(246b) Optimal Synthesis for the Feed-Water-Heater Network of a Pulverized Coal (PC) Power to Minimize Water Consumption

1. Introduction

Coal-fired power plants contribute to almost 50% of the
United States' total electric power production. At the same time, pulverized
coal (PC) power plants are large water consumers to the point that construction
and operability of PC power plants have started to be constrained by water
availability in some regions of the country. Two process systems engineering
approaches have been studied to minimize the water consumption of power plants.
First, a better optimization of nonlinear uncertain systems (BONUS) algorithm
has provided solutions to the minimization of water consumption under uncertain
atmospheric conditions. The calculated conditions for a 550 MW PC plant
predicted reductions of 6.4%, 3.2%, 3.8% and 15.4% in the average water
consumption for the four different seasons from fall to summer respectively. A
second approach pursues the reduction of the residual heat in the steam cycle
of the PC process by formulating an optimal design of the feed water heat
exchange network (HEN). The proposed methodology uses Aspen Energy Analyzer
(AEA) to determine the mass flowrates of the bleeding
streams while generating alternative designs that can potentially reduce the
water consumption by reducing the total cooling requirement. The optimization
approach to process synthesis involves the selection of an optimal solution
from the superstructure with a simulated annealing (SA) capability built in
Aspen Plus. Results show at least a 5% reduction in water consumption for the
PC power plant via this enhanced HEN synthesis technique

2. Process Description

The PC power plant model studied in this paper is based
on Case 11 referenced by the

DOE/NETL's report on cost and performance of fossil
energy plants [1] . This power plant is a re-heating cycle with feed-water heating. The
process comprises a boiler section where coal is burned and the combustion heat
is transferred to water generating steam, and a steam section where the high
pressure steam is expanded through a train of high, intermediate and low
pressure turbines. Steam is initially generated in the boiler and expanded in
the high pressure (HP) turbine; then, it is re-heated at the boiler for later
expansion at the intermediate (IP) and low pressure (LP) turbines. Then,
exhausted steam is condensed with liquid water and returned to the boiler while
heated with bleeding streams of steam from the turbines (feed-water heater).
Cooling water is sent to the cooling tower where heat from the cycle is
ultimately rejected to the environment by evaporative cooling.   

3. Optimization under uncertainty

3.1. Stochastic Simulation

As it was
previously said, water consumption is strongly affected by weather conditions like
air temperature and humidity. Detailed information about air conditions for
different locations within the US can be obtained from the website (http://www1.eere.energy.gov). Histograms of the frequency distributions
were fitted to the appropriate probability density functions to represent the
air conditions variability of an average U. S. Midwestern urban center during
each of the four seasons [7] . Stochastic simulation
is carried out by efficiently sampling these distributions to generate a set of
scenarios under which the plant models are evaluated and a corresponding probability
distribution of water consumption (output variable) is calculated.

3.2. Stochastic Optimization

Minimization of water consumption in power plants is a
stochastic non-linear programming (SNLP) problem where one of the moments (expected
value, standard deviation, etc.) of the water consumption's probability distribution
is the objective function, the model is the set of constraints and model
parameters are the decision variables. BONUS algorithm [4]  is a ?here and now? approach to solve this
problem [8]  determining the process
conditions under which the expected value of the water consumption can be
reduced compared to the base case originally reported, as shown in  REF _Ref319918871 \h Table 1 08D0C9EA79F9BACE118C8200AA004BA90B02000000080000000E0000005F005200650066003300310039003900310038003800370031000000
[9]
.  It can be seen that the reduction on
the consumption strongly depends on the season consumption during drastic
weather conditions as summer can be reduced by changing some of the parameters
of the process.   

Table  SEQ
Table \* ARABIC 1 Minimization of average water
consumption under uncertain air conditions for a 550 MW PC power plant located
at Midwestern US for four different seasons [9] .

