p
ps vent pd max
out;hu
fip ; fip ; fip ; fD ; hbip ; hii0 tp ; hip ; hcp ; tchkp ; tckp ; tcp
hu
thckp ; thkp ; Tgp ; Tgbp ; qbp ; qbup ; qcp ; qgp ; qhckp ; wgp ; wgjp ; wii0 tp
imp;e
exp;e
>
max
max
max
>
w
; wp ; zb ; zbp ; zbi ; zbip ; zhu
>
ii0 tjp ; wg ; wt ; wmjp ; wm ; wp
cp
>
>
>
>
zg ; zgp ; zgb ; zgbp ; zgj ; zgjp ; zhckp ; zcu
; zt ; zii0 t ; zii0 tp ; zii0 tj ; zii0 tjp ; zii0 i00 tp
>
hp
>
>
>
0 00
>
>
>
>
:
"b B; c C ; d D ; g G ; h H ; i; i ; i I
j J ;k K ;m M ;p P;t T ;u U
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;
5. CASE STUDIES
In this section, two case studies are presented to demonstrate
the application of the proposed MINLP model. In case 1, SDN
design with the given steam demands is studied. The process data
are taken from the work of Bruno et al.,4 which was originally
solved for the single period operation only. The other period
demands are added in the present example to facilitate a multiperiod SDN design. In case 2, simultaneous design for SDN and
HEN is studied, where the interaction between a steam system
and a heat recovery system can be optimized.
The site conditions for case studies are presented in Table 1.
The optimization platform employed was the General Algebraic
Modeling System (GAMS).14 The solver used was SBB15 for the
MINLP model. An Intel Core 2 Duo CPU 2.53 GHz computer
with 1 GB of RAM was used.
45
257.4
369.0
17
204.3
265.0
147.9
148.0
4.5
5.1. Case 1. SDN Design Problem Associated with Multiperiod Demands. Let us first consider the SDN design problem
ARTICLE
CP (kW/C)
Tin (C)
Tout (C)
H1
205
388
110
H2
152
210
60
119.90
C1
753
100
200
scenario 1
scenario 2
scenario 3
139.72
134.75
120.63
121.75
123.78
7.60
0.00
10.73
C2
377
140
255
9.91
10.51
11.03
C3
143
70
140
ARTICLE
4500
2000
1200
P, pressure (MPa)
(2) large package boiler
F, maximum steam ow rate (tons/h)
4954F0.77fp2
fp2 = 1.3794 0.5438P
0.1879P2
P, pressure (MPa)
(3) heat recovery steam generator
941Ffg0.75
81594 18.052Wst
321350 67.618Wgt
8141 0.6459Weg
1601 27.288Wel
7271 79.25FB
3977 1.84Fc
(475.3 34.95Pw
0.0301Pw2)fpw
6. DISCUSSION
In scenario 1 of case 1, temperatures of steam headers are
specied before a network structure is available. On the other
ARTICLE
hand, header temperatures are treated as temperature-independent and therefore operation and design possibilities may be
restricted. In scenario 2, the header temperatures are considered
as decision variables and a one-step procedure is developed with
the ability to optimize the network structure and the operating
conditions simultaneously. It is appreciated that the better design
and operation can be accomplished under this approach. In
scenario 3, four periods with electricity import and export are
studied. The result reveals that steam turbines for the electricity
generation are preferred due to the higher operating exibility.
Thus, the steam system can maintain higher operating eciency
throughout all periods.
In case 2, simultaneous design for SDN and HEN is studied.
The proposed model can determine both the moderate operating
conditions and the corresponding network for the steam system
and the heat recovery system. The operating condition determination can aect the operation eciency for steam systems
and the heat recovery circumstance for the given chemical
process. In this work, the interaction between two systems can
be optimized.
7. CONCLUSION
Changes in specications, composition of feed, and seasonal
product demands may cause several process conditions with
variation in the energy requirements during an annual horizon. In
the rst part, an MINLP model, based on unit superstructures,
has been developed to design a steam system with variable utility
demands. Complex multiperiod scenarios were studied that
together consider the design and operation of steam power
systems in an industrial plant. In the second part, a novel
methodology has been developed to address the design of a
steam system and a heat recovery network. This work determines
the optimal structure for both SDN and HEN, and also estimates
the moderate operating conditions. The results from the case
studies demonstrate that better energy management and utilization can be realized with the proposed model.
APPENDIX
The investment cost data according to Bruno et al.4 are
itemized in Table 7.
AUTHOR INFORMATION
Corresponding Author
ACKNOWLEDGMENT
Financial support of the National Science Council of ROC
(under Grants NSC98-3114-E-002-009 and NSC100-3113E-002-004) is appreciated.
