You are on page 1of 19

Journal of Marine Engineering & Technology

ISSN: 2046-4177 (Print) 2056-8487 (Online) Journal homepage: http://www.tandfonline.com/loi/tmar20

Reactive power sharing in ship energy systems


with shaft generators

J M Prousalidis, E Xanthopoulos & K Voutzoulidis

To cite this article: J M Prousalidis, E Xanthopoulos & K Voutzoulidis (2009) Reactive power
sharing in ship energy systems with shaft generators, Journal of Marine Engineering &
Technology, 8:1, 21-38, DOI: 10.1080/20464177.2009.11020216

To link to this article: http://dx.doi.org/10.1080/20464177.2009.11020216

Published online: 01 Dec 2014.

Submit your article to this journal

Article views: 292

View related articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tmar20

Download by: [195.29.81.98] Date: 26 July 2016, At: 22:45


Reactive power sharing in ship energy systems with shaft generators

Reactive power sharing in ship


energy systems with shaft
generators
JM Prousalidis, E Xanthopoulos and K Voutzoulidis, National Technical University of Athens,
School of Naval Architecture & Marine Engineering

This paper deals with the well-based mathematical formulation of active and reactive
Downloaded by [195.29.81.98] at 22:45 26 July 2016

load sharing of the synchronised generators of a ship’s electric energy system. Thus, both
the first- and second-stage regulation during load sharing is presented via mathematical
equations. The novelty introduced in the paper is, on the one hand, that the hybrid
parallel operation of shaft and conventional generators is covered, while on the other
hand, that the analysis results in solving the combined reactive and real load sharing
within the generator rated capacity. Furthermore, the dynamic behaviour of the ma-
chines and their controllers during the load sharing problem is also discussed with the
aid of simulations in the MATLAB/Simulink environment. The entire analysis, which it is
hoped can also be used for educational purposes, is enriched by figurative results
obtained from an actual ship case study, in which this work assisted in resolving a series
of load sharing problems

AUTHORS’ BIOGRAPHIES
verters are installed.1,2 In this paper, a thorough theoretical
Dr J Prousalidis (Electrical Engineer from NTUA/1991, PhD analysis of the load sharing problem in its entirety is pre-
from NTUA/1997) is Assistant Professor at the School of sented, highlighting the differences between conventional
Naval and Marine Engineering at the National Technical and shaft diesel generators. The mathematical formulation
University of Athens, dealing with electric energy systems of the so-called primary and secondary regulation in terms
and electric propulsion schemes in shipboard installations. of active and reactive load sharing among all generators
during a load change, is presented.
E Xanthopoulos and K Voutzoulidis are graduate students This analysis is supported by simulations on a two-stage
of the School of Naval and Marine Engineering at the approach. At the first stage, the so-called ‘static problem’ is
National Technical University of Athens. Their scientific in- faced, with the load sharing problem considered without
terests include marine electrical engineering issues. any time constants. At this stage, the significance of having
well-tuned load sharing is highlighted. At the second stage,
the complete ‘dynamic problem’ is faced, taking into ac-
INTRODUCTION count the dynamic response of all generator systems in-
oad sharing among the shipboard generator sets is volved, including main electric machine dynamics, as well

L considered to be a well-defined and rather simple


practice. However, this is erroneous: several pro-
blems can be encountered in an actual ship’s elec-
trical power grid if proper attention has not been paid at the
design and trials stages. These problems increase if reactive
as those of the speed and voltage controllers. In this latter
case, the problem is solved via the MATLAB/Simulink3
computer simulation tool. Two different sub-groups of pro-
blems are studied, ie, load sharing in the cases of making
and braking a passive load and a large power motor, respec-
load sharing is sought and especially so in ships where tively.
modern shaft generator systems with power electronic con- The work presented in this paper has been provoked by

No. A13 2009 Journal of Marine Engineering and Technology 21


Reactive power sharing in ship energy systems with shaft generators

an actual case study of a ship’s electrical power system Nevertheless, due to the presence of the dc link, the SG can
comprising both shaft and conventional generators operating supply the ship electric system only with active power.
in parallel. In this case the generators did not share either Therefore, the system reactive power demands are covered
active or reactive load in the manner required by the appro- by a Rotary Compensator (or Synchronous Condenser, SC),
priate standards. ie, a synchronous machine drawing only the real power
The source programs developed during this study are necessary to sustain its own mechanical losses and therefore
intended to be used for educational purposes as a tangible running at a power factor of, essentially, zero and producing
tool to help students understand the load sharing problems or absorbing only reactive power. In this way, the synchro-
of synchronised generators. nous condenser, which slightly increases the installation and
operation cost, acts as voltage regulator for the entire sys-
tem keeping the voltage downstream the SG unit almost
constant. Hence, the combination of the SG and the SC,
SHAFT GENERATORS WITH DC LINK hereinafter called the Shaft System (SS), behaves as an
AND CONDENSER ordinary generator supplying the system with energy via the
circuit breaker ACB (Fig 1). Furthermore, in the case of
Shaft generators (SGs) have long been exploited, in both
short circuits, the SC also acts as a current source supplying
commercial and warship applications, as they can produce
the short circuit with fault current and permitting the pro-
significant amounts of electric energy and take advantage
tection scheme to offer discrimination. Moreover, the SC
of a ship’s main propulsion engines at very low cost.2 Thus,
covers the reactive losses due to harmonic distortion of both
in certain cases, one shaft generator can cover the energy
voltage and current waveforms provoked by the power elec-
demands of the entire ship without the dedicated generator
tronic devices of the dc link.
sets being used. As the main engine rotates at a variable
Downloaded by [195.29.81.98] at 22:45 26 July 2016

It is worth noting that:


speed range, the frequency output of the SG is not constant.
In most configurations, this problem is resolved by a dc
link, ie, a power electronic device comprising a rectifier in
• In former applications, the SGs could not operate in
port. Nowadays, however, the propeller can often be
series with an inverter (Fig 1) which transforms the electric
easily decoupled from the main engine allowing the
power output of the SG into constant frequency and voltage.
operation of SGs.
This is the scheme that will be considered in this paper as it
is the one most often used in applications of parallel opera-
• The ac/dc/ac conversion at the dc link provokes harmo-
nic power quality problems to the entire electric energy
tion of conventional diesel generating sets with shaft gener-
system, ie, harmonic distortion to voltage and current
ating systems.
waveforms and increased associated reactive losses.
However, as with all power electronic applications, this
problem can be resolved by introducing certain alterna-
tive harmonic quality improving measures, eg, instal-
ling harmonic filters (such as a DCL reactor coil on the
dc side and an ACL coil on the ac side of the inverter
(see Fig 1), applying sophisticated switching techniques
or using less harmonic-polluting power electronic
bridge topologies.