4. Heat Exchange Network synthesis for
water management

4.1. Modified HEN synthesis in AEA

Conventional methodology for the heat exchange network
synthesis of the feed water heating system is based on equal enthalpy change on
the liquid stream for each of the heaters. Mass flow rates of the bleeding
streams are calculated based on such heat load distribution. AEA generates
alternative designs with a mixed integer linear programming (MILP) algorithm.
The solution of the MILP problem is based on the thermodynamic characteristics
of the involved streams and on fixed mass flowrate of the bleeding
streams.  This approach was modified
based on the early work by Linhoff [5]  that proposes the employment of pinch
technology to define mass flowrates from the bleeding streams for maximum heat
recovery and improved cycle efficiency. This cycle efficiency is directly
associated to the PC process water consumption through the amount of heat
rejected by the cycle.

The main idea of this work is employing the AEA MILP
algorithm to determine the mass flowrates of the bleeding streams while
generating alternative designs that can potentially reduce the water
consumption by reducing the total cooling requirement. For this purpose, the
bleeding streams have been treated as utilities (instead of process streams)
whose mass flowrates are determined by the calculated heat load assigned to
them by the MILP algorithm.

4.2. Cost modification for alternative
utility streams

Optimization algorithms based on process costs (as the
one used by AEA) yield designs that employ mostly hot streams because they
minimize heat transfer area leaving low temperature streams unused. To avoid
this situation a cost penalty was assigned to bleeding (utility) streams. The
approach assumes that feed water heating is expensive for the process when
using high pressure steam and using low pressure steam is inexpensive to the
point that some income can be generated. The main reason is that employing
large amounts of high pressure steam may decrease the process productivity and
preheating the feed water with low pressure steam will increase the process
efficiency. The drain streams are mixed and also are used to transfer sensible
heat to the feed water. The costs assigned to these drains are naturally
related to utility streams.   On the
other hand the remaining drain streams are mixtures of two or three drains.
Therefore their costs have been assumed to be a weighted average of their
bleeding constituents having more weight the cost of the bleed that is more
expensive. This formulation of the problem and its solution with AEA yielded
interesting alternative designs whose cold utility consumption is lower than
that of the base case as shown in  REF _Ref314440794 \h Table 2 08D0C9EA79F9BACE118C8200AA004BA90B02000000080000000E0000005F005200650066003300310034003400340030003700390034000000
and
it is expected that the water consumption will also be lower.

Table  SEQ
Table \* ARABIC 2 Cooling requirements for designs
that consider the bleed streams as utilities

5. Conclusion

This paper presents a framework for integrated water
management in power systems.  It has been
found that water consumption depends on the weather conditions and the
uncertainties in weather affect the consumption considerably.  We presented an optimization under
uncertainty problem for savings water consumption. The second approach is heat
integration.  An algorithmic framework
(Figure 1) based on simulated annealing (and OA/ER/AP MINLP) for discrete
decisions, BONUS algorithm for the stochastic NLP, and Hammersley
sequence sampling for the sampling saved 97% of computational time to solve
this problem. Optimization under uncertainty allowed us to save up to 15% in
the expected value of water consumption in a PC plant and a pinch technology
approach to the heat exchange network synthesis allowed a 38% in cooling
requirement which will be translated in water savings as well.

Figure 1:
Algorithmic Framework

References

[1]DOE/NETL,
Cost and performance baseline for fossil energy plants, (2007).
DOE/NETL-2007/1281.

[2]DOE/NETL,
Power plant water usage and loss study, (2007).

[3] U.
Diwekar and E.S. Rubin, Comput. Chem. Eng. 15 (1991) 105-114.

[4] K.
Sahin and U. Diwekar, Annals of Operations Research 132 (2004) 47-68.

[5] B.
Linhoff and F.J. Alanis, ASME Advanced Energy Systems 85 (1989) 10-15.

[6] J.M.
Salazar, U.M. Diwekar and S.E. Zitney, Comput. Chem. Eng. 35 (2011) 1863-1875.

[7] J.M.
Salazar, U.M. Diwekar and S.E. Zitney, Energy Fuels 24 (2010) 4961-4970.

[8] U.M.
Diwekar, Introduction to Applied Optimization, 2nd ed., SpringerLink, New York,
2008.

[9] J.
Salazar and U. Diwekar. Energy Systems 2 (2011) 263-279.