Documente Academic
Documente Profesional
Documente Cultură
2 February 1982
381
interval.
INTRODUCTION
In most power systems computer control which include
Automatic Generation Control (AGC) and Economic Dispatch, the
Economic Dispatch (EDC) program computes an economic BASE
value for designated units every few minutes, based on the load change
over the time since the last execution of the EDC program. In between
computations of this economic BASE, the frequency deviation and
inadvertent interchange is controlled by an algorithm which distributes
the Area Control Error (ACE) over certain designated regulating
units, and linearly distributes the deviation of the units from their
economic base over the economic units.(5) In using this approach the
economic BASE value calculated is usually restricted not only by the
dispatcher entered economic high and low limits, but also by the rate
at which the unit can respond over the nominal period of the economic
dispatch (5 minutes). Hence, a unit with a response rate of 5
MW/MIN, economic high and low limits of 200, 100 respectively, an
actual power of 110 MW, and a nominal economic dispatch period of
5 minutes, will be restrained in the On-Line Economic Dispatch by an
effective upper and lower limits of 135 MW and 100 respectively. The
result of doing this is that each unit is assigned an economic BASE
which can be achieved by moving it at its sustained rate of response
from the present value, and hence controlling the system in a smooth
manner.
The above methodology, however, only ensures that the units are
assigned within their capability for the next 5 minutes or so. An
extension to this method, is that of including a constraint into the
economic dispatch problem to guarantee that sufficient regulating
and/or reserve margin is maintained over a short (5 minute) period.
This problem has been resolved in reference 1, and in this paper a
more general algorithm is proposed. This method will now ensure that
the economic Base values assigned to the units will be achievable
within 5 minutes of operation, and will, if possible maintain a
sufficient short term regulating reserve.
Under the strategy described above, it is well known that the least
expensive units will be allocated close to their limits during the early
stages of the load pickup, leaving the more expensive units for the
final stages. This results in the company not being able to meet its load
pickup in the later stages, except by purchasing energy from a
neighbor, or 'dragging the ties'.
At present, this is usually handled by the dispatcher's ramping of
the more expensive units above their 'economic' assignment early in
the load pickup period and thus keeping some less expensive units
below their operational limits in this early stage so that they can help
satisfy the load pickup required at the later stages. This requires, of
course, the dispatcher to make these decisions at his own discretion.
He will tend to choose a safe solution rather than an economical one.
This problem can be resolved by using an algorithm to look ahead
at the predicted load pickup and to restrict the economic BASE assignment of the units in such a manner that the load pickup will be
satisfied. This problem has been expressed and solutions to the
problem proposed in references 2, 3 and 4. The solutions proposed in
these papers involve methods which are more general than that
proposed here, their drawback being, of course, that they require
more extensive computer resources to solve the problem.
The remainder of the paper is organized into three main sections.
The first section describes the static dispatching problem, starting with
a definition of the traditional economic dispatch, then developing a
reserve limited dispatch algorithm, and finally showing the effect of
reserve limits which are above the dispatch limits. The second section
describes the dynamic dispatching problem and its solution algorithm;
it then goes on to discuss certain aspects of including such an
algorithm in an on-line operating system. The last section presents and
discusses the results.
Xi
X4
: upper
limit of unit i
Xi
: lower
limit of unit i
Si
xi(L)
: effective upper
xi(L)
: effective lower
interval)
: amount
UP(L)
Si
: min
S(L)
1982 IEEE
Si
idl
(x4
382
fi(xi)
Problem Definition
: system load
Loss
system losses
system lambda
PFi
(PFi
minimize
=
iI
Xi
iEI
2 Xi
= 1
il
fi (xi)
LOSS
(1.)
=
(1.2)
(1.3)
xi > xli
(1.4)
Where equation 1.1 represents the fact that the sum of the production costs over the set of units I is to be minimized. The constraint
1.2 states that the sum of the generation over these units must equal
the load plus losses and 1.3 and 1.4 are the inequality constraints
stating that each generation has an assignable minimum and maximum
output.
The solution to this problem is found by using traditional
Lagrangian multipliers, and is well known in the literature (6). The
losses are represented by a B-matrix, and the penalty factors are used.