ACTIVE LOAD SHARING


Active load sharing among ac generators is generally based
upon their frequency droops, ie, the gradients of their f-P
Fig 1: Typical shaft generator system (frequency versus active power generation) characteristics,
which in turn reflect the relationship between the ac gen-
Before power electronics had been developed to suffi- erators and their associated prime movers (Fig 2).
cient capacity, several other solutions to the variable electric Considering the example of two generators depicted in
frequency (and perhaps voltage) problem were applied, such Fig 2, given a total load Po, the corresponding partial load-
as: ing of each generator as well as the common operating
frequency is obtained from solving the following set of
• the variable rotation speed was stabilised by rather linear algebraic equations:
complicated (and therefore expensive) gearboxes
• the main engine speed was kept constant, while the PO1
ship’s controllable pitch propeller provided the control Generator 1: f ¼ f L1  x P1   f N1 (1)
PN 1
of ship’s speed (this also had significant increased cost) PO2
• the variable frequency of the SG was used to supply an Generator 2: f ¼ f L2  x P2 
PN 2
 f N2 (2)
independent electric system comprising loads insensi-
tive to frequency variations. PO1 þ PO2 ¼ PO (3)

22 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators

movement is achieved by increasing or decreasing the fuel


injection of the generator prime mover (Fig 4).

Fig 2: Parallel operation of two generators in f-P common


plane
Fig 4: Graphical representation of fuel injection change in
where the frequency droops of the two generators are de- generator 2 working in parallel with generator 1
fined as follows:

f L1  f V 1 f L2  f V 2 Mathematical relationships P- and Q-V


x P1 ¼ x P2 ¼ (4)
f N1 f N2 The well-known single phase equivalent circuit of a three-
Downloaded by [195.29.81.98] at 22:45 26 July 2016

phase synchronous generator is shown in Fig 5.


In the special case where one generator, say generator 2, is
significantly larger than the other one (acting thus as an
infinite bus), its f-P characteristic is almost horizontal (Fig
3). Moreover, this slack generator defines the operating
frequency covering any active load fluctuations.

Fig 5: Single-phase equivalent circuit of a three-phase


synchronous generator

The voltage-current relationships are:


E~  V~
E~ ¼ Ie  ð R þ j  X Þ þ V~ ) Ie ¼ (5)
Rþ j X

Fig 3: Parallel operation of a generator with an infinite bus The single-phase complex power is obtained from
S~ ¼ V~: I~ (6)
The other generator with non-zero frequency droop
works at the frequency set by the slack generator. In the Considering the generator terminal voltage V as the refer-
case that none of the plant generators are of horizontal f-P ence phasor, (ie, its argument equals zero or V~ ¼ V j 0 o )
characteristic, one is rather arbitrarily selected to play the equation (5) turns into:

slack role. Ie ¼ I   ¼ I cos  þ jI sin 
When the load condition changes, the new operating
point (ie, load sharing among generators at a new operating ) I cos   jI sin  ¼
frequency) is defined by the generator’s frequency droops.
This first settlement, the so-called primary regulation or E cos  þ jE sin   V R  jX
 )
self-regulation corresponds to a frequency decrease in case R þ jX R  jX
of a load increase and vice versa. However, as this new
operating frequency can be beyond the permissible operat- E~ ¼ Ej  ¼ E cos  þ jE sin 
ing limits the so-called secondary regulation is performed. ½ð E cos   V Þ R þ X  E  sin 
According to this procedure, the f-P characteristic of the þ j½ E  R  sin   X ð E  cos   V Þ
regulating unit moves upwards or downwards (but without I cos   jI sin  ¼
R2 þ X 2
changing its slope) so that the new loading conditions are
covered at a frequency close to the nominal one. This The separation of real and imaginary parts leads to:

No. A13 2009 Journal of Marine Engineering and Technology 23


Reactive power sharing in ship energy systems with shaft generators

1
I cos  ¼  ½ð E cos   V Þ R þ X  E  sin  (7)
R2 þ X 2
1
I sin  ¼ 2  ½R  E  sin  þ X ð E  cos   V Þ
R þ X2
(8)
The complex output power S~ at the generator output term-
inals is:

S~ ¼ V~  I~ ¼ V j 0 o  I  
(9)
¼ V  I  cos  þ j  V  I  sin   P þ jQ Fig 7: Voltage vs reactive power (V-Q) curve of synchronous
P ¼ V  I  cos  (10) generator
Q ¼ V  I  sin  (11)

By substituting (7), (8) into (10) and (11) P and Q expres- are significantly more restricted (perhaps down to almost
sions are yielded: 40%) of the static stability limit values given in expressions
V (16) and (17)4 and illustrated in Figs 6 & 7. This additional
P¼ 2
 ½ð E  cos   V Þ R þ x  E  sin  (12) restriction is due to the nature of transient stability which
R þ X2
considers the system’s capability to recover to steady-state
V
Q¼ 2  ½ð E  cos   V Þ X  R  E  sin  (13) operation after fast transients, such as short circuit faults
R þ X2
Downloaded by [195.29.81.98] at 22:45 26 July 2016

