Sei sulla pagina 1di 6

2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004)April 2004 Hong Kong

A Test System for Combined Heat and Power


Economic Dispatch Problems
Cameron Algie and Kit Po Wong, Fellow, IEEE

Abstract-Cogeneration and Independent Power and steam turbines in early 1900s saw cogeneration and
Producers are playing an increasing role in deregulated IPP fall increasingly out of favour for residential use,
electricity supply markets. The scheduling of combined heat being replaced by boilers and utility-supplied power. By
and power outputs from multiple-generator cogeneration the 1940s IPP cogeneration was only suitable for the
plants is a complicated problem that needs powerful largest industrial power users and users of process heat.
methods to solve. Representative test systems are required to
test any new method in the development of new algorithms.
This paper presents a test system and a set of solutions for Interest in cogeneration was renewed in the early 1980s
demand cases. The proposed new test system could be used following the PURPA reform act in 1978. Initially the
as a standard test system for assessing cogeneration focus was on more efficient use of fossil fuels and the
economic dispatch algorithms. issues investigated were primarily economic feasibility
studies and case studies for conversion of existing plant to
Index Terms-Cogeneration, Combined Heat and Power, cogeneration. Throughout the 1980s technological
Economic Dispatch Evolution Computation, Independent advances in gas turbine components and an increased
Power Producers, Test System supply of natural gas lead to IPP gas turbine cogeneration
and combined cycle plants becoming economically viable
1. NOMENCLATURE alternatives to utility supplied electricity. The efficiency
of single cycle gas turbines is now comparable (or better)
Pi Active Power MWe, than coal-fired steam turbine plant, and combined cycled
hj Thermal Power (Steam) MWth gas turbines are approaching 60% thermal efficiency.
C,(pi) Generation Unit Cost Economic scheduling strategies for supplementing utility
Cj(hj) Thermal (Boiler) Unit Cost supplied power and independently meeting internal power
Ck(pk,hk)Cogeneration Unit Cost needs from on-site cogeneration [1]-[5] continued to be
Gena An Electricity Unit in a Cogeneration System investigated.
Cogp A Cogeneration Unit in a Cogeneration System
Boil, A Process Steam Unit in a Cogeneration System However, unlike the traditional, electricity-only
H-P Heat(MWthtPower(MWe) Cogeneration Demand economic dispatch (ED) problems, cogeneration ED
problems do not have a standard set of test problems for
11. INTRODUCTION assessing algorithms. The demonstration system and
demand case used by Guo, Henwood and van Ooijen [ 11,
Cogeneration, or combined heat and power (CHP), and has become a defacto standard test case because: a) it is a
Independent Power Producers (IPP) have had a role in cogeneration system, and; b) it is to-date the only test
industrial and residential power supply and heating since system published in the literature. This test system has a
the inception of commercial power supply in the 1880s. number of strong points, especially concerning the
The small capacity and short transmission distance modelling cogeneration unit Feasible Operating Regions,
achievable from DC power generation meant that however it has inherent weaknesses that make it
electricity users were forced to a certain extent to utilise unsuitable as a “standard” test system for algorithm
waste heat from the generators’ coal furnaces for heating development.
purposes. The increased transmission distances and
capacity increases made possible by AC power generation As part of the development of an Evolution
Computation (EC) algorithm for solving IPP cogeneration
C. Algie obtained his Master of Engineering Science from the School ED [ 5 ] , a new test system has been designed by the
of Electrical Electronic & Computer Engineering, The University of authors. The new system augments the old system model
Western Australia in 2003 and now works for CSR Sugar.
(calgie@graduate.uwa.edu.au)
from [ l ] with new generation unit models. In this paper,
the strengths and weaknesses of the old test system are
Professor K.P. Wong is Chair and Head of the Department Electrical discussed and the new test system is described. The
of Engineering, Hong Kong Polytechnic Univcrsity, as well as member solutions of three heat-power demand cases solved by the
of Staff of the School of Electrical Electronic & Computcr Engineering, authors’ EC algorithm are presented for the purposes of
The University of Wcstcm Australia.
demonstrating the different types of dispatch solutions
that the new test system can provide.