NOMENCLATURE
Indices
Cx
* = xed coecient function for units, where * = {b, c, d, g,
h, m, t}
Cvar
* = variable coecient function for units, where * = {b, c, d, g,
h, m, t}
Cwp = cost per unit mass of demineralized water makeup in time
period p, $ kg1
cw
Cp = cost per unit mass of cooling water in time period p, $ kg1
Cimp,e
= specic cost of imported electricity in time period p,
p
$ kWh1
exp,e
Cp = specic cost of exported electricity in time period p,
$ kWh1
Cu = cost per unit mass of fuel u, $ kg1
Fcp = heat capacity ow rate for cold process stream c in period p,
kW/C
Fhp = heat capacity ow rate for hot process stream h in period p,
kW/C
G* * = coecient function of units, where * = {b, c, d, g, h, m, t}
= enthalpy of saturated steam at steam header i level,
Hsat,l
i
kJ kg1
ps
Hip = enthalpy of steam supplied by processes and delivered at
header i in period p, kJ kg1
w
Hp = enthalpy of demineralized water makeup in period p,
kJ kg1
deaer
H
= enthalpy of water leaving a deaerator, kJ kg1
hrs
tp = number of operating hours in time period p, h period1
Tin
cp = inlet temperature of cold process stream c in period p, C
Tout
cp = outlet temperature of cold process stream c in period p, C
Tin
hp = inlet temperature of hot process stream h in period p, C
Tout
hp = outlet temperature of hot process stream h in period p, C
= shaft power demand j in time period p, kW
wdem,s
jp
ARTICLE
h ii0 t, ii0 t = upper and lower bounds of steam ow rate for steam
h
turbine t, kg s1
h g,
h
turbine g, kW
h ii0 t, ii0 t = upper and lower bounds of power generation for steam
h
turbine t, kW
dthu
cp = temperature approach for the match of cold stream c and
hot utility in period p, C
dthckp = temperature approach for match (h, c) at temperature
location k in period p, C
dtcu
hp = temperature approach for the match of hot stream h and
cold utility in period p, C
fmax
= maximum steam ow rate for boiler b, kg s1
b
fbip = steam output from boiler b to steam header i in time period
p, kg s1
bd
fbip = blowdown water for boiler b at pressure i in time period p,
kg s1
bfw
fbp = boiler feed water for boiler b in time period p, kg s1
fbup = fuel u consumed in boiler b in time period p, kg s1
fchkp = heat capacity ow rate for cold stream c at stage k in time
period p, kW/C
= maximum water ow rate for deaerator, kg s1
fmax
d
fgbp = exhaust gas from gas turbine g to HRSG b in time period p,
kg s1
fgp = gas turbine g exhaust mass ow rate, kg s1
fgup = fuel u consumed in gas turbine g in time period p, kg s1
fhckp = heat capacity ow rate for hot stream h at stage k in time
period p, kW/C
fii0 tp = steam ow rate from header i to header i0 through a steam
turbine t in time period p, kg s1
f0ii0 tp = input steam ow rate from header i to header i0 through a
multistage steam turbine t in time period p, kg s1
0 0 00
fii i tp = steam ow rate from stage i0 to stage i00 for a multistage
steam turbine t in time period p, kg s1
fii0 p = steam ow rate from header i to header i0 in time period p,
kg s1
fip = steam ow rate from header i to deaerator in time period p,
kg s1
ld
fip = desuperheating boiler feed water injected into header i in
time period p, kg s1
pd
fip = steam process demand at header i in time period p, kg s1
1
fps
ip = steam from process entering header i in time period p, kg s
vent
1
fip = vented steam at header i in time period p, kg s
fcp = condensate return in period p, kg s1
fwp = demineralized water makeup in time period p, kg s1
fcw
p = cooling water mass ow rate for condensers in time period
p, kg s1
hbip = enthalpy of steam generated by boiler b entering header i in
period p, kJ kg1
8108
ARTICLE
REFERENCES
(1) Nishio, M.; Itoh, J.; Shiroko, K.; Umeda, T. A thermodynamic
approach to steam-power system design. Ind. Eng. Chem. Process Des.
Dev. 1980, 19, 306.
(2) Chou, C. C.; Shih, Y. S. A thermodynamic approach to the design
and synthesis of plant utility systems. Ind. Eng. Chem. Res. 1987, 26, 1100.
(3) Papoulias, S. A.; Grossmann, I. E. A structural optimization
approach in process synthesisI. Comput. Chem. Eng. 1983, 7, 695.
(4) Bruno, J. C.; Fernandez, F.; Castells, F.; Grossmann, I. E. A
rigorous MINLP model for the optimal synthesis and operation of utility
plants. Chem. Eng. Res. Des. 1998, 76, 246.
(5) Chang, C. T.; Hwang, J. R. A multiobjective programming
approach to waste minimization in the utility systems of chemical
processes. Chem. Eng. Sci. 1996, 51, 3951.
(6) Hui, C. W.; Natori, Y. An industrial application using mixedinteger programming technique: A multi-period utility system model.
Comput. Chem. Eng. 1996, 20, S1577.
(7) Iyer, R. R.; Grossmann, I. E. Optimal multiperiod operational
planning for utility systems. Comput. Chem. Eng. 1997, 21, 787.
(8) Maia, L. O. A.; Qassim, R. Y. Synthesis of utility systems with
variable demands using simulated annealing. Comput. Chem. Eng. 1997,
21, 947.
(9) Micheletto, S. R.; Carvalho, M. C. A.; Pinto, J. M. Operational
optimization of the utility system of an oil renery. Comput. Chem. Eng.
2008, 32, 170.
(10) Aguilar, O.; Perry, S. J.; Kim, J.-K.; Smith, R. Design and
optimization of exible utility systems subject to variable conditions,
Part 1: Modelling Framework. Chem. Eng. Res. Des. 2007, 85, 1136.
(11) Aguilar, O.; Perry, S. J.; Kim, J.-K.; Smith, R. Design and
optimization of exible utility systems subject to variable conditions,
Part 2: Methodology and Applications. Chem. Eng. Res. Des. 2007,
85, 1149.
(12) Yee, T. F.; Grossmann, I. E. Simultaneous optimization models
for heat integrationII. Heat exchanger network synthesis. Comput.
Chem. Eng. 1990, 14, 1165.
(13) Chen, J. J. J. Letter to Editors: Comments on improvement on a
replacement for the logarithmic mean. Chem. Eng. Sci. 1987, 42, 2488.
(14) GAMS: A Users Guide; GAMS Development Corp.: Washington,
DC, 2008.
(15) GAMS: The Solver Manuals; GAMS Development Corp.:
Washington, DC, 2007.
8109