and switching operations. (Operating at the very limits of


static stability – the peaks in Figs 6 & 7—is effectively
Considering that in most cases R  X, equations (12) and impossible as any perturbation which increases the load will
(13) are simplified by eliminating the R-terms: move the machine into the unstable area. The amount of
VE margin necessary to ensure transient stability depends on
P¼  sin  (14) the overall system design and the level of destabilising
X
perturbations arising form both normal and abnormal opera-
V
Q ¼  ð E  cos   V Þ (15) tion.)
X In contrast, static stability refers to extremely slow per-
Equations (12) (13) (or their plausibly simplified versions turbations from normal operation, which set higher limit-
in (14) (15) which are the most commonly used) comprise ations. In between stands the dynamic stability problems,
the P-ä and Q-V operation rules respectively – they are corresponding to case studies of such duration that certain
graphically presented in Figs 6 and 7. automatic control sub-systems like the Automatic Voltage
The shaded areas in Figs 6 & 7 correspond to static Regulator (AVR) are capable of reacting and participating
stable operation areas, ie, the following conditions (inequal- effectively in the entire phenomenon. It is to be noted that
ities) are fulfilled: frequency regulation that relies on mechanical speed gover-
 VE nors can never be fast enough to materially affect dynamic
< & P< (16) stability (even mechanically speed regulated diesel-genera-
2 X
tors).
1 1
 E  cos  < V < E  cos  & Q <  E2  cos2 
2 4X
(17) REACTIVE LOAD SHARING
Reactive load sharing among generators is similar but not
It is worth noting that, in practice, transient stability areas identical to active load sharing. The main difference is that
although both the V-Q and f-P relations are non-linear only
that for f-P can be sensibly linearised around the operating
point (and then only where margin for transient stability is
included in the maximum load operating point). Conversely
the V-Q relationship can only be approximated by a linear
curve in certain cases. However, as shown in Fig 8, the
more the generator emf, E, increases the more the curve
can be considered to be linear, at least for small reactive
loads and small reactive load changes.
It is worth noting that the AVR acts faster than the
speed governor – by at least an order of magnitude – and
this allows the ‘P-’ relationship during active power reg-
ulation to be simplified by regarding the emf, E, to be of
Fig 6; Active power P vs power angle  diagram of constant value. On the contrary, the ‘Q-V’ relationship
synchronous generator senses changes of the -angle, but via the cos-function, ie,

24 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators

ing of the generator (it is regulated by the speed-governor


intervention to the f-P rule).
On the other hand, referring to the regulation time
delays of the dependent variables:
Emf E: very fast regulation by the AVR (in the order of
100-200ms)
Power angle : fairly slow regulation by the speed
governor (in the order of 1000-2000ms for diesel generating
sets).
Therefore during load changes (referring to either active
Fig 8: Effect of increased emf E on Q-V curve or reactive power):
• the AVR detects variation of voltage V (due to load
change) and reacts fast, regulating emf E and therefore
in a rather insensitive manner especially for the small range
reactive power Q, too.
of -values (0,,408) where the generator normally oper-
ates (because of transient stability limits).
• the speed governor detects variation of f and reacts
relatively slowly regulating active power P via the fuel
By inspecting equations (13) or (15), it is easily seen
injection rule. This regulation sets the power angle 
that, because the V-Q relationship is not linear, the voltage
(eg, it is obtained from (14) that  ¼ sin1 VPXE ).
droop varies with the operating point (unlike the real load
droop relationship). If voltage droop is defined as: However, this variation of  is reflected in the Q-V relation-
@V 1 ship as well, spoiling the fast reaction of the AVR (it is to
xQ ¼ ¼ (18)
Downloaded by [195.29.81.98] at 22:45 26 July 2016

@Q be noted that the voltage droop xQ depends on both E and


@Q
cos). This leads to a new (secondary) regulation of the
@V
reactive power.
then the accurate expression (13) leads to:
1
x1
Q ¼
½ð E  cos   V Þ  R  E  sin 
R2 þ X 2 THE SPECIAL CASE OF THE SHAFT
(19)
V GENERATOR SYSTEMS
þ (X ) As previously mentioned this paper considers the most
R2 þ x 2
common scheme for shaft power generation, namely a shaft
generator (most often a synchronous three-phase one) along
whereas the approximate expression (15) results in:
with a synchronous condenser (Fig 1). The shaft system SS,
1 V as a whole, operates in parallel to the conventional genera-
x1
Q ¼
ð E  cos   V Þ  (20)
X X tor sets via the intervention of a power ac/dc/ac converter
with the generator producing only active power and the
As a mean representative value, the gradient between the condenser supplying the corresponding reactive power.
stable area limits can be considered: Thus, the f-P mechanism refers to the SG while the V-Q
1 mechanism refers to the SC. It is worth noting that due to
˜V E  cos    E  cos  power electronics, the SG could fix the frequency despite
x^Q ffi ¼ 2
˜Q 1 2 any load changes acting as an infinite bus. However, in
 E  cos2 
most cases, the SGs are adjusted so that they have a non-
04
X zero frequency droop working in a similar manner to con-
(21) ventional generators. In this way, parallel operation between
1
 E  cos  2X SGs and conventional generating sets can be accomplished.
¼ 2 ) x^Q ¼  The V-Q mechanism is controlled by properly adjusting
1 2 E  cos 
 E  cos2  the AVR of the synchronous condenser of the shaft system.
4
X In this machine, the active power is almost zero, ie, its
power angle is zero:
As yielded from equations (19 or (20) voltage droop xq P 0 ) sin  ffi 0 )  ’ 0 (22)
depends on the following parameters:
R: armature winding resistance (varies slightly with (in the equation below, with sin  ffi 0, cos  ffi 1)
temperature) V
X: synchronous reactance (constant, but transient and )Q¼  ð E  cos   V Þ ) Q
x
sub-transient reactances differ) (23)
V: network voltage (held constant by proper regulation V  ðE  V Þ E  V  V 2
¼ ¼
through the V-Q relationship) x x
E: generator emf (it is regulated by the AVR interven-
tion to the field circuit) As deduced from equation (23) compared to the conven-
: power angle reflecting mainly the active power load- tional generators:

