Professional Documents
Culture Documents
Automatic Generation Control in A Deregulated Power System
Automatic Generation Control in A Deregulated Power System
Automatic Generation Control in A Deregulated Power System
PE-I 83-PWR6-6-11-1007 A paper recommended and approved by FIB. I Structwre of 7+4 ccreu n d c l uf Nanvcgiurr urrd
the IEEE Power System Operations Committee of the IEEE Power Swedish power systems
Engineering Society for publication in the IEEE Transactions on Power
Systems. Manuscript submitted July 9, 1997; made available for Each physical inter-area tieline is identified and modelled in
printing November 7, 1997. an aggregate way such that between any two areas only one
tieline at 300 kV and one at 400 kV (if existlng) are repre-
\ented Lines trom 132 kV and down are omitted. Two 400
0885-8950/98/$10.00 0 1997 IEEE
1402
kV tielines between Norway and Sweden are included in the dled onl>-by the instantaneous primar! control baszd on unit
model. The values of generation and load in each area are droop characteristi2s. acci.ptin=. a fluctuation of s>.stemfre-
adjusted based on generation and consumption statistics to quency of +/-0.1 Hz. H o n e i a . xhen thz S J s t s m Operator
get a reasonably good correspondence with the inter-area notices an increasing deviation, he calls upon capacity
load flows of two different load situations: Winter peak load offered at the Regulating Power Market. This capacity will
and summer low load. then be manually activated after a certain time dela!, i S
mins. is chosen as illustration), and in gradual steps if moi-e
Based on frequency measurements from actual generation than one unit i s required to regulate. For further information
outages, dynamic parameters of the generators are then on this manual secondary control system. the reader is
adjusted to give a response similar to the complete Nordel referred to references [4, 51.
grid to casually chosen faults at both heavy load and low
load (830 M W and 590 MW nuclear power units in Sweden, The starting point (T=O) in these simulations can be any
respectively); first the stationary droop of the turbine con- instant during the hour when there is a momentarily balance
trollers to approximate the stationary situation. then the gen- between a constant (scheduled) generation and a continuall!
erator inertias and the controller transient droops and time changing load. Note that in order to get a stable start of the
constants to approximate the transient response. All parame- simulation. the load ramp starts after a 10 sec. delay. This is
ters are kept within 'normal' ranges of variation. done in all simulations shown in the following sections.
Note that the integrator time constant is chosen rather high 1001
to ensure a smooth unit operation and to prevent the LFC
from interfering with the normal unit primary conltrol 1 0 0
-.0 ~ -
1
response [81.
0 999
f *-fo
-0 020
0 100 200 300 400 500 600 700 800 900
I Time (s)
PXO
Fig. 4 System response ut peak load with central LFC to
loud rumps of 80 MW/min domestic i- 800 MW/10
min HVDC
Fig. 3 Classical loadfrequency controller 161
However, the general LFC system is not very well suited to
The integrator time constant T is set to 120 sec. with 13=O a deregulated energy market like Norway and Sweden, a fact
(integrator). Values of f3 within the range 0.1-0.5 are also which is recognized also by other authors [9]. In typical ver-
tested without changing the results significantly. The sunn of tically integrated systems the power producer is also the
all participating factors C, in the signal distributor is 1.O. owner and operator of the grid, but in Norway the grid com-
pany Statnett is responsible for the system operation, and
By introducing this controller scheme, the HVDC connec- thus the secondary control, while the generating units are
tions will be treated as internal loads in the Norwegian sys- owned by several other companies (Gencos). The System
tem. The system is now able to withstand both the domestic Operator would have to buy all reserve capacity from the
load ramp from Figure 2 and an additional steep ramping of Gencos on a commercial basis, and in the case of automatic
the HVDC connectioris without violating the allowed fre- LFC for rather long periods of time and for large and fre-
quency range. As an illustration, Figure 4 shows the quent unit load changes. (This is similar to the ‘Charged
response when the HVDC connection to Denmark is ramlped LFC’ in reference 191.) Thus, the price needed to get suffi-
from initially 200 MW load to full load of 1000 MW in I O cient amounts of reserve capacity might be high. On the
minutes (starting after 100 seconds in the simulation). ‘This other hand, it would be very difficult to place any kind of
doubles the load change rate in the Norwegian system from responsibility for system deviations at a given time, thus the
80 MW/min to 160 MWlmin. financing of LFC would most probably have to be made
with a general fee.