The solution is expressed as:
a f(xi)=
a xi
PFi
iI
Xi
Ii
xi= L
(2.2)
(2.3)
xi >
Xl
(2.4)
(2.5)
xl'
si > S
(2.6)
UP(L)
(x4
i- xi(L))
icIl
Vi
*
Vil
DOWN(L)
iEI2
- (X-xi(L))
(xi
loss
(2.1)
<
iI
Xi < Xiu
Xi >
fi(xi)
Si
= 1
xi +
1-a Loss
a xi
Loss)
The total reserve S can be computed from the results of the traditional
EDC.
S(L)
and A
iI2
S(L)
Si
-
reserve.
We
can
now
383
2) The reserve requirement cannot be satisfied. That is we have
a deviation from reserve requirement A, and either the rate
limited units cannot be moved up sufficiently to cover the
deviation (UP(L)<A), or the upper limited units cannot be
moved down sufficiently to cover the deviation
(DOWN(L)<A).
(A<0 and either UP(L)<A
or
3) The dispatcher may use a more constraining dispatch or control limit to save the last portion of the unit's capability for
load growth problems at a later time.
2i(Ll)
xi(L)
Xi >
xI(Ll)
Xi
icIl
xi(L)
+A
run a traditional economic dispatch with the conditions listed above to yield xi(Ll) as the result.
xiuVril
Si < 3i Vi6l (unchanged)
il
iEI2
xi(L)
x!Via (unchanged)
xi <
3i
ViaIl
xi(L) - A
Ll=
iI2
xi(L) + A +
(iEIl)
iEIl
LI + L2 = L.
x4'
The proposed method has a unique way of solving this problem as desqribed below. The essence of the problem is that
we know how much generation (A) is to be switched from the
units in the set I2(L) to those in the set I1(L), but we do not
know how to distribute this among the units. The method
described below achieves this.
Note: LI + L2 =
ii(L2)
-i
Step
= X
DOWN(L)<A)
A < 0,
and UP(L) >iA
and DOWN(L) > A'
xi(L2)
A'
si > S (unchanged).
xP
1 >lgi ViElA
2) xr
384
ViEI
S- < u
x* +
Xi < Xi
0
iEI
VE
(3.6)-
(3.7)
(3.1) Is the minimization criteria. It specifies to minimize the production cost over the set of units I required to meet the load, and
over N time intervals.
(3.3) This specifies that for each unit, the sum of its assigned generation and its reserve contribution must be less than or equal to its
high limit.
where
Si* O
Si* = Si
Si* = Si
ViElA
SF
iI
(3.2) This equation specifies that the sum of the generation must
satisfy the load requirements and the transmission losses
incurred. The losses are handled by the traditional B-Matrix,
penalty factor method.
FS
Si <Si*
Si > S*
S -
(x -X i )
i-IA
Si-
fiI
-VIEI, iEIAUIB
iEIB
(xf - xi')
(3.5) This specifies that the unit assigned generation be greater than a
lower limnit.
(3.6) This specifies that the sum of the unit reserves must be greater
than a prespecified reserve requirement.
: time interval
xi(n)
fi(xi(n))
si(n)
: reserve
(3.7) This specifies that the unit's response between time intervals is
restricted to ri.
xiU
xi(n)
xli
xi(n)
Feasible Solution
ri
: response
Si
l(n)
: system reserve
xi(n)
n = N, N - l,
requirement at time n.
1.
J(n) =
minimize
2 xi(n)
iI
Problem Definition
xi(n)
'xi(n)
iI
l(n)
n=l,
fi(xi(n))
(4.1)
l(n)
(4.2)
N
J = E I fi(xi(n))
n=l idl
minimize
iEI
ijViEI
0 6 si(n) < Si
xi(n) >
(4.4)
(3.1)
; si(n) > S(n)
iI
(3.2)
(4.3)
VilIVn= 1.
VidIVn =1.
xlViElIVn= 1,
(3.3)
(3.4)
.i(n)
(3.5)
5xi(n)
Vidl
(4.5)
(4.6)
where
=
=
(4.7)
(4.8)
max
(xi(n + 1)
385
Ri(n)
(n)
= min(x4,
(4.9)
Results
(4.10)
Algorithm
(xi)
Yi(N) = i(N)
xi(N) Xi(N)
=
Step 3 Solve the reserve limited economic dispatch for the time interval n, yielding xi(n), si(n), S(n).
interval.