No. A13 2009 Journal of Marine Engineering and Technology 25


Reactive power sharing in ship energy systems with shaft generators

8   9
• this Q-V has a point of maximum with increased both >
> (14) Ek  V
sin  k
>
k ¼ 1, 2, . . . M t >
Q and V co-ordinates, as cos¼1
>
< ! Pk ¼ Xk
>
=
• as this Q-V relationship is independent on the power >
>
X
Mt
V >
>
angle , see equation (21), reactive loading of the >
:  ð E k  cos  k  V Þ ¼ Q tot >
;
x
k¼1 k
synchronous condenser (and hence the entire shaft gen-
erator system) is defined only by appropriate setting of (26)
the emf, E, via the AVR without any iteration with the
P-f relationship. These non-linear sets of equations can be solved by apply-
ing iterative numerical procedures such as the Newton-
As cos()¼1, the voltage droop is of a lower value, see Raphson method, which is presented in the following.
equations (20) and (21), ie, the V-Q characteristic is more The generic formula of Newton-Raphson is:
horizontal (Fig 9). This leads to the condenser being able to  1
act as a regulating unit, stabilising the voltage, V, and cover- @g
y ¼ yo  : g( yo ) (27)
ing all reactive load demands. @y

where in this case:


y ¼ [ 1 2 ... M t V ]T
yo ¼ [ 1,o 2,o ... M t ,o V o ]T
Ek  V
g(k): sin  k  Pk ¼ 0 k ¼ 1, 2, . . . M t
Downloaded by [195.29.81.98] at 22:45 26 July 2016

xk
(28)
X
Mt
V
g(M t þ 1):  ð E k  cos  k  V Þ  Q tot ¼ 0
k¼1
xk

Fig 9: Linearised approximation of V-Q curve Whereas regarding the Jacobian matrix elements:
@ g(1) E1  V @ g(1)
Primary load sharing ¼ sin 1
@V X1 @1
Consider a set of Mt generators (comprising M1 shaft sys-
E1  V @ g(1) @ g(1)
tems and M2 diesel generator sets) working in parallel at an ¼ cos 1 ¼ ... ¼ ¼0
instant, when the total active and reactive load is Ptot and X1 @2 @ M t
Qtot respectively, and while the generators’ no load frequen- @ g(2) E2  V @ g(2) @ g(2)
¼ sin 2 ¼0
cies and emfs are fL1 , fL2 , . . ., fLMt and E1 ,E2 , . . ., EMt @V X2 @1 @2
respectively.
E2  V @ f (2) @ g(2)
The f-P set of equations is as follows: ¼ cos 2 ¼ ... ¼ ¼0
8 9 X2 @3 @ M t
>
> Pk >
>
< f ¼ f Lk  x Pk  PNk  f Nk >
> = ...
XMt (24) @ g(M t ) E M t  V @ g(M t ) @ g(M t )
>
> >
> k ¼ 1, 2, 3, . . ., M t ¼ sin 6 ¼ ... ¼0
>
: P k ¼ P tot >
;
k¼1
@V X Mt @1 @5
@ g(M t ) E M t  V
¼ cos  M t
ie, a Mt xMt set of linear equations, where the unknown @ M t X Mt
quantities are: P1, P2 , . . ., PMt and f.
@ g(M t þ 1) E1 cos 1  2V E2 cos 2  2V
On the other hand, regarding the V-Q set of equations, if ¼ þ
@V X1 X2
the approximate linearised V-Q equations (those with the
voltage droop xQ ) are used, the corresponding set of equa- E M t cos  M t  2V
þ ... þ
tions is very similar to the f-P one: X Mt
8 9
>
> Qk >
> @ g(M t þ 1) VE1 @ g(M t þ 1)
< V ¼ E k  xQk  Q Nk  V Nk >
> = ¼ sin 1
@1 X1 @2
XMt (25)
>
> >
> k ¼ 1, 2, 3, . . ., M t VE2 @ g(M t þ 1)
>
: Q k ¼ Q tot >
; ¼ sin 2 . . . (29)
k¼1 X2 @ M t
VE M t
ie, a Mt xMt set of linear equations, where the unknown ¼ sin  M
X Mt t
quantities are: Q1 , Q2 , . . ., QMt and V.
Alternatively, the unknown variables, ie, 1 , 2 , . . ., Mt
and V can be calculated via the following Mt xMt set of hence, the Newton Raphson general equation in matrix
exact but non-linear expressions: form is as follows:

26 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators

2 3 2 3
1 1,o
6 7 6 7
6 7 6 7
6 2 7 6 2,o 7
6 7 6 7
6 7 6 7
6 . . . 7 ¼ 6 . . . 7
6 7 6 7
6 7 6 7
6 7 6 7
6 M t 7 6 M t ,o 7
4 5 4 5
V Vo
2 31
E1 V  E1
6 sin 1,o cos 1 0 ... 0 7
6 X1 x1 7
6 7
6 7
6 E2 V o  E2 7
6
6 sin 2,o 0 cos 2,o ... 0 7
7
6 X2 X2 7
6 ... ... ... ... ... 7
6 7
6 7
6 EM t Vo  E M t 7
6 sin  M t ,o 0 0 ... cos  M t 7
6 7
6 X Mt X Mt 7
6 7
6 7
4 E1  cos 1,o  2V o E M t  cos  M t ,o  2V o V  E1 V o  E2 Vo  EM t 5
þ ...  sin 1  sin 2,o ...  sin  M t ,o
X1 X Mt X1 X2 X Mt
2 3
E1  V o
sin 1,o  P1
Downloaded by [195.29.81.98] at 22:45 26 July 2016

6 X1 7
6 7
6 7
6 E2  V o 7
6
6 sin  2,o  P 2 7
7
6 X2 7
6 ... 7
36 6 7
7
6 EM t  Vo 7
6 sin  M t ,o  PM t 7
6 7
6 X Mt 7
6 7
6 7
6X Mt
Vo 7
4 5
 ð E k  cos  k,o  V o Þ  Q tot
k¼1
X k

(30)

