Anda di halaman 1dari 12

Ind. Eng. Chem. Res.

1998, 37, 1051-1062 1051

A Rule-Based Steam Distribution System for Petrochemical Plant


Operation
Heui-Seok Yi, Yeong-Koo Yeo,* and Jin-Kuk Kim
Department of Chemical Engineering, Hanyang University, Seoul 133-791, Korea

Myung Kil Kim and Sun Sup Kang


Yochon Plant, Daelim Industrial Company, Yochon, Chunnam, Korea

A rule-based expert system for the optimal operation of plantwide steam distribution systems
is proposed to minimize the net cost of providing energy to the plant. The system is based on
the steady-state modeling and simulation of steam generation processes and steam distribution
networks. Modeling of steam generation processes and steam distribution networks was
performed based on actual plant operation data. Heuristic operational knowledge obtained from
experienced plant engineers is incorporated in the form of IF-THEN rules. The proposed system
could provide operational information when there were changes in the grade and amount of
steam demand. The letdown amount from the very high pressure steam (VS) header and the
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

amount of VS produced at the boiler showed good agreement with those of actual operational
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

data. The prediction of an increase of boiler load caused by self-consumed steam made it possible
to prevent an unexpected sudden increase of electricity demand.

1. Introduction deaerator and steam header except the boiler system.


Nishio et al. [1982] discussed a two-level approach
Most petrochemical plants consume great amounts of determining the optimal supply and demand relation-
energy in the form of steam and electrical power. ship of steam and power based on thermodynamic
Electrical power is purchased from an electrical power analysis of the system performance. In their later work
company and part of it is produced at the plant site from [1985] they used a linear programming method to solve
turbine generators. Electrical power drives motors and a problem regarding the selection of heating devices
is used directly in some processes. Steam is produced under given loads. Petroulas and Reklaitis [1984] also
in boilers, generated in processes, or recovered from heat utilized a linear programming method to solve problems
exchangers. Steam is a working fluid when powering of driver allocation and selection of the header and
turbines and a heat-transfer medium and sometimes is power source. Maia et al. [1995] set up a combinatorial
injected directly into a process. optimization problem by discrete representation of
In general, energy requirements from the process equipment capacities and the conditions of steam to
units are not constant. Especially, demands for energy solve the synthesis problem of utility systems.
change as production rates of individual units change. So far most of the operations concerning steam
Energy efficiencies of boilers change with steam loads, distributions have heavily relied on the heuristics of
and efficiencies of turbines are a function of throughput. experienced engineers. In some cases they use various
The cost of fuels consumed in each boiler varies with modeling and simulation software tools to make deci-
the type of fuel and the rate of consumption. The cost sions. But, changes in plant demands and operating
of power purchased from the electrical power company conditions to minimize energy costs often require repeti-
is a complex function of the base consumption level, tive optimization computations within a very short
actual power consumption, peak demand, time of day, period (sometimes within half an hour). Therefore, for
and day of year. the steam distribution system to be successful, utiliza-
To minimize the net cost of energy supplied to the tion of heuristic knowledge in addition to modeling and
plant, it is obvious that the operation of the energy computational simulations for the energy distribution
management system should be optimized. Many efforts system is the essential prerequisite. The objective of
have been devoted to the study of minimization and the present work is to develop a rule-based energy
synthesis of heat integration and utility systems. Gor- distribution system based on the steady-state steam
don and Hashemi [1978] presented a simple boiler- distribution systems to optimize the operations of the
turbine system using steady-state mass and energy plantwide energy management system.
balances. Nishio et al. [1980] described a thermody-
namic approach to steam-power system design with
2. Overview of the Steam Distribution System
some numerical design examples. Clark and Helmick
[1980] employed an iterative linear programming algo- The steam distribution system for a plant basically
rithm in the design of steam systems. They handled consists of a network of headers connecting the distribu-
all the main components of steam systems such as the tion system to the process units. The steam headers
are a very high pressure steam (VS) header, a high-
* Author to whom correspondence is addressed. Fax: +82- pressure steam (HS) header, a medium-pressure steam
2-291-6216. E-mail: ykyeo@email.hanyang.ac.kr. (MS) header and a low-pressure steam (LS) header.
S0888-5885(97)00214-5 CCC: $15.00 © 1998 American Chemical Society
Published on Web 02/04/1998
1052 Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

Figure 1. Process flow diagram of the steam distribution unit.