0-7803-8237-4/04/$17.0002004IEEE
96
2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DWT2004) April 2004 Hong Kong

111. COMBINED HEAT AND POWER ECONOMIC


DISAPTCH PROBLEMS

Scheduling problems entail finding the generators, and


individual generator outputs, required to meet power
demand over a period. Economic means that the settings
are to be chosen to produce the minimum possible cost. In
the most basic problem, static economic dispatch, a group
of generators, which all have to be used, meet a set
demand for one scheduling interval. The CHP problem is
more complicated as it has two types of generation unit
output and demands to simultaneously meet (heat/ steam o Heat ( M W t h ) i
and active power). Also, CHP systems can include three
types of generation units: cogeneration (heat and power),
power-only generators and heat-only boilers. A generic Fig. 1. Cogeneration Unit FOR
CHP solution is shown in Table I, where each generation
unit is subject to output constraints, as described by (la), Like individual cogeneration units, cogeneration
(Ib), (IC)and (Id). systems also have allowable sets of heat-power demands
that it can meet. By combining cogeneration units, FOR
models, boiler heat output ranges and generator power
TABLE I output ranges, the system Feasible Demand Region (FDR)
REPRESENTATION OF A CHP ED SOLUTION
is formed. A typical FDR of a system with auxiliary heat
and power generators is shown in Fig. 2 .
Gen, Gena Cog,, Cogp Boilpl Boil,
PI ... Pa Pa+] ... PO
ha+l ... he hb+l ... hn

To be considered a viable solution, the power and heat


demands have to be met, as per the equality constraints of I . . . . . . . . . . , . .
(2a) and (2b). 0 Heat (MWth)

P I + p2 + ... + Pa = Power Demand (2a) Fig. 2. Cogeneration System FDR


ha+l + ha+?+ . . . + h, = Heat Demand (2b)
VI. CHARACTERISTICS OF SOLUTIONS
Generation unit cost functions (heat characteristics), C,
- C,, are used to form an objective function, i.e. (3), that is Optimal solutions for 25 heat-power demand cases
used to assess the strength of the solution. In ED problems, within the new test system’s FDR were found. Initially
the optimal generator settings will produce the lowest based on the ability of the author’s EC algorithm [ 5 ] to
possible operating cost. find solutions to demand cases, the demand cases were
classified as either simple or difficzilt. However, further
investigation was able to identify the underlying
Cl(pl) +...+ Cu(pu) + Cu+l(Pa+lha+l)+...Cp(~php)
+Cp+,(hp+,) +.. .+ C,(h,) = Total Cost conditions that are independent of the test application and
(3) the solution technique used.
To date in the field of CHP research, only methods for
The reasons for the simple case being solved relatively
solving the static economic dispatch problem have been
easily are: 1) that some generation units are passive to the
reported.
solution process (by being at maximum or minimum
output), effectively reducing the number of variable to
The constraint inequalities of (lb) and (IC) are used to
solve, and I or; 2) that the dollar cost of the optimal
describe the Feasible Operating Region (FOR) of
solution is not overly sensitive to generator outputs. There
cogeneration units, i.e. the area in a two axis diagram that
contains all possible heat-power outputs from a given will still only be one true minimum, but if this is only
cogeneration unit. A typical FOR is an irregular distinguished from other solution costs in terms of tenths
or hundredths of cents, outputs can vary slightly from the
quadrilateral, as shown in Fig. 1.
optimal setting and still return the “minimum” cost. A
generic simple solution is demonstrated in Table 11,