As soon as these unknown variables, ie, 1 , 2 , . . ., Mt possible, while each generator is loaded by its proportionate
and V are calculated, then each generator’s reactive power active and reactive load, PK and Q K :
can be assessed, too:
V ð E1  cos 1  V Þ PN K
Q1 ¼ PK ¼ PO¸ k ¼ 1, 2, . . . Mt (32)
X1 X
M
V ð E2  cos 2  V Þ PN K
Q2 ¼ (31) K¼1
X2
  QN K
V E M t  cos  M t  V QK ¼ QO¸ k ¼ 1, 2, . . . Mt (33)
QM ¼ X
M
X Mt QN K
K¼1
The solution yielded in the linear V-Q, equation (25), is less
accurate than that obtained by the iterative procedure which
is based on the more precise equation (26). However, this By setting the desired (proportionate) active power produc-
linearised method can be used as a means to get an initial tion for each generator, Pk , the no load frequencies fLk ’s
guess vector y0 for use in the precise method. are calculated from:

Secondary load sharing Pk


f Lk ¼ f N þ x Pk   f Nk (34)
As soon as assessment of the primary loading is finished, PNk
assessment of the secondary load sharing can begin. At this
stage the central energy management system makes Similarly, the desired proportionate reactive load sharing
the effort to converge the system frequency and voltage as among the generators is also known, ie, Q k , therefore, the
close to their nominal values, fN and VN respectively, as is generator emf’s, Ek ’s, can be calculated from:

No. A13 2009 Journal of Marine Engineering and Technology 27


Reactive power sharing in ship energy systems with shaft generators

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!2
u
u P X 2 Qk X k
behaviour is best, and most easily, studied in a computer
Ek ¼ t k k
þ þ VN k ¼ 1, 2, . . ., M t environment dealing with automation and control systems.
VN VN In this paper, the well-known simulation environment of
(35) MATLAB/Simulink3 has been exploited. Typical control
schemes for speed governors and AVRs are considered (Figs
Finally should the load angles k be required they are 10 & 11). Furthermore, the model of a conventional diesel-
obtained from: generator set is shown in Fig 12, while the circuit model of
a shaft system is depicted in Fig 13. According to the
Pk X k
tan  k ¼ k ¼ 1, 2, . . ., M t (36) approach followed, a Load Flow subroutine runs first so that
Qk X k þ V N 2 initial conditions are determined and then, the full scale
simulation is run. Interest is focused on the dynamic reac-
It can be seen that, at this stage, no iterative calculation tion of all system components, which takes place during the
procedure is required. On the other hand, due to practical primary regulation, and the subsequent elimination of fre-
restraints, in a real generator system the target of propor- quency and voltage errors at the secondary regulation.
tionate active and reactive loading at both nominal fre-
quency and voltage cannot be attained precisely. This is the
reason why most standards5-7 specify a tolerance range LOAD SHARING AMONG
within which the proportionate loading is considered suc- CONVENTIONAL AND SHAFT
cessful.
GENERATOR SYSTEMS –
k˜PK k ¼ kPK  PK k ,  P (37)
MATHEMATICAL CASE STUDY
k˜QK k ¼ kQ K  Q K k ,  Q
Downloaded by [195.29.81.98] at 22:45 26 July 2016

(38) This case study is obtained from an actual ship electric


energy system and comprises six 60Hz generators (Mt ¼6)
These tolerances, P and Q are based on the rated power working in parallel, two of which are shaft systems (M1 ¼2)
capacities of all generators participating in load sharing. and four conventional diesel generating sets (M2 ¼4), as
shown in Fig 14. The shaft systems, as described in the
preceding sections, are composed of shaft generators
DYNAMIC SYSTEM SIMULATION (450V@1000kW) along with ac/dc/ac power converters and
The analysis made so far is static having no inherent rotating condensers (450V@750kVAr). Generators 1-3 are
dynamic character. Hence, no time delay nor overshooting installed in the bow electric power plant, while generators
related to the primary or secondary load sharing made by 4-6 in the stern plant; each plant has a directly associated
machine dynamics or the control schemes of the speed generator bus. Nevertheless, the main generator bus is con-
governors and the AVRs has been introduced. This dynamic sidered to be a unique one, as the two buses are directly

Fig 10: Speed governor


model in MATLAB
environment

Fig 11: AVR model in MATLAB environment

28 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators

 
Pk  X k
sin  k ¼ k ¼ 1, 2, . . . 6 (41)
Ek  V

These equations set a non-linear system solved via the


Newton-Raphson iterative numerical procedure as described
above. Hence:
 1
@g
y ¼ yo  : g( yo ) (42)
@y
Ek  V
g(k): sin  k  Pk ¼ 0 k ¼ 1, 2, . . . 6 (43)
xk
Fig 12: Diesel generator model in MATLAB environment X
6
V
g(7):  ð E k  cos  k  V Þ  Q tot ¼ 0 (44)
k¼1
xk

connected and the voltage drop between them is negligible.


The rated values of the diesel generators are 770kW and Whereas regarding the Jacobian matrix elements:
577.5kVAr @ 450V. Almost all speed and voltage droops @ g(1) E1  V
xP and xQ * varied significantly from the ideal value of 3% ¼ sin 1
@V X1
prescribed by the manufacturer (Table 1). This was due to
inadequate calibration of their governors and AVRs. Thus, @ g(1) E1  V
¼ cos 1
Downloaded by [195.29.81.98] at 22:45 26 July 2016

the generators did not share either active or reactive load in @1 X1
the proportionate manner that the standards stipulate, while
in many cases reverse power flow was noticed leading to @ g(1) @ g(1)
¼ ... ¼ ¼0
unpleasant tripping of the protection scheme. This is the @2 @6
problem that has provoked the work presented in this paper. @ g(2) E2  V @ g(2)
The mathematical analysis developed in the preceding ¼ sin 2 ¼0
@V X2 @1
sections is exploited to highlight the significance of propor-
tionate load sharing among generators and of having equal @ g(2) E2  V @ f (2) @ g(2)
¼ cos 2 ¼ ... ¼ ¼0
values of frequency and voltage droops. Furthermore, the @2 X2 @3 @6
accuracy of the non-linear set of equations of the primary @ g(6) E6  V @ g(6) @ g(6)
regulation versus the linearised is examined. All cases are ¼ sin 6 ¼ ... ¼0
@V X6 @1 @5
compared to a base reference one, which refers to equal
droops and equal initial operating points for all generators @ g(6) E6  V
¼ cos 6
(the so-called ideal). In all cases where the total load con- @6 X6
sidered is about 60% the nominal capacity of all six gen- @ g(7) E1 cos 1  2V E2 cos 2  2V
erators is 3000kW and 2200kVAr). ¼ þ
@V X1 X2
E6 cos 6  2V
þ ... þ
Primary load sharing X6
The set of active power balance, ie, P-f, equations is: @ g(7) VE1 @ g(7) VE2
8 9 ¼ sin 1 ¼ sin 2
> Pk > @1 X1 @2 X2
> >
< f ¼ f Lk  x Pk  PNk  f Nk >
> = (45)
@ g(7) VE6
X6
k ¼ 1, 2, 3, . . ., 6
(39) ... ¼ sin 6
>
> >
> @6 X6
>
: Pk ¼ Ptot >
;
k¼1