Table 1. Temperature and Pressure of Each Steam


Header
VS HS MS LS
header header header header
temperature (°C) 520 380 285 200
pressure (atm) 120 40 16 4

Figure 1 shows a steam distribution system of a typical


petrochemical plant. Pressures and temperatures of
each steam header of the particular plant considered
in the present study are listed in Table 1.
One or more utility boilers and/or waste heat boilers
in the process units generate VS. The boiler feed water
Figure 2. Schematic of a simple steam distribution system.
(BFW) is supplied from deaerator pumping systems. VS
from the boiler is fed to a VS header. Turbine genera-
tors use VS to generate electricity and release MS, LS, pose that there are three BFW pumps and that one is
and steam condensate (SC). The HS header is supplied driven by an electrical motor and the others are driven
with steam by VS/HS letdown desuperheaters. To by steam turbines. Also suppose that one steam turbine
maintain the pressure and temperature of the HS consumes 10 ton/h of steam, and 25.5 ton/h of steam is
header, VS is desuperheated by cooling water. The MS required in process units. If two BFW pumps are
header is supplied with steam by HS/MS letdown needed, there can be two operation modes as follows:
desuperheaters and by release from turbine generators. Mode 1: number of steam turbines ) 1, number of
Steam-air heater (SAH), high-pressure feed-water electrical motors ) 1, amount of steam from letdown )
heater (HPH), and process units consume MS. The LS 15.5 ton/h.
header is supplied with steam by MS/LS letdown Mode 2: number of steam turbines ) 2, amount of
desuperheaters, by release from turbine generators, and steam from letdown ) 5.5 ton/h.
by release from turbines connected to utility pumps. Because the amount of steam from letdown is smaller
Utility pumps are driven either by steam turbines or in mode 2, we can say that the operation mode 2 is more
by electrical motors. For example, consider the simple effective.
steam distribution unit shown in Figure 2. The boiler In Figure 1, the boiler blowdown rate is determined
in the figure produces only HS which is used to drive in order to maintain the specific boiler drum water
the BFW pump. The LS header is supplied with steam characteristics. This rate is a function of the deminer-
by release from the BFW pump and by the HS/LS alized water quality, combined with the boiler steam
letdown valve. LS is consumed in process units. Sup- output rate. The blowdown water is flashed in the
Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998 1053

Figure 4. Flow diagram of the deaerator.

Enthalpy balances around the deaerator and HPH


give
Figure 3. Schematic diagram of the boiler system.
(mBFW + mPLT)HDEA ) mWTRHWTR + mHPHHlMS +
boiler blowdown flash tank to produce LS and low- mDEAHLS + mSAHHlMS (2)
pressure condensed effluent.
mBFW(HlHPH,o - HlHPH,i) ) mHPH(HMS - HlMS) (3)
3. Steady-State Modeling of the Steam
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

Distribution Process 3.2. Steam Generator. The amount of energy


Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

required in the steam generator is equal to the differ-


Steady-state modeling of the boiler is in essence heat ence between the enthalpy of the feed-water and that
and mass balances which make use of specified input of the VS steam (comprised blowdown water from the
operating parameters such as the boiler blowdown boiler). In the plant considered in the present study,
fraction, the steam loss from the deaerator, and others. the blowdown rate is assumed to be 1% of the VS
From the model we can estimate the amount of steam produced.
required in the steam-air heater, in the high-pressure From simple mass balances on the steam generator
feed-water heater, and in the deaerator and the we have
amount of the fuel (bunker-C oil). Figure 3 shows a
mCBD ) 0.01mVS (4)
schematic of the boiler system considered in the present
work.
mBFW ) mVS + mCBD (5)
For the modeling, the boiler is regarded as consisting
of three main parts: deaerator, steam generator, and Thus, the amount of energy required in the steam
steam-air heater. The deaerator produces gas-free generator is given by substitution of (4) into (5) with
water to be used as BFW and process water in other some rearrangements as
plant units. The steam generator is supplied with BFW
to produce VS (520 °C, 118 kg/cm2). The steam genera- ∆H ) (HVS + 0.01HCBD - 1.01HBFW)mVS (6)
tor consists of a steam drum, a water drum, water tubes,
first superheaters, and second superheaters. Bunker-C 3.3. Steam-Air Heater. In the steam-air heater
oil is used as fuel. In the steam-air heater, combustion the combustion air (15 °C) is heated to 80 °C and fed
air is preheated up to a suitable temperature and fed into the air heater (gas type). From simple enthalpy
to the gas-air heater. The preheating of air is required balance we have
to avoid corrosion of the gas-air heater at low temper-
ature when the fuel contains a relatively large amount mSAH(HMS - HlMS) ) (CP80O mO2 + CP80N mN2)∆To -
2 2
of sulfur.
(CP15O mO2 + CP15N mN2)∆Ti (7)
3.1. Deaerator. A deaerator is used to separate 2 2
gases from BFW. In the deaerator, oxygen, hydrogen,
and CO2 dissolved in the water are removed by intro- Typical values of the air temperature are 15 °C at the
ducing steam (259 °C, 4 kg/cm2) into BFW. The gas- inlet and 80 °C at the outlet. The MS used in the
free water produced by the deaerator is fed into the steam-air heater is condensated and returns to the
steam generator, letdown units, turbine generator, and deaerator. Bunker-C oil is used as the fuel, and the flow
process units. Figure 4 shows a schematic of the rate of the combustion air is determined by the flow rate
deaerator. Simple mass balance on the deaerator gives of bunker-C oil. The composition of typical bunker-C
oil used in the plant considered here is shown in Table
mBFW + mPLT ) mWTR + mSAH + mDEA + mHPH (1) 2. By assuming x % excess air in the combustion, we
have
The water from the deaerator with a temperature of mSAH ) A′(x + 1)mBC (8)
132.9 °C and a pressure of 2 kg/cm2 is introduced into
the high-pressure feed-water heater (HPH) via the where
high-pressure water-feed pump. The water in HPH at
135.8 °C and 140 kg/cm2 is heated up to 195 °C and A′ )
130.4 kg/cm2 by MS and fed into the boiler. MS 3.1875(80CP80O - 15CP15O ) + 10.6112(80CP80N - 15CP15N )
2 2 2 2
consumed in HPH is condensed and is introduced into
the deaerator. HMS - HlMS
1054 Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998