97
2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004)April 2004 Hong Kong

showing passive variables, Units 1 and 4, and a minimum generators available to meet the remaining demand. In
cost that is insensitive to a small variation, 6, in the output practical terms, a test system for cogeneration ED
of Units 2 and 3. problems should consist of two or more generation units.
By including two different cogeneration unit models, a
boiler and a power generator, the effects of these two
auxiliary units can also be taken into account.
Unit 1 Unit 2 Unit3 Unit4 Cost The cogeneration units used are of significantly
Min A B Max $min.xxxy different maximum capacities; one is 247 MWe of
Min A+6 B-6 Max $min.xxxz electricity and 180 MWth of in steam, the other 125.8
MWe and 135.6 MWth. This corresponds to modelling a
Difficult cases are characterised by extreme variable new smaller gas-turbine unit (125.8 MWe unit) being
sensitivity and the involvement of all generators in the added to increase power and steam output capacity of an
solution process. Though outputs may be minimums or older coal-fired (247 MWe unit) plant.
maximums, they will not be passive. Difficult cases tend
to occur near the edge of the FDR and involve generation Not only are the cogeneration units modelled for this
units competing for minimum or maximum output, though system of different output capacities, but they also use
they can also occur in the central regions (see Section VI). different FOR models; an irregular quadrilateral like
Table 111 demonstrates a generic “maximised” difficult Figure 1 (247 MWe unit) and a two-segment model (125.8
solution where a slight change, 6, to Unit 2 output affects MWe unit). The irregular quadrilateral is the “standard”
the outputs of Units 3 and 4, in turn significantly FOR model and is suitable for modelling all types of
increasing (+ A) the minimum operating cost. cogeneration unit. Though better suited for modelling
possible outputs from gas turbine cogeneration units, the
TABLE I11 authors of this paper found [6] that in practical terms, the
DIFFICULT
SOLUTIONCHARACTERISTICS two-segment model has no advantages over a single
quadrilateral. However, the two-segment model provides
Unit 1 I Unit 2 I Unit 3 I Unit 4 I Cost challenge to CHP dispatch problem solution
max I A I B I max I Smin methodologies; the FOR model for the 125.8 MWe unit
I max I A+6 I B-%6 Imax-%i3I $min+A I will be shown in Section VI.

B. Weaknesses
V. AN EXISTING TEST SYSTEM
Though the cogeneration unit capacities and generator
Guo, Henwood and van Ooijen [l] developed a mix for this test system are sound in principle, the models
Lagrangian Relaxation method for solving cogeneration used for the auxiliary boilers and generators have
dispatch problems. To demonstrate the method, they compromised the validity of the system. Instead of using
chose a test system and heat-power demand case that quadratic models for heat characteristics, single variable,
would allow a complete solution process to be shown in linear coefficients have been used. Because of this, these
only seven steps. This demonstration system has become a auxiliary units are quickly discounted in the Lagrangian
starting point for comparison of other methods by other Relaxation solution process. The linear coefficients have
researchers. Though the test system contained all elements been selected to be larger than the heat-only and power-
necessary to describe cogeneration plant system, certain only partial derivatives of the cogeneration heat
parameters were chosen as to ensure the optimal result characteristic models. Also the minimum viable outputs
was found in minimal time, thus removing the element of for auxiliary units have been set at 0 MWth and 0 MWe
challenge that is required for a more general test. The respectively.
strengths and weaknesses of the existing test system are
discussed below. A linear model is an acceptable model for boiler heat
characteristics, however the lack of an offset term is
A. Strengths questionable. Using 0 MWth as minimum output is a
reasonable approximation. However the maximum
This test system can be tailored to cover the majority of capacity of the boiler model (approximately 2700 MWth)
IPP cogeneration configurations. The defining features of needlessly expands the FDR.
cogeneration system are that they contain at least one
cogeneration unit and that the total power and heat outputs The simplified generator model is definitely not
of the system cannot be separated. In most cogeneration acceptable. The heat characteristic at least should be
system’s FDR either the maximum heat or power output piecewise linear, or a continuous quadratic, with a
will significantly outweigh the other. In these cases, it is significant offset term. Also the minimum output should
more economical to be able to meet the smaller of the be at least 25% of the maximum, i.e. around 40 MWe
maximum demands with cogeneration only, and then have would be expected given the maximum output is 150
auxiliary heat-only units (coal-fired boilers, or auxiliary MWe.
firing of heat recovery steam generators) or power

98
2004 IEEE Intemational Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong

The last point of discussion is the power-heat demand


Cogeneration Unit 3
case used. The heat and power demands are very small in
comparison to the system FDR and not a practical case for
the simulation of CHP plant operation, e.g. the combined
load can be met entirely by the larger cogeneration unit
alone. To rectify this, different, more relevant demand
cases could be used on the same test system. However,
because the relatively large size of the FDR due to the
boiler model and the unacceptable generator model used,
a new test system is needed.

