Hydro Turbin - 3 Paper
Hydro Turbin - 3 Paper
Hydro Turbin - 3 Paper
a r t i c l e
i n f o
Article history:
Received 28 October 2011
Received in revised form 13 February 2012
Accepted 14 February 2012
Available online 14 March 2012
Keywords:
Computer simulation
Hydro turbine models
Hydro governor models
Power system modeling
Power system stability
Free and open source software
PSAT
Power system simulation
FOSS
a b s t r a c t
This article studies accurate and detailed hydro turbine and governor models, and implements these models in Power System Analysis Toolbox (PSAT)a free and open source software. The implementation of these models in PSAT is approached using a general
methodology, which is described in detail. To evaluate the performances of the developed
hydro turbine and governor models, simulation studies are carried out on test systems of
different scales, from a single-machine innite-bus (SMIB) system to a larger system which
includes 20 generators. Further more, transient stability analysis and small signal stability
analysis are carried out to assess the performance of the implemented models.
2012 Elsevier B.V. All rights reserved.
1. Introduction
Hydro-power in Sweden accounts for one half net production of electrical power generation [1]. Furthermore, as one of
the most important renewable energy sources, the exploitation of hydro-power naturally attracts more and more attention
worldwide. In hydro power production systems, the functions of hydro turbine and governors cannot be neglected because
they participate in primary frequency control of power systems. To highlight this fact, Fig. 1 shows the relationship of a generators turbine and governor (indicated by the dashed block) with the overall power system.
Different power sources determine the physical features and categories of hydro turbine and governors, thus each type
requires its unique modeling. For example, time delays in hydro turbine and governor models are due to transient droop
compensation, pilot valves and gate servomotors. Additionally, the effect of water inertia in hydro turbines signicantly
inuences the governors requirements and must be compensated by a rate feedback, which must be reected in the modeling of these devices [2]. In this case, it is important that hydro turbine and governors are adequately modeled so that the
models can accurately represent the behaviors and capacities of hydro turbine and governors as close as possible. To examine model performance, the developed models are implemented in test power systems of different scales and their stabilities
are evaluated: i.e. if the system remains in equilibrium after being subjected to a disturbance [3]. To this end transient (large
signal) stability and small signal (small disturbance) stability analysis are frequently utilized as system stability analysis
approaches [4]; in this study both approaches are used.
85
Fig. 1. Functional block diagram showing relationship of the turbine and governor with the overall power system.
On the other hand, computer-based simulation tools are used by engineers for modeling and simulation to assess the stability dynamic and performance of the power system. With advances in computer software, there has been a large number of
power system simulation software packages developed, which can be basically divided into three categories: proprietary
software, open source software and free software [5]. Proprietary software are conceived by the general public to be
well-tested and computationally efcient. However, license agreements restrict their use by imposing different conditions.
Users are prohibited and generally have no possibility to modify the source code nor to distribute binaries (i.e. executables).
Additionally, they seldom allow to import or export data in different formats [6].
Nevertheless, free and open source software are usually freely (at no cost) distributed on line and are less cumbersome for
education and research purposes. More importantly, they allow users to change the source code, add new algorithms, or
implement new components. In addition, free software stands on an ethical pillar which aims to warrant intrinsic freedom
of computer uses that are jeopardized by proprietary software [7]. Despite theses advantages, there still exist some implementation challenges; particularly for new users. One of these challenges is that it is seldom documented in a detailed fashion the necessary approaches for implementing new components or algorithms in free and open source software. This
severely restricts free and open source software to be accepted and popularized by engineers, and also decreases individuals
cognition of the softwares exibility characteristics.
The objective of this paper is to develop hydro turbine and governor models and implement them in MATLAB-based Power
System Analysis Toolbox (PSAT)a free and open source software. To evaluate models performance, the relevant simulation
will be performed on a single-machine innite-bus (SMIB) system and another larger system, e.g. the KTH-NORDIC 32 system, for both transient and small signal stability analyses.
In view of the description above, the contributions of this paper can be summarized as follows:
To develop four detailed hydro turbine and governor models which can represent the behaviors of hydro turbine and governor precisely, e.g. water hammer effect.
To exhibit a systemic, clear and general approach for implementing new components in PSATa free and open source
software.
To illustrate a detailed description for how to implement new hydro turbine and governor models in PSAT.
To perform different system stability analysis approaches and their application in PSAT, e.g. transient stability analysis
and small signal stability analysis.