Table 2. Composition of Bunker-C Oil


kg/kg of combustible
heat of
combustion required for combustion flue products
composition wt % (gross), kcal/kg O2 N2 air CO2 H2O N2
C 84.0 7828.24 2.66 8.86 11.53 3.66 8.86
H2 11.5 33898.28 7.94 26.41 34.34 8.94 26.41
S 4.0 2212.44 1.0 3.29 4.29 SO 2.00 3.29
N2 0.5
bunker-C 100.0 10562.52 3.1875 10.6112 13.7987 3.0744 (0.08) 1.0281 10.6112

Table 3. Parameters of Heat Capacity (J/kg °C) of Combustion Gases


component O2 CO2 H2O N2 SO2
A 28.11 19.8 32.24 31.15 23.85
B -3.680 × 10-2 7.344 × 10-2 1.924 × 10-3 -1.357 × 10-5 6.699 × 10-2
C 1.746 × 10-5 -5.602 × 10-5 1.055 × 10-5 2.680 × 10-5 -4.961 × 10-5
D -1.065 × 10-8 1.715 × 10-8 -3.596 × 10-9 -1.168 × 10-8 1.328 × 10-8

If we regard the boiler and the gas-type air heater as From these facts we have
a single unit, the energy required for BFW to be VS and
blowdown water should be supplied from the bunker-C (a + bx + HBC)mBC ∝ mVSHVS + mCBDHCBD +
oil and combustion air. Thus, we can write mCombCPts
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

∆H + (CPCO mCO2 + CPH OmH2O + CPN mN2 + In the normal operation the amount of blowdown
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

2 2 2
water is kept nearly constant. The flow rate of the stack
CPSO mSO2 + CPO mO2)∆T ) mBCHBC + (CP80CO mO2,i + gas can be represented in terms of the flow rate of
2
2 2
bunker-C oil.
CP80N mN2,i)∆T (9)
2 mCBDHCBD ) R (R ) constant)
The heat capacity of each gas component is given by (a + bx + HBC)mBC ∝ mVSHVS + c + (d + emBC)ts
2 3
CP ) A + BT + CT + DT Considering the efficiency of the boiler we finally have

Where values of A, B, C, and D for the main components mBC ) I + JmVS + (K + LmVS)ts
of combustion gases are given in Table 3. Rearrange-
ment of (9) by using the heat capacity relations gives From (11) and (12) the flow rate of bunker-C oil
required in the production of a given amount of VS can
mBCHBC - ∆H easily be computed.
) (B′ + C′x)mBC 3.4. Steam Distribution Network. In the steam
Ts distribution system shown in Figure 1, there are six
letdown units. As the amount of letdown increases, the
where loss of available energy increases and the overall energy
B′ ) 3.0744a + 1.0281b + 10.6161c + 0.08d - efficiency of the plant decreases. In order to decrease
3.1875f - 10.6112g the letdown and to avoid unnecessary vent, precise
adjustment of the amount of extraction through the
and steam turbine generator is imperative.
The steam turbine generator is driven by VS to
C′ ) 10.6112c + 3.1875e - 3.1875f - 10.6112g generate electrical power. From the steam turbine
generator, MS, LS, and steam condensate are released.
Introducing (6), we have from the above equation From simple mass balance we have
mBCHBC - D′mVS mVS ) mMS + mLS + mSC
) (B′ + C′x)mBC (10)
Ts
Mass balances on each steam header have to satisfy the
where following criteria:

D′ ) HVS + 0.01HBC - 1.01HBFW ∑


VS
m ) 0, ∑
HS
m ) 0, ∑
MS
m ) 0, ∑
LS
m)0
header header header header
In the actual plant operation we can use the following
relation without loss of generality: In the steam distribution system there are several
desuperheaters through which water is introduced to
mBCHBC + (1 + x)mairHair ∝ mVSHVS + mCBDHCBD + keep the temperature and pressure of each header
mCombCPts (11) constant. Mass and energy balances around the let-
down desuperheater give
From the fact that Hair is relatively constant, we can
say that the flow rate of the combustion air is propor- ∑i mi ) 0
tional to the flow rate of bunker-C oil:

(1 + x)mairHair ∝ (a + bx)mBC
∑i miHi ) 0
Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998 1055

Figure 6. Amount of bunker-C oil consumption.


Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

°C and that at the outlet is 80 °C. The consumption


rate shows a linear relationship as expected.
MS is used to maintain the temperature and pressure
of boiler feed-water at desired levels, and Figure 9
Figure 5. Computational procedure for steam generation. shows the amount of MS consumed for this purpose.
BFW extracted from the deaerator is fed into the boiler
Some chemicals such as ammonia and hydrazine via HPH. In order to remove gases solved in BFW, LS
(N2H4) are used to remove air in the boiler feed-water. is mixed directly with feed-water in the deaerator. The
But these chemicals may cause a change of the pH consumption rate of LS in the deaerator is shown in
value, which in turn results in damages to process units. Figure 10 as a function of VS generated.
Therefore, the boiler should be operated within the 4.2. Steam Distribution Process. To solve the
permissible range of water quality. To maintain the model equations, we used the well-known Newton’s
permissible range of water quality, saturated water is iteration method. The computational procedure is sum-
continuously withdrawn from the boiler drum. The so- marized in Figure 11. In simulations the amount of
called continuous blowdown rate reaches up to 1% of steam consumption at each process unit should be
VS produced and can be approximated by the relation known in advance in addition to the amount of ex-
hausted steam from turbine generators. Three typical
1
mFT ) mCBD (13) cases adopted from actual operation data were consid-
3 ered in the simulation and are summarized in Table 4.
The minus sign in HS consumption amounts means that
4. Simulation Results and Comparisons much HS steam is introduced from the heat exchanger
4.1. Steam Generation Process. Figure 5 shows at NCC units which are connected with steam distribu-
the flow diagram for boiler simulation. The mass flow tion networks. For each one of the three cases the
rate of bunker-C oil is assumed first and the enthalpy amount of letdown steam and VS produced in the boiler
of the outlet from the boiler is computed. The computed were computed and compared with corresponding op-
value of the stack gas temperature is compared with eration data. Results are shown in Table 5.
the measured value, and the mass flow rate of the From the results shown in Table 5 we can see that
deaerator is obtained finally. letdown amounts from each header are adjusted by
The amount of bunker-C oil to be supplied to produce simulation. As a result the amount of VS produced at
VS fed is plotted in Figure 6 as a function of VS to be the boiler is decreased compared to operation data. This
produced. During the combustion process in the boiler fact means that the overall energy efficiency of the
15% of excess air is maintained. steam distribution system is enhanced. The operational
Figure 7 shows the boiler efficiency as a function of cost of the utility unit is given by
VS generated. The boiler efficiency is given by
operational cost ) (cost of bunker-C oil) ×
mVSHVS + mCBDHCBD - mBFWHBFW ∆H
eff ) ) (bunker-C fuel to generate steam) +
mBCHBC mBCHBC (cost of industrial water)(BFW to generate steam)
(14) (15)
As can be seen, the computed efficiency is about 88%.
Figure 8 shows the consumption rate of MS to preheat As seen in (15), the operational cost of the utility plant
combustion air as a function of VS produced. The is composed of the cost of bunker-C oil and that of
temperature at the inlet of the steam-air heater is 15 industrial water, which are the main factors in reducing
1056 Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998

Figure 7. Boiler efficiency. Figure 9. Amount of steam consumption in HPH.


Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

Figure 10. Amount of steam consumption in the deaerator.


Figure 8. Amount of steam consumption in SAH.

using the class and object editors provided by the


operational cost because they are heavily influenced by
NEXPERT Object, we could conveniently express pro-
the VS generation, i.e., the amount of steam generated
cess equipments with related properties. For example,
by the boiler.
if we take a turbine generator as an object, the amounts
By using (15), we can compare the actual operational of steam extraction and electric power consumption are
cost with the cost obtained from simulations, and the defined as properties of the object.
results are shown in Table 6 for each case. From the The expression of the performance of the turbine
results we can see that VS generation in each case is generator obtained from the related model can also be
decreased noticeably (Table 5) and the decrease of VS incorporated in the rule using the method provided by
generation caused a reduction of the operational cost NEXPERT Object. The expression consists of a se-
about 2-5%. This fact means that the unnecessary vent quence of actions under certain conditions during infer-
of steam in the utility equipment could be reduced in ence processes. The following shows an example of
great amount by the exact estimation of the amount of knowledge expressions:
steam required in each header and equipment through
simulations. Reduction of the vent of steam is directly OBJECT
related to the decrease in the operational cost, as can The turbine generator
be seen in Table 6. PROPERTY
The amount of electric power generation
5. Construction of the Knowledge Base The amount of SC extraction
The amount of LS extraction
5.1. Expression of Process Equipments. We The amount of MS extraction
employed NEXPERT Object (version 3.01, Neuron Data) The amount of VS consumption
as a construction tool to build the knowledge base. By method
Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998 1057