(Xi)
IOOR
*100
: means
386
CONCLUSION
It was the author's intent in analyzing these problems to stay as
close to the traditional economic dispatch methods as possible, and to
develop algorithms which would have good performance
characteristics. In doing so he initially accepted that some suboptimality would probably be necessary. In the case of the spinning
reserve limited dispatch, the algorithm developed and described is
both elegent and simple, it uses the available traditional economic
dispatch algorithm as a subroutine, and, in itself introduces a trivial
amount of new code. There are some circumstances in which it is sub-
SYSTEM LOAD
1A
1:
1'
-- REQUIRED RESERVE
10
TOTAL
73
93
143
70
93
350
248
190
190
113
1563
12
26
42
18
30
100
100
40
70
40
478
RESPONSE MW
10
15
23
10
13
50
48
20
30
20
235
RESERVE BREAKPOINT
63
78
120
60
80
300
200
170
160
93
1328
HIGH LIMIT
(xi)
10
70
40
796
11 .
167
12X
43.9
44.4
18x
3Ox
1OOx
190x
867
12.3
167
12x
47.4
48.3
18x
30x
100X
190x
898
12.5
167
12x
49.8
50.9
18x
30x
100X
190x
963
13.
160.6
16.6
55.7
57.3
18x
30x
1072
13.8
137
*27.3
65.2
67.6
18x
42
125.7
l9ox
182.6 *104.4
1190
14.
121
30.1
68.2
70.2
19.6
*64
*201
19ox
190x
1391
14.6
69
38.7
82.0
78.2
24
*77.2
*350(
190X
I190X 113X
l9ox
113.3
59.8
134
65.8
166.4
80.7
113X
387
SYSTEM
LOAD LAMBDA RESERVE
3.
100
796
11.0
167
12X
43.9
44.4
18X
3Ox
867
12.3
167
12X
47.4
48.3
18x
30x
898
12.5
167
12X
49.8
50.9
18X
30x
963
13.
160.6
16.6
55.7
57.3
3OX
1072
150
27.7
65.6
68
18X
18X
1190
150
31.3
70.1
71.2
19
1391
150
*63
78
43.5
137
*300
*80
10
248X I goX
248X l9ox
00OX 248X l9ox
00OX 248X 190
*55.5 *229.4
107.4 *5 1.3
2.48X
l9ox
248X
248X
l9OX
I goX
70
40
*1 13.3
59.8
134
65.8
166.4
80.7
175.9
98.9
175.9
98.9
175.9
98.9
LOAD
RESERVE
796
167
12X
43.9
44.4
867
167
12X
54
55
898
167
12X
51
52.2
963
167
17.9
56.9
58.6
18X
I18X
18X
18X
1072
150
27.7
65.7
68.1
18X
1190
150
37.7R
78
91.1R
26.9
1391
106
47.7R
93R
1 14.1l R
36.9
19OX
IOOR
g9oX
10
TOTAL
'1OR
68.6
l9ox 160R
l9ox 175.9
84.4
19O0
175.9
98.9
19oX
190
98.9
113
1339.2
TABLE 4 - RESERVE LIMITED ECONOMIC DISPATCH WITH CONSIDERATION OF UNIT RESPONSE RATES
LOAD
RESERVE
796
167
12X
43.9
44.4
867
167
13R
54
898
167
23
963
167
1072
10
18X
30X
LOOX
248X 19OX
55
18X
30X
1X
248X
48.8
49.9
18X
30X
lOOX
248X
190X 1211R
33R
48
48.9
21.3
41
150R
167
43R
48
61.4
31.3R
54R
200R
248X
190X 131
1190
202.5
53R
63R
844R
41.3 R
67R
250R
236.5
170
1391
150
63R
78R
1074R
51.3R
80R
300R
70X J4X
145.9
63.6
61.3
| 65
78.9
388
LOAD
RESERVE
10
796
167
12X
43.9
44.4
18.
3oX
00X
248X
190X
7ox
4ox
867
167
16
52.7
54.1
18X
30X
100X
248X
190X
ioR
60R
963
167
26R
63.8
68.1
18X
38.5
100X
248X
19oX
130R
80R
898
167
33
48.8R 49.9
21.3
41
|50R
248X
19OX
126.1
63.6
1072
167
43R
48
61.4
313 R
54R
200R
248k
190X
131
65
1190
202.5
53R
63R
844R
41.3R
67R
250R
236.5
170
145.9
78.9
1391
150
63R
78R
1074R
51.3R
80R
300R
REFERENCES
1. W. 0. Stadlin, "Economic Allocation of Regulating Margin".
IEEE Trans. Power App. Syst., Vol. PAS-90, No. 4, July/
August 1971.
989R