VI. NEW TEST SYSTEM

Retained from the demonstration system of Section 4


were the generation unit mix (1 boiler, 1 generator, 2
I o 50 75 loo 135.6150
hnt.3 (MWth)

cogeneration units) and the approximately 2: 1 capacity Fig. 4. Cogencration Unit 3 2-Segment FOR
ratio between cogeneration units. Only the gas-turbine
cogeneration unit, and its 2-segment FOR model, was The cogeneration unit FOR regions and output ranges
reused. A smaller cogeneration unit replaces the 247 of equations (8) and (9) have been combined to form the
MWe unit from Section 4 and fresh boiler and generator new test system's FDR. This is shown in Fig. 5.
models have been sourced. Heat characteristic / cost
functions of the units are shown in (4) - (7).

Costl($) = 254.8863 + 7.6997 PI + 0.00172


+ 0.000115 p13 (4)
Costz($)= 2650 + 34.5 p2 + 0.1035 pZ2 + 2.203 h2
+ 0.025 h12 + 0.05 1 pzh2 (5)
COS^^($) = 1250 + 36 p3 + 0.0435 p3* + 0.6 h3
+ 0.027 h32+ 0.01 1 p3h3 (6)
Cost4($)= 950.002 + 2.0109 h4 + 0.038 b2(7)
50
The objective function, i.e. total cost per scheduling 25.
period, is the sum of (4) - (7). Boiler and generator output
constraints are listed in (8) and (9). Cogeneration unit
FOR are shown in Fig. 3 and Fig. 4.
Fig. 5. Ncw Tcst System FDR

Each point within the FDR of Figure 5 is a demand


case that has its optimal generation unit settings and
minimum cost calculated. The solid dots are simple cases,
Cogeneration Unit 2
and the open circles are difficult cases. The three
numbered cases will be considered in greater detail in
Section VII.

30 -

(0 20 40 60
Hnt.2 (Mwthl
Fig. 3. Cogeneration Unit 2 FOR

99
2004 IEEE Intemational Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004)April 2004 Hong Kong

VII. THREE DEMAND CASES optimal settings (again causing knock-on effects on the
others) without affecting the total cost.
To demonstrate the difficult and simple solutions to
demand cases, studies were run for three cases using the C. Case 3: 240 MWth - 250 MWe (Diflcult)
authors’ Cogeneration Economic Dispatch - Evolution
Program solution algorithm, “CED-EP” [ 5 ] . All cases TABLE VI
SOLUTIONS FOR 240-250 DEMAND CASE
were run for ten trials each. For the two simple cases, “1”
and “2” in Figure 5, each trial was allowed a maximum of
500 iterations, and the difficult case, “3” in Figure 5, was
allowed a maximum of 2000 iterations. In each of the set
of results shown, Tables IV, V, and VI, the top row (in
italics) is the optimal solution, followed by three
representative cases from the 10 trials run. The columns
are the generation units’ outputs followed by the total cost. Of the 2000 iterations available, between 640 and 1993
iteration were used. None of the trial runs found the
A. Case 1: 125 MWth - 205 MWe (Simple) optimal solution. In terms of variable settings, all
TABLE IV
solutions were very close, but not in absolute dollar terms.
SOLUTIONS FOR 125-205 DEMANDCASE Apparent failures like this lead to the investigation of the
underlying factors in difficult cases.