Table 4. Three Typical Operation Cases


item case I case II case III
VS consumption in process units (ton/h) 170.0 175.0 175.0
HS consumption in process units (ton/h) -104.7 -132.0 -130.0
MS consumption in process units (ton/h) 167.3 201.6 221.8
LS consumption in process units (ton/h) 51.8 74.2 75.9
steam consumption at the air dryer (ton/h) 0.3 0.3 0.3
steam consumption at the flare stack (ton/h) 1.0 1.0 1.0
steam consumption at MISC (ton/h) 4.0 4.0 0.0
electricity produced at the turbine generator (MW) 24.5 25.5 25.5
LS amount at the turbine generator (ton/h) 12.0 20.0 26.0
SC extraction at the turbine generator (ton/h) 41.0 28.0 28.0

Table 5. Results of Simulation Based on Raw Operational Data


case I case II case III
amount of steam(ton/h) data simulation data simulation data simulation
letdown from the VS header 6 54. 9 3 53.9 8 86.3
letdown from the HS header 2 1.52 0 7.48 6 4.6
letdown from the MS header 7 2.41 0 4.76 0 2.39
VS produced at the boiler 440 423.3 453 443.75 505 479.7

Table 6. Comparison of Operational Cost (thousand s


W/h)
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

case I case II case III


Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

data simulation data simulation data simulation


operational cost 6748.7 6494.1 6946.9 6805.9 7739.6 7353.9

IF the electricity supplied from the power company is not


THEN acceptable.
Compute VScon Even with the low efficiency, a change of the amount
of the outlet from the letdown desuperheater has little
If the expert system requires the amount of steam
effect on the other utility equipment. Due to the ease
consumption of the turbine generator during the infer-
in maneuvering desuperheaters, this method is recom-
ence processing, the method part of the object is trig-
mended in executing operations of steam distribution
gered. Because the IF clause has no condition, the
at first.
method is executed whenever the object is called.
In the case of a steam turbine, the amounts of steam The driving force of the utility motor is electric power
consumption and electric power consumption of a motor generated by the turbine generator. This autogenerat-
are constant. In this case, the constant values can be ing system has inherent constraints in switching from
initialized by the meta-slot as the utility pump to the motor because the turbine
generator has a chance of being overloaded with the
OBJECT increase of electricity. Thus, a change of the driving
The steam turbine force in the utility pump for steam distribution should
PROPERTY be preceded by adjustments of the turbine generator. If
The amount of steam consumption the method described above is not enough to adjust the
meta-slot amount of demand in the steam distribution network,
Initial value: constant we finally change the amount of VS generation by the
boiler.
5.2. Expression of Operational Heuristics. The If demands of steam grade are varied, we adjust the
steam distribution system should be activated if there utility equipment in the order of letdown desuperheater,
are changes in steam demands from process units. The turbine generator, utility pump, and boiler. Moreover,
steam distribution system adjusts vibrations in the steam distribution is always accomplished from low-
amount of steam consumed in process units and fulfills pressure steam header to high-pressure steam header.
steam needs due to changes in operating conditions of A summary of the heuristics concerning steam distribu-
utility equipmentsboiler, utility pumps, turbine gen- tion at each header can be expressed as follows:
erators, and letdown desuperheaters.
The final objective of the present work is to develop LS steam header:
a steam distribution system for on-line use at the plant, (a) Letdown of the desuperheater from the MS steam
but presently the system developed is being tuned and header to the LS steam header
used off-line at the plant. Thus, if demands of a specific (b) Low-pressure extraction of the turbine generator
steam grade are varied or the utility equipment are out (c) Change of the utility pump
of order or shutdown for maintenance, then the user
(d) Letdown of the desuperheater from the VS steam
activates the steam distribution system to get appropri-
header to the LS steam header
ate operational commands or information. The graphi-
cal user interface developed by using the Open Interface MS steam header:
(by Neuron Data) tool makes it very easy and convenient (a) Letdown of the desuperheater from the HS steam
for the user to use the system developed. The use of header to the MS steam header
the steam distribution system is also required when the (b) Medium-pressure extraction of the turbine genera-
source of electrical power should be varied due to tor
changes of operating conditions or when the quality of (c) Change of the utility pump
1058 Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

Figure 12. Rule base hierarchy of LS demand change.

Figure 11. Computational procedure for the steam distribution


system.