The remainder of this paper is organized as follows. Section 2 describes how to develop models for hydro turbine and
hydro governor, respectively, and presents an example model with its structure and realization description. In Section 3,
a general approach for implementing components in PSAT is described, as an example, this approach is illustrated with
the implementation of a hydro turbine and governor model in PSAT. Next, Section 4 shows simulation results and discusses
the performance of the implemented models in the SMIB system as reected by transient stability analysis and small signal
stability analysis. Implementation of turbine and governor models in a large power system is carried out in Section 5. Conclusions are drawn in Section 6 highlighting the contributions reported in this paper.
2. Hydro turbine and governor modeling
In this section the hydro turbine and hydro governor models which are implemented in PSAT are throughly described and
their realization diagrams are derived.
2.1. Hydro turbine modeling
As shown in the hydro turbine schematic diagram in Fig. 2, at steady state the net force on the water in the penstock is
equal to the rate of momentum change of the water according to Newtons Second Law [3]:
86
F net qL
dQ
dt
where q the mass density of water, L the length of penstock, and Q the volumetric ow rate.
The net force on the water in the penstock can be obtained by considering the pressure head [3]:
where Hs, Hl and H are static head, head loss, and head at turbine gate, respectively. Combining the Eqs. (1) and (2), the rate of
change of the volumetric ow rate can be described as:
dQ Hs Hl HAg
dt
L
Common bases (hbase and qbase) are used to normalize the above equation, resulting in
dq 1 hl hAghbase 1 hl h
dt
Tw
Lqbase
Lqbase
where T w Agh
is the water starting time or water time constant. Tw denes the time period it takes for water at the head
base
hbase to obtain the ow rate of qbase. hl is proportional to the ow rate square and penstock friction factor, as described by
2
hl = kfq2. The pressure head across the turbine is related to the owrate as h Gq 2 , where G describes the gate position from
0 (closed) to 1 (fully open) [3].
The mechanical power developed by the hydro turbine is proportional to pressure head and ow rate. As there is no-load
ow qnl, it should be subtracted from the actual ow rate. If damping effects are also taken into account [3], then
Pm At hq qnl DGDx
where At accounts for the difference in per units for both sides of the equation, as follows
At
turbine power MW
1
generator rating MVA hr qr qnl
The nonlinear turbine model encompassing the equations above is drawn in Fig. 3.
Linearizing the turbine characteristics around an operating point can aid in simplifying the turbine model. In hydro turbine systems, the mechanical torque/power increment m and the water ow increment q are dependent on the gate opening
increment g, the rotational speed increment n, and the head increment h. The following linearized equations are valid to represent the turbine for small perturbations around a steady state condition [2].
where
@q
@q
@q
; a12
; a13
@h
@n
@g
@m
@m
@m
; a22
; a23
@h
@n
@g
a11
a21
When the rotational speed deviation is small (n 0), the turbine transfer function which relates mechanical power increment in attaining gate position increment is
g
1 a11 T w s
87
Fig. 4. A simplied schematic of a mechanical hydraulic governor (adapted from Ref. [8]).
b
K1
a 1 Tps
The transfer function between the disturbing valve b and gate position change DG is
DG K 2
b
s
10
DG
K1K2
1
1
a
T p s 1s T g T p s 1s
11
where T1g K 1 K 2 . Tg is equal to the time in seconds for a 1 p.u. change in frequency to produce a 1 p.u. change in gate position.
Assuming the ow of dashpot uid through the needle valve is proportional to the dashpot pressure, the compensating dashpot transfer function is
88
Fig. 5. Transfer function block diagram of a typical hydro turbine governor model.
c
dT r s
DG 1 T r s
12
Through the action of the systems oating levers, the pilot valve input signal a is determined by the reference speed xref,
rotor speed x, permanent droop rDG, and temporary droop signal c:
a xref x rDG c
13
The relationship between the speed difference xref x and the gate position change DG can be described as shown in
Fig. 5. The rate limit may occur for large or rapid speed deviation, and the position limit may correspond to wide open valves
or the setting of a load limiter.
2.3. Hydro turbine and governor models
The hydro turbine and its governor are normally combined together for representation. However sometimes the output of
turbine governor is the derivative of gate position, which does not match the input of the turbine. Therefore, a desired gate
position reference is applied to add to the gate position derivative.
The report [9] presents four developed hydro turbine and governor models. Since PAST already has two turbine and governor models, Model 1 (thermal turbine and governor model) and Model 2 (simiplied turbine and governor model), the new
implemented models will be ordered sequentially afterwards. As shown in Table 1, the rst new model, namely, Model 3,
consists the typical hydro turbine and governor model. Model 4 contains a simple PI controller in front of Model 3 [10]. A
nonlinear turbine model [10,11] is used in Model 5 [12], and the nonlinear hydro turbine and governor model which is available in SPS [13] as HTG [14] and implemented in PSAT is called Model 6. As space is limited, only Model 3 is used as an example in this paper. Detailed structure and model realization for software implementation is described below.