ie, a 7x7 system with unknown variables: P1, P2, . . ., P6 Secondary load sharing
and f. In this case the set of active load sharing equations results
The corresponding reactive power balance equation is: in obtaining the required no-load frequency setting from
proportionate active loading Pk and nominal system operat-
X
6
V
 ð E k  cos  k  V Þ ¼ Q tot k ¼ 1, 2, 3, . . ., 6 ing frequency fN :
k¼1
xk
Pk
(40) f Lk ¼ f N þ x Pk   f Nk k ¼ 1, 2, . . ., 6 (46)
PNk

Where angles k are interrelated via the active power quan- Concerning assessment of reactive load sharing through the
tities: accurate non-linear Q-V relationship, the no-load voltage is
obtained from proportionate reactive loading Q k and nom-
* Mean values of voltage droops, ie, mean value of Q-V gradient inal system operating voltage VN:

No. A13 2009 Journal of Marine Engineering and Technology 29


Downloaded by [195.29.81.98] at 22:45 26 July 2016

30
Reactive power sharing in ship energy systems with shaft generators

Fig 13: Shaft system model comprising a shaft generator, a rotary condenser and a ac/dc/ac link in MATLAB environment

Journal of Marine Engineering and Technology


No. A13 2009
Reactive power sharing in ship energy systems with shaft generators

Fig 14: Case study


scheme: two shaft
systems synchronised
with four diesel
generators
Downloaded by [195.29.81.98] at 22:45 26 July 2016

Xp1 (%) 2,20 Xq1 (%) 3,89 fL1 (Hz) 60,65 E1 (V) 460,47
Xp2 (%) 2,68 Xq2 (%) 0,69 fL2 (Hz) 61,59 E2 (V) 460,68
Xp3 (%) 2,10 Xq3 (%) 0,53 fL3 (Hz) 60,67 E3 (V) 458,06
Xp4 (%) 2,19 Xq4 (%) 0,35 fL4 (Hz) 60,96 E4 (V) 456,40
Xp5 (%) 2,58 Xq5 (%) 0,12 fL5 (Hz) 60,53 E5 (V) 456,66
Xp6 (%) 2,94 Xq6 (%) 0,88 fL6 (Hz) 60,85 E6 (V) 459,58
Table 1: Non-equal speed droops and mean voltage droops Table 2: No-load frequencies and voltages, fLk and Ek,k¼1,2,...6
of the ship case study as obtained from field test measure- as obtained from field test measurements
ments
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
!2
u
u P X 2 Q X
(xpk ¼3%, k¼1,2,. . .,6), while during the primary reac-
t k k k k tive load sharing the linear set of Q-V equations is
Ek ¼ þ þ VN k ¼ 1, 2, . . ., 6
VN VN used. Thus, in this case all generator voltage droops
(47) have a constant common value (xQk ¼3%, k¼1,2,. . .,6).
• Case III: frequency and mean voltage droops are as
Concerning assessment of reactive load sharing through the tabulated in Table 1, while fLk and Ek , k¼1,2,. . .6 are
linearised Q-V relationship, the no-load voltage is obtained as tabulated in Table 2. Primary reactive load sharing is
from: performed based on the non-linear set of Q-V equa-
tions. This was the case to ‘be corrected’.
Qk
E k ¼ V N þ xQk   V Nk k ¼ 1, 2, . . ., 6 (48)
Q Nk
The results of the primary load sharing assessment are
summarised in Table 3, from which the following remarks
The initial operating point (no-load frequencies and vol- can be made:
tages, fLk and Ek,k¼1,2,...6 ) for the Newton-Raphson proce-
dure of the primary regulation calculations is obtained from • As expected there is no error between cases I and II
a set of field measurements performed aboard (Table 2). with respect to active power load sharing as in both
The following cases have been considered: cases the corresponding P-f curves are the same. Con-
versely, the relative error in reactive load sharing is of
• Case I: all frequency droops are equal to one another the order of 15% for power generation, and 2% for
(xpk ¼3%, k¼1,2,. . .,6), while during the primary reac- voltage. This is the error of linearising the Q-V curve.
tive load sharing the non-linear set of Q-V equations is • With respect to the load sharing of the actual ship
used. The no-load frequencies and no-load voltages are generation system with unequal droops, the no-load
also of common value for all generators (ie, fLk ¼60.6 set-points (fLk and Ek ) vary and make the assessment
Hz, Ek ¼450 V, k¼1,2,. . .6). This case is considered the problematic. Thus, case III demonstrates how important
reference case to which all the others are compared. the good calibration of generator speed governors and
• Case II: all frequency droops are equal to one another voltage regulators is for proportionate loading. Nega-