(d) Letdown of the desuperheater from the VS steam


header to the MS steam header
HS steam header:
(a) Letdown of the desuperheater from the VS steam
header to the HS steam header
(b) MS steam generated by the boiler
VS steam header:
(a) VS steam generated by the boiler
According to the hierarchy of operational heuristics
for steam distribution, the computational procedure and
hierarchy of rule base are shown in Figures 11-15.
Following is an example of a rule when the requirement
of LS is changed.
IF the requirement of LS decreases
AND it is possible to finish the steam distribution of
the LS header by diminishing the amount of the
outlet of the letdown desuperheater 3
THEN decrease the outlet of the letdown desuperheater
3
If the amount of LS required decreases and the Figure 13. Rule base hierarchy of MS demand change.
letdown desuperheater 3 could reduce steam outlet flow
as much as the amount of LS decrement in the process, AND it is impossible to accomplish the steam distribu-
the system reduces the downstream of the letdown tion of the LS header by reduction of the amount
desuperheater 3. In general, the minimum amount of of the letdown desuperheater 3
outlet flow of letdown desuperheaters is 5% of the AND it is possible to accomplish the steam distribution
maximum value. Therefore, reduction of the outlet of the LS header by diminishing the amount of
cannot meet the requirement of LS in some cases. In outlet of the letdown desuperheater 3 and the
these cases, the low-pressure extraction of a turbine low-pressure extraction of the turbine generator
generator must be controlled. The following rule de- THEN decrease the outlet of the letdown desuperheater
scribes this operation: 3
AND decrease the low-pressure extraction of the tur-
IF the requirement of low-pressure steam decreases bine generator
Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998 1059

letdown desuperheater 3 and the low-pressure


extraction of the turbine generator
AND the user selects the pumps to change their
driving forces
THEN decrease the outlet of the letdown desuperheater
3
AND decrease the low-pressure stage extraction of the
turbine generator
AND change the driving forces of the utility pumps
AND adjust the letdown desuperheater 6
In order to adjust the letdown desuperheater 6, the
system examines if the letdown desuperheater 6 can
accomplish the steam distribution. In this case, a
method is defined to check the conditions and to
terminate the steam distribution as shown below.
OBJECT
The letdown desuperheater 6
PROPERTY
The amount of outlet steam
method
IF the steam requirement decreases
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

AND it is possible to meet the demand of steam


Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

by the letdown desuperheater 6


THEN decrease the outlet of the letdown desuper-
Figure 14. Rule base hierarchy of HS demand change. heater 6
In the computation of increase of the boiler load by
the self-consumed steam, much iteration is needed and
the following rule should be used repeatedly to search
the knowledge base in predefined order.
IF the steam distribution of the LS header is exe-
cuted
AND the steam distribution of the MS header is exe-
cuted
AND the steam distribution of the HS header is exe-
cuted
AND the steam distribution of the VS header is exe-
cuted
AND the boiler load does not converge
THEN update the amount of self-steam consumption
AND reset all of the rules

6. Steam Distribution Based on the Heuristic


Knowledge Base
The expert knowledge base system developed in the
present work is utilized to perform steam distribution
operation. Construction of the knowledge base was
described in the previous section. Using NEXPERT
Object, we create objects for process equipment with
related properties by the object editor as mentioned
Figure 15. Rule base hierarchy of VS demand change.
before. To perform inference processing, various process
conditions and hierarchical rules as shown in Figures
In the plant considered in the present work the 11-15 are evaluated step by step. In the inference
minimum flow rate of low-pressure extraction is usually processing, hypotheses and conditions based on a rel-
maintained at 12.0 ton/h. If the supply cannot meet the evant knowledge base are examined, and opportune
demand by decreasing the extraction down to the actions described by rules are executed and triggered
minimum, we must change the driving forces of the to examine subsequent conditions of other rules. The
utility pumps. Because a steam turbine always con- computational procedure for calculating mass balances
sumes a fixed amount of steam, a letdown desuper- at each steam head is incorporated in the rules. Lists
heater 6 must always be adjusted in order to meet the of hypotheses or conditions in a queue were constructed
demand of the process exactly. This fact can be sum- in a sequential fashion by using the method of rule
marized as follows: chaining provided by NEXPERT Object. Based on the
methodology described so far, we have constituted about
IF the requirement of LS decreases 1500 rules containing the heuristics of experienced plant
AND it is impossible to meet the requirement of LS operators and practical operation knowledge accumu-
by reducing the amount of the outlet of the lated in the plant.
1060 Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998

Table 7. Operational Status of the Utility Plant (I)


operational operational
status status
process equipment (ton/h) process equipment (ton/h)
VS boiler 441.0 letdown desuperheater 6 0.0
HS boiler 0.0 turbine generator
HPH 57.33 steam consumption 180.0
SAH 16.758 MS extraction 131.0
deaerator 40.8807 LS extraction 19.0
oil heater 1.7523 condensate extraction 30.0
steam atomizer 4.6611 electric generation (MW) 23.0
flash tank 2.6667 VS user 195.0
letdown desuperheaters 1 and 4 71.9997 HS user -170.1926a
letdown desuperheater 2 9.08225 MS user 208.22715
letdown desuperheater 3 5.852 LS user 70.6837
letdown desuperheater 5 0.0
a Because much steam is generated in the heat exchanger at the NCC unit, the amount of steam consumption is negative.

Table 8. Operational Status of the Utility Plant (II)


process operational process operational
equipment status (ton/h) equipment status (ton/h)
cooling water pump 3T1Ma air compressor 1T1M
decarbonated water pump 1T1M oil pump
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

polished water pump 1T B-C oil pump 1T


Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

forced draft fan 3T1M PFO pump not operated


boiler feed water pump 2T2M C9-C10 pump not operated
induced draft fan 1T3M
a Four pumps worked by three steam turbines and one motor.