None of the units’ outputs in the optimal solution of


this demand case are passive. The output of most efficient
power generator, GI, has moved below its maximum,
while steam generating units C3 and Bq have moved to
their maximums. The high thermal energy demand (this
case is close to the edge of system FDR) has forced C3 to
This case was chosen as it is close to the centre of the
higher power output, in turn reducing the amount of
system FDR. Of the 10 trials run, none took more than
power that GI can produce. Even though all solutions are
329 iterations to find a minimum cost solution. It can be
seen that in this case the power generator, GI, is operating very close to optimal settings (except Cr discrepancies in
Table VI are less than 2 MW), the snowballing effect of
at its maximum and is passive to solution. The problem
small increments at the high end of the heat characteristic
has become how to split the remaining 70 MWe between
functions, primarily on unit C2 output setting, greatly
the cogeneration units C2 and C3, which in turn affects
increases the total dollar values of the solutions found.
how much thermal energy is available for the most
efficient steam generator, Bq, to produce.
VIII. CONCLUSIONS
The “optimal” solution outputs for unit C2 lies on
This paper has outlined the growing importance of
lower edge of the FOR of Figure 3. All the solutions
found were within a narrow margin surrounding the cogeneration, or CHP facilities, particularly in the case of
Independent Power Producer in modem electricity
optimal setting, causing a slight knock-on effect on for the
markets. It has highlighted the lack of a standard problem
output settings of C3 and B4. But these were not
set for researchers studying CHP dispatch to benchmark
significant enough to alter the total cost in terms of dollars
the performance of their solution methodologies.
and cents.
The CHP ED problem and cogeneration unit Feasible
B. Case 2: 175 MWth - 250 MWe (Simple) Operating Region have been restated, and the concepts of
the CHP system Feasible Demand Region and di@cult vs.
TABLE V
SOLUTIONS FOR 175-250 DEMAND
CASE simple demand cases introduced.

A demonstration CHP system model [l] that has


become the de facto test in recent years has been
examined critically. In the light of this, a new test system
that overcomes some of these weaknesses has been
designed and optimal solutions for various demand cases
have been found.
Of the 500 available iterations, no more than 387 Three demand cases from the new test system have
iterations were taken to find a solution. Even though both been shown in detail to illustrate the concepts of difficult
the power and heat demands have increased over case 1, and simple solutions. The proposed new test system could
the output of unit GI remains a passive variable. All other be used as a standard test system for assessing
units’ outputs in the optimal solution are not at the cogeneration economic dispatch algorithms.
extreme points, and hence all can shift slightly from their

IO0
2004 IEEE International Conference on Electric Utility Deregulation, Restructuring and Power Technologies (DRPT2004) April 2004 Hong Kong

XI. REFERENCES

Guo, T., Henwood, M.I., van Ooijen, M. "An Algorithm for


Combincd Heat and Power Economic Dispatch, IEEE Trans. on
Power Systems, Vol. 11, No. 4, Nov 1996, pp 1778-1784.
Rooijers, F.J., van Amerongen. R.A.M. " Static Economic Dispatch
for CO-Generation Systems", IEEE Trans. on Power Systems, Vol.
9, No. 3, Aug 1994, pp 1392.1398.
Moleshi, K., et. al. "Optimization of Multiplant Cogeneration
System Operation, Including Electric and Steam Networks", IEEE
Trans. on Power Systems, Vol. 6, No. 2, May 1991, pp 484-490.
L.L. Lai, J.T. Ma and J.B. Lce, "Multi-timc interval scheduling for
daily operation of a two-cogeneration system with evolution
programming", International Joumal of Electrical Power & Energy
Systems, Vol. 20, No. 5, 1998, pp 305-31 1.
K.P. Wong, C. Algie "Evolutionary Programming Approach for
Combincd Heat and Power Dispatch, Electric Power Systems
Research, Vol. 61, No. 3, April 2002, pp 227-232.
C. Algie, K.P. Wong "Reduced Cogeneration Feasible Operating
Region for Combined Heat and Power Scheduling Problems",
Procecdings of the Fifth International Conference on Power Systcm
Management and Control, IEE, London, United Kingdom, April
2002, pp 323-328.

101

Potrebbero piacerti anche