Note that as opposed to a step excitation, the I-controllers
are unable to catch up with the ramp. The size of the fre- As the main cause of the future increased operational strain
quency deviation is dependent on the total system frequency on the Norwegian power system is the 4 HVDC connec-
bias. The frequency bi,w of the Norwegian and Swedish sys- tions, an alternative secondary control system can be intro-
tems are nearly equal in the model. The ACE are given i n pu duced in which special power stations are selected to follow
of aggregated MVA base of controlled units. the HVDC load automatically, while the “rest” of the system
deviations are handled manually via the existing Regulating
4. RAMP FOLLOWI.NG CONTROLLER Power Market as today. In this case there is a clear connec-
4.1 Introduction tion between the responsibility for the “disturbance” - the
contractual partner(s) of the HVDC connections - and the
Fully automatic load-frequency control as illustrated above provision of sufficient reserve power. Thus, the contractual
is technically a rather simple way to handle the secondary partner would have to supply the reserve from own genera-
control in the Nordel system. The fsct that the control func- tion capacity or by purchase, not the System Operator. This
tions are generally aimed at keeping system frequency and system is similar to the concept of ‘bilateral LFC’ in ref. [9].
power interchanges at their nominal values cnsures that all
deviations in the systt:m are taken care of, no matter where 4.2 Controller layout
the deviation occurs and what might be thc cause. The only
condition is that there are sufficient reserves and transmis- An HVDC ramp following controller (RFC) is now intro-
sion capacity available. duced as shown in Figure 5. (No controllers are needed on
1404
the Swedish side.) The different HVDC ramps (e.g. daily Let:
plans of operation) are added to a total MW demand signal
A P,, to the controller. This signal is compared to changes
A P, i t ) = RC A PqOy (t)
in the generator electric power output: = R C .p P i i ( t ) + $
1' (~,~-o(t))
* PG ,
t.---*'
a
~
cn ,~~~*
-,
As the purpose of the ramp following controller is to handle
L...
the HVDC ramps automatically, while the normal domestic
Fig.5 HVDC rump following controller (RFC) with droop load changes will be handled by the current Regulating
cornpensation PoLver Market. the controller used in these simulations must
be tuned such that the system frequency follows the uncon-
One important problem now arises. As shown in Figure 6. trolled response of Figure 2 as closely as possible no matter
the additional output demand signal A P , to the turbine gov- ho\{. the HVDC connections are operated. The operation of
ernor is added to the normal stationary droop signal of the the connections is thus 'hidden' for the System Operator.
governor. The turbine governor input aPnUo,. thus consists of
two elements: In Figure 7 the s ~ s t e mfrequency with HVDC ramping is
compared to the basic response of Figure 2. To emphasize
the difference in controller response, the rather steep ramp
of 100% in 10 minutes is used for all four connections, giv-
where 6 - unit stationary droop ing a total HVDC ramp rate of 230 MW/min (in addition to
a,,,,w, ( t ) - rated and actual speed the domestic load ramp). The stability of the system is
ensured b!. inspecting the eigenvalues in each case.
-..._ . .
while the process transfer function can be written as a func-
tion of general polynomials with one or more zero poles
0 996
A e ( s ) = AP,,(s) -AP,,(s)
0 100 200 300 400 500 600 700 800 900
Time (s)
-
A p,, ($1 -
-
A pD,
(10)
Fig. 7 System frequency respon.se (pu) to a total HVDC 1+ H R ( s ) . H p ( s ) 1 + A (s)
ramp oj 2300 MlV in 10 minutes with and without
droop compensa'ion 1 + spT n (s)
where A (s) = ~ ~
ST SI . d (s)
Figure 8 shows a comparison of the HVDC load ramp and
the output of the controlled generator for the simulation of If this system is excitated with a ramp function CIS2 the fol-
Figure 7b. The difference in slope is due to the subordinated lowing result can be found:
unit primary control response.
Load (pu)
,"
'"I
08 /-----
si+ 2Td (s)
s " ' T d ( s ) + s p T n ( s )+ n ( s )
= 0 (11)