Table 9. Operational Range of Process Equipment


operational operational
process equipment range (ton/h) process equipment range (ton/h)
VS boiler 0-550.0 letdown desuperheaters 1 and 4 0.0-171.0
HS boiler 0-10.0 letdown desuperheater 2 0.0-14.959
turbine generator letdown desuperheater 3 0.0-8.512
steam consumption 35.0-233.5 letdown desuperheater 5 0.0-190.0
MS extraction 10.0-186.0 letdown desuperheater 6 0.0-100.0
LS extraction 12.0-99.0
condensate extraction 13.0-45.0
electric generation 10.0-30.0
When the expert steam distribution system is acti- Table 10. Design Data of the Steam Turbine and Motor
vated, the system first computes the mass balances of
driving steam turbine motor
four steam heads using operational data. The rules in process equipments force (ton/h) (kW)
the knowledge base containing overall computational
cooling water pump 4T4M 50.268 1770.0
procedures examined which of the balances are varied
decarbonated water pump 1T2M 2.13 90.0
or require more [less] steam. If steam distribution polished water pump 1T1M 7.59 220.0
operation is needed, the NEXPERT Object put a hy- forced draft fan 4T4M 8.12 560.0
pothesis (for example, the increase of the LS header boiler feed water pump 4T2M 17.298 1250.0
steam demand) on the inference engine processing induced draft fan 4T4M 4.6 250.0
performed by volunteering data or suggesting a hypoth- air compressor 2T2M 9.612 540.0
oil pump
esis directly. Then the agenda mechanism performs to B-C oil pump 2T1M 2.34 45.0
interpret the evaluation process and to govern how and PFO pump 1T1M 1.95 60.0
when subsequent rules should be executed. C9-C10 pump 1T1M 0.889 22.0
The operational status of the actual utility plant
studied in the present work is shown in Tables 7 and 8. Table 11. Amount of Change of Steam Demand
The operational ranges of process equipment are shown steam grade current status (ton/h) future status (ton/h)
in Tables 9 and 10. VS 195.0 206.0
Consider the steam distribution when the require- HS -170.1926 -145.1474
ment of steam demand is changed as shown in Table MS 208.22715 214.8137
11. Steam distribution is always accomplished from LS 70.6837 74.0128
low-pressure steam headers to high-pressure steam AND it is possible to finish the steam distribution of
headers. Because a letdown desuperheater 3 has a very the LS header by increasing the amount of the
narrow operational range, the desuperheater and a letdown desuperheater 3 outlet and the extrac-
turbine generator perform steam distribution. This tion of a low-pressure stage of the turbine
operation is executed by the following rule. generator
IF the requirement of LS increases THEN increase the outlet flow of the letdown desuper-
AND it is impossible to meet the steam demand by heater 3
increasing the outlet of the letdown desuper- AND increase the extraction of a low-pressure stage
heater 3 of the turbine generator
Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998 1061

In order to distribute steam to the MS header, the Table 12. Comparison with Steam Distribution
amounts of extraction from the high-pressure stage of Results (I)
the turbine generator and from the letdown desuper- result by
heater 2 are adjusted because the operational range of operational expert system
the letdown desuperheater 2 is narrow, as shown below. process equipments status (ton/h) (ton/h)
VS boiler 498.0 486.95
IF the requirement of MS increases HS boiler 0.0 0.0
AND it is impossible to finish the steam distribution HPH 64.74 63.3
of a MS header by increasing the outlet of the SAH 18.924 15.5
letdown desuperheater 2 deaerator 46.1646 45.14
AND it is possible to finish the steam distribution of oil heater 1.9793 1.94
a MS header by increasing the amount of the steam atomizer 5.2651 5.15
flash tank 2.667 2.667
outlet of the letdown desuperheater 2 and ex- letdown desuperheaters 1 and 4 102.5446 102.52
traction of the high-pressure stage of the turbine letdown desuperheater 2 14.959 14.96
generator letdown desuperheater 3 8.512 8.51
THEN increase the outlet flow of the letdown desuper- letdown desuperheater 5 0.0 0.0
heater 2 letdown desuperheater 6 0.0 0.0
AND increase the extraction of the high-pressure turbine generator
steam consumption 198.0 187.42
stage of the turbine generator high-pressure stage extraction 144.0 142.08
The following rule can be used in the case of increase low-pressure stage extraction 25.0 23.96
condensate extraction 29.0 21.78
of HS demand. electric power (MW) 22.0 22.0
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