No. A13 2009 Journal of Marine Engineering and Technology 31


Reactive power sharing in ship energy systems with shaft generators

Case I Case II % error Case III %error


f(Hz) 59.54 59.54 0.00 60.56 -1.72
P1 (kW) 590.55 590.55 0.00 65.47 88.91
P2 (kW) 454.72 454.72 0.00 491.87 -8.17
P3 (kW) 454.72 454.72 0.00 65.00 85.70
P4 (kW) 454.72 454.72 0.00 231.98 48.98
P5 (kW) 454.72 454.72 0.00 -16.70 103.67
P6 (kW) 590.55 590.55 0.00 162.37 72.51
V(V) 459.68 452.21 1.63 458.60 0.23
Q1 (kVAr) 518.06 433.07 16.40 2992.19 -477.58
Q2 (kVAr) 290.97 333.46 -14.60 1870.19 -542.74
Q3 (kVAr) 290.97 333.46 -14.60 -490.12 268.44
Q4 (kVAr) 290.97 333.46 -14.60 1985.82 782.48
Q5 (kVAr) 290.97 333.46 -14.60 1751.50 701.95
Q6 (kVAr) 518.06 433.07 16.40 1565.05 -202.10
Table 3: Primary load sharing results of all cases

tive active power flow (reverse power flow) is observed condenser and a diesel generator). On the contrary, in
for generator No 5. It is worth mentioning that in the case II, where the linearised Q-V equation is used, a
actual ship system, as soon as this problem was cor- different set value (about 12V lower than that of cases I
rected a set of measurements close to that of case I was and II) is yielded, which is common for all six genera-
Downloaded by [195.29.81.98] at 22:45 26 July 2016

measured. tors.

Secondary regulation, where voltage and frequency values


are brought back close to their nominal values, while the Simulation of the dynamic behaviour of load sharing
proportionate loading in Table 4 is sought, is summarised In this case, the system considered is modelled in the
on Table 5. MATLAB/PSB-Simulink environment, as shown in Fig 15.
For each generator an AVR and a speed governor like
P1 (kW) 590.551 Q1 (kVAr) 433.071 the ones shown in Figs 10 & 11, are considered. Thus,
P2 (kW) 454.724 Q2 (kVAr) 333.465 during the simulation, the actual primary load sharing in-
P3 (kW) 454.724 Q3 (kVAr) 333.465
cluding the dynamic transition from one loading stage to
P4 (kW) 454.724 Q4 (kVAr) 333.465
another one, due to generator self-regulation, can be stud-
P5 (kW) 454.724 Q5 (kVAr) 333.465
P6 (kW) 590.551 Q6 (kVAr) 433.071 ied. The initial conditions of the simulations are yielded
from a preliminary load flow procedure run for the entire
Table 4: Proportionate loading among all six generators circuit. The generators are initially loaded by a resistive-
inductive static load (P¼5120kW and Q¼3840kVAr) along
Case I Case II Case III
with a large power asynchronous motor (emulating the be-
fL1 (Hz) 61.063 61.063 61.063
haviour of a 1340hp thruster motor with QC ¼500kVAr).
fL2 (Hz) 61.063 61.063 61.063
fL3 (Hz) 61.063 61.063 61.063 Two distinct representative load changes based on this
fL4 (Hz) 61.063 61.063 61.063 scheme are considered:
fL5 (Hz) 61.063 61.063 61.063 • Case A: The resistive-inductive load is initially con-
fL6 (Hz) 61.063 61.063 61.063
nected but immediately disconnected at 0 seconds and
E1 (V) 459.947 447.795 458.874
E2 (V)
re-connected at 5s. The motor is de-energised through-
460.047 447.795 458.974
E3 (V) 460.047 447.795 458.974 out.
E4 (V) 460.047 447.795 458.974 • Case B: The static load and motor are connected
E5 (V) 460.047 447.795 458.974 throughout. A step-up increment (from 5.3kNm to
E6 (V) 459.947 447.795 458.874 6.2kNm) takes place on the shaft torque load of the
large power motor at 3s.
Table 5: Secondary load sharing results of all case studies
In case A, Fig 16 illustrates the main generator bus voltage
and Fig 17 illustrates the system frequency. During initial
From the results of Table 5, it can be seen that: disconnection both voltage and frequency encounter a tran-
sient decaying increment due to load shedding, while during
• In all three cases, the setting required for the no-load re-connection at 5s, they both suffer from a sag due to the
frequency is the same for all generators and equal to abrupt load increment. The voltage sag is in the order of
61.063Hz. 70V (about 15%), while the corresponding frequency fluc-
• The situation alters in the case of no-load voltage. In tuation has a transient dip of 5.5% followed by a swell of
cases I and III, where the Q-V curve is considered non- 2.5%. The speed and voltage controllers manage to restore
linear, a value of 460 and 459 respectively is required voltage and frequency nominal values within a time frame
(a small difference is noticed between a shaft system of 2-3s.

32 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators
Downloaded by [195.29.81.98] at 22:45 26 July 2016

Fig 15: Ship electric network simulated in MATLAB environment

Regarding the loading of the generators, representative hand, in Figs 20 and 21 the corresponding reactive power
curves are shown in Figs 18 to 21. More specifically, in quantities are presented, the one supplied by one shaft con-
Figs 18 & 19 where the active power production of one denser and the one provided by one diesel generator. In this
shaft generator and one diesel generator are shown, it can case, at the first interval when the load is disconnected
be seen that in the initial interval the generator does not there is some reactive power produced mainly covering the
actually produce any active power; the situation changes requirements of all the choke filters of the ac/dc/ac conver-
rapidly at the instant of the load re-connection. On the other ter. As soon as the load is re-connected the reactive power

No. A13 2009 Journal of Marine Engineering and Technology 33


Reactive power sharing in ship energy systems with shaft generators

Fig 16: Main bus


voltage fluctuation
(case A)
Downloaded by [195.29.81.98] at 22:45 26 July 2016

Fig 17: System


frequency fluctuation
(case A)

Fig 18: Active power


produced by one shaft
generator

34 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators

Fig 19: Active power


produced by one diesel
generator
Downloaded by [195.29.81.98] at 22:45 26 July 2016

Fig 20: Reactive power


produced by one shaft
condenser

Fig 21: Reactive power


produced by one diesel
generator

No. A13 2009 Journal of Marine Engineering and Technology 35


Reactive power sharing in ship energy systems with shaft generators

Fig 22: Main bus


voltage (case B)
Downloaded by [195.29.81.98] at 22:45 26 July 2016

Fig 23: System


frequency (case B)

Fig 24: Induction motor


electromagnetic torque
(case B)