IF demand of HS increases
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

Table 13. Comparison with Steam Distribution Results


AND it is possible to meet the steam demand by the (II)
letdown desuperheaters 1 and 4
THEN increase the amount of the outlet of the desu- result by
operational expert system
perheaters 1 and 4 process equipments status (ton/h) (ton/h)
If the requirement of VS demand increases, the expert cooling water pump 3T1M 3T1M
system achieves steam distribution by controlling the decarbonated water pump 1T1M 1T1M
production rate of the main steam boiler. polished water pump 1T 1T
forced draft fan 3T1M 3T1M
IF the amount of required VS steam increases in boiler feed water pump 2T2M 2T2M
process units induced draft fan 1T3M 1T3M
AND it is possible to meet the steam demand by air compressor 1T1M 1T1M
oil pump
increasing the load of the main steam boiler B-C pump 1T 1T
THEN increase the production rate of the main steam PFO pump not operated not operated
boiler C9-C10 pump not operated not operated
It is obvious that the steam consumption of HPH,
SAH, and the deaerator, which cover most of the self- network were performed. Newton’s iteration method
steam consumption, should be varied according to the and a simple linear programming algorithm were
load of the boiler. In general, the amount of self-steam employed in the simulation. To minimize the net cost
consumption does not match the amount required when of providing energy to the plant, heuristic operational
the load of the boiler increases. The amount of self- knowledge and steady-state models developed were
steam consumption should be updated and the outlet incorporated in the construction of an expert system for
steam of the utility equipment should be redistributed the steam distribution network. In the applications of
again from the LS header until the load of the boiler is the present steam distribution system for the steady-
converged. state operation status, the letdown amount from the
During the inference processing, it is necessary to find very high pressure steam (VS) header and the amount
out new steam demand of self-steam consumption for of VS produced at the boiler showed good agreement
the equipment (HPH, SAH, deaerator, etc.) subjected with those of actual operational data. Three typical
to the boiler. So, when the knowledge base needs new operation cases were considered in the test applications.
data for self-steam consumption, the NEXPERT Object For each one of these cases the amounts of letdown
is again supplied with the simulation results shown steam and VS produced in the boiler were computed and
before. compared with actual operational data.
The results of steam distribution by the system A more economical steam distribution was accom-
developed in the present work are summarized in Tables plished by using the present expert system compared
12 and 13. As can be seen, the results well agreed with to the prior steam distribution operations. The predic-
the actual operation data. This fact means that present tion of the increment of boiler load by self-consumed
operations are well within the optimal operational steam improved the operation of the steam generation
range. We can expect that the steam distribution unit.
system by the present study may give dependable
operational guidelines even when there are unexpected Nomenclature
sudden changes in operating conditions.
CP ) heat capacity [J/kg °C]
7. Conclusions eff ) efficiency
Steady-state modeling and simulation of the steam H ) enthalpy [J/kg]
generation process and the plantwide steam distribution m ) mass flow rate [kg/h]
1062 Ind. Eng. Chem. Res., Vol. 37, No. 3, 1998

ts ) real stack gas temperature [K] l ) saturated


T ) temperature [K]
Ts ) simulated stack gas temperature [K] Literature Cited
x ) excess air [%]
a, b, c, d, e, f, g, A′, B′, C′, D′, I, J, K, L ) constant Clark, J. K.; Helmick, N. E. How to Optimize the Design of Steam
Systems. Chem. Eng. 1980, March, 116.
Subscripts Gordon, E.; Hashemi, M. H. A Versatile Steam Balance Program.
air ) combustion air Chem. Eng. Prog. 1978, July, 51.
Maia, L. O. A.; Vidal de Carvalho, L. A.; Qassim, R. Y. Synthesis
BC ) bunker-C oil of Utility Systems by Simulated Annealing. Comput. Chem.
BFW ) boiler feed-water Eng. 1995, 19, 481.
CBD ) continuous blowdown Nishio, M.; Itoh, J.; Shiroko, K.; Umeda, T. A Thermodynamic
Comb ) combustion Approach to Steam-Power System Design. Ind. Eng. Chem.
DEA ) deaerator Process Des. Dev. 1980, 19, 306.
FT ) flash tank Nishio, M.; Shiroko, K.; Umeda, T. Optimal Use of Steam and
HPH ) high-pressure feed-water heater Power in Chemical Plants. Ind. Eng. Chem. Process Des. Dev.
i ) inlet 1982, 21, 640.
o ) outlet Nishio, M.; Koshijima, I.; Shiroko, K.; Umeda, T. Synthesis of
Optimal Heat and Power Supply Systems for Energy Conversa-
LS ) low-pressure steam
tion. Ind. Eng. Chem. Process Des. Dev. 1985, 24, 19.
MS ) medium-pressure steam Petroulas, T.; Reklaitis, G. V. Computer-Aided Synthesis and
PLT ) plant feed-water Design of Plant Utility Systems. AIChE J. 1984, 30, 69.
SAH ) steam-air heater
SC ) condensate Received for review March 12, 1997
VS ) very high pressure steam
Published on February 4, 1998 on http://pubs.acs.org | doi: 10.1021/ie970214y

Revised manuscript received November 24, 1997


WTR ) raw water
Downloaded by NORTH CAROLINA STATE UNIV on August 5, 2009

Accepted November 25, 1997


Superscript IE970214Y

Anda mungkin juga menyukai