36 Journal of Marine Engineering and Technology No. A13 2009


Reactive power sharing in ship energy systems with shaft generators

Fig 25: Induction motor


rotation speed (case B)

demands are increased and the generators tend to respond mathematical equations. Among the difficulties faced in the
Downloaded by [195.29.81.98] at 22:45 26 July 2016

to this increment. paper are the consideration of hybrid parallel operation of


In case B, following the motor starting-up, it initially it shaft and conventional generators where each shaft system
takes about 2s for voltage and frequency to stabilise (Figs participates via a pair of machines, one for active power
22 & 23) and also Figs 24 & 25 for the motor’s electromag- and one for reactive power generation. The analysis also
netic torque and its rotation speed. As the motor encounters covers the combined proportionate load sharing of not only
a small step-increase in torque at 5s, its rotation speed active power demands but also of reactive ones. In the case
slightly decreases. The system frequency encounters a sag study presented, which is based on an actual ship’s electric
which is restored completely within 1.5s. Only a very small network, the significance of having well-tuned controllers
sag is noticed in system voltage following this increase in and droops is highlighted. The paper is further enriched by
power demand of the motor; and in any case this is very representative load sharing simulations during abrupt load
quickly eliminated. changes, where the dynamic behaviour of the generators
It is worth noting that no regulation directly stipulates themselves in combination with their speed and voltage
the time window within which the primary and secondary controllers play a significant role.
regulations are performed. The only restrictions often met5-9
mention that no voltage transient fluctuations should exceed
10% the nominal value, but within a time interval equal REFERENCES
to half the fundamental period. The corresponding con- 1. IEEE Standard 45-1993 Recommended practice for
straint for the frequency is 5%, without any time refer- electrical installations on shipboard, IEEE Press, New York,
ence. Anyhow, it is deduced that in case A, both voltage 1993.
and frequency limits are temporarily exceeded and they take 2. Prousalidis J, Hatzilau IK, Michalopoulos P, Pavlou I,
at least two seconds to recover (instead of few milliseconds) Muthumuni D. 2005. Studying ship electric energy systems
to their nominal values. This problem can result in serious with shaft generator, Proceedings of Electric Ship Sympo-
power quality problems such as ‘voltage stability’ and sium ESTS05, USA.
‘flickering’. Moreover, when some large power loads (the 3. MATLAB: The Mathworks Inc. Matlab User’s Guide
so-called ‘pulsed loads’) are cyclically connected, discon- (2004).
nected and re-connected to the grid, voltage and frequency 4. Van Cutsem T and Vournas C. 1998. Voltage stability
modulation phenomena can emerge.10-12 The effect of dy- analysis of electric power systems, Vluwer Academic Pub-
namic behaviour of both the speed governor and the voltage lications.
regulator is of primary importance in such cases and has to 5. Germanischer Lloyd (GL). General Requirements and
be thoroughly analysed so that precautionary measures can Instructions, 2002.
be taken.11-12 6. Lloyd’s Register of Shipping (LRS). Rules and Reg-
ulations for the Classification of Naval Ships – Electrical
Engineering, 2001.
CONCLUSIONS 7. Det Norske Veritas (DNV). Rules for ships – July
This paper deals with analysing and solving the mathemati- 2001 – Pt 4, Machinery and Systems – Main Class, 2001.
cal problem of active and reactive load sharing among the 8. Registro Italiano Navale (RINA). Rules and Regula-
synchronised generators of a ship’s electric energy system. tions for the Classification of Naval Ships – Electrical
Thus, both the first- and second-stage regulation during Engineering, 2001.
load sharing is presented via the solution of well-based 9. Hellenic Register of Shipping. Rules and Regulations

No. A13 2009 Journal of Marine Engineering and Technology 37


Reactive power sharing in ship energy systems with shaft generators

for the Classification and Construction of Ships – Part 6/ X: synchronous reactance (constant with the exception of
Electrical Installations - 1999, (Issue No 1). the short circuit transient period)
10. STANAG 1008: Characteristics of shipboard electri- V: network voltage (by proper regulations on V-Q rule,
cal power systems in warships of the North Atlantic Treaty this voltage is to remain constant)
navies, NATO, Edition 8, 21 Febr. 1994/Edition 9 under E: generator emf (it is regulated by the AVR intervention
ratification. to the field circuit)
11. Kanellos F, Hatzilau IK, Prousalidis J and Styvakta- : power angle reflecting mainly the active power
kis E. 2006. Simulation of a shipboard electrical network loading of the generator (t is regulated by the speed-
(AES) comprising pulsed loads, IMarEST Proceedings of governor intervention to the f-P rule)
Engine as a Weapon II, International conference, UK. xP : frequency droop
12. Kanellos F, Hatzilau IK and Prousalidis J. 2007. xQ : voltage droop
Investigation of voltage/frequency modulation in ship elec- fLk : no-load frequency of k-generator
tric networks with pulsed loads according to STANAG 1008 fN : nominal system frequency
design constraints, IMarEST Proceedings of All Electric fNk : nominal generator frequency of k-generator
Ship International Conference AES 2007, UK. fvk : full-load frequency of k-generator
Mt : total number of generators installed
M1 : total number of shaft systems installed
ACKNOWLEDGEMENTS M2 : total number of diesel generators installed
P1 : proportionate active power loading of k-generator
The authors wish to express their gratitude towards Elefsis
POk : active power operating point of k generator
Shipyards for all the support provided, including funding,
Q1 : proportionate active power loading of k-generator
technical data accumulation and field measurement per-
Downloaded by [195.29.81.98] at 22:45 26 July 2016

SG: shaft generator


formance. In addition, they are indebted towards Dr John
SC: shaft (rotary) condenser
Dermentzoglou for his assistance on setting-up the simula-
SS: shaft system comprising a shaft generator with its
tions in the MATLAB environment.
associated condenser
VN: nominal system voltage
VNk: nominal terminal voltage of k-generator
p : tolerance set by standards for proportionate active
NOMENCLATURE
power load sharing in secondary regulation
R: armature winding resistance (varies slightly by the Q : tolerance set by standards for proportionate reactive
thermal load) power load sharing in secondary regulation

38 Journal of Marine Engineering and Technology No. A13 2009

You might also like