Normally, a general device model can be described by a set of differential and algebraic equations (DAEs) of the form
x_ i f i xi ; yi ; xe ; ye ; pi ; ui ;
0 gi xi ; yi ; xe ; ye ; pi ; ui
14
where the sub-index i indicates specic device internal variables and the sub-index e indicates external variables from other
devices; f are the differential equations, g are the algebraic equations, x are the state variables, y are the algebraic variables, p
are device parameters assigned in the data le, and u are controllable variables (such as reference signals in control loops)
[15].
For the turbine and governor model (14) can be simplied:
x_ fx; y; u;
0 gx; y; u
15
where the most important state variables x are generator rotor angles d and generator rotor speeds x. The vector of algebraic
variables y includes system variables (e.g. bus voltage magnitudes), internal variables (e.g. internal input and output signals)
and output variables (e.g. mechanical power); and u are controllable and/or specied signals (e.g. speed references) [15].
2.3.1. Model 3typical hydro turbine and governor
The structural diagram of Model 3 is shown in Fig. 6. It consists of a typical hydro turbine governor model and a linearized
hydro turbine model. The output of turbine governor is the gate position derivative (DG), while the input of the turbine is the
gate position (G). Consequently, a position reference Gref, which is regarded as equal to Pref, is required between them.
To implement models in computer software and perform simulations, a set of state variables from a set of rst-order coupled differential equations is necessary. However, it is difcult to map the state variables in the structural diagram shown in
Fig. 6. The solution is redrawing the models by using only integrators and gain blocks. In this way, along the signal ow each
state variable is located behind each integrator, while the derivative of state variables labeling in front of each, as shown in
89
Table 1
Summary of governor models and turbine models combinations.
Model
Model
Model
Model
Model
3
4
5
6
Typical model
Typical mode with PI controller in front
PI controller combined with servomotor
PID controller combined with servomotor
Linear model
Linear model
Nonlinear model
Nonlinear model
Fig. 7. The number of state variables is equal to the total number of integrators in a system. This is an specic realization of
the mathematical model described by the block diagram in Fig. 6 suitable for computer implementation [16].
From the block diagram realization in Fig. 7, it is now possible to derive a set of differential and algebraic equations which
can be utilized for implementing the model in PSAT and to calculate the integrators initial values. The derived DAEs are
x_ g1
1
d
1
xref x r dDG xg3 xg1
TgTp
Tr
Tp
x_ g2 v
8
>
< xg1
v max
>
:
v min
if
if
if
90
x_ g3 DG
x_ g4
1
xg3
Tr
a13 a21
1
DG Pref
xg4
a11 T w
a211 T w
8
>
< xg2
DG Gmax Pref
>
:
Gmin Pref
Pm xg4
When the system is in steady state, x = xref, the rate of the gate movement v = 0, and the gate is xed as DG = 0. The initial values for Model 3 can be obtained by setting the derivatives of the differential equations to zero, as follows
xg4
a13 a21
P ref ;
a11
Pm a23 Pref
This initial values can be also determined by analyzing Fig. 7. When the system is in steady state, x = xref, all the derivatives of all state variables are zero. Consequently, point a and b are zero. Because of c b = a, c is accordingly equal to zero.
Similarly, point d is also zero as (xref x) and c are zero. With this recurrence method, it is possible to obtain all the state
variables initial values that are as same as those obtained from the DAEs above. This approach can be used to check if the
calculation using the DAEs is correct.
3. Implementation in free and open source software
PSAT is an open source MATLAB and GNU/Octave-based software package for analysis and design of electrical power systems [5,15,17,18]. It is a very exible and modular tool for power ow (PF), continuation power ow (CPF), optimal power
ow (OPF), small signal stability analysis (SSSA) and time domain simulation (TDS). In addition, a variety of static and dynamic models are provided. Both graphic user interface (GUI) and command line execution can be utilized for calculations
and simulations. Since MATLAB is a proprietary software, PSAT can also run on the latest GNU/Octave [19] release to realize its
full free software potential. The greatest advantage of PSAT is that it is free and open source, allowing the user to thoroughly
know about components inside structures and power ow calculations, and even to develop their own models.
PSAT is written by using classes and object-oriented programming techniques. Each device is dened by a class with attributes and methods [15]. The advantages of using classes include: easing the maintenance of the code, facilitating modularity
and reusing well assessed code. Moreover, class modularity allows one to quickly master the code. Normally, each type device in power system corresponds to one class, and all classes PSAT are documented in the PSAT manual [15].
3.1. A general approach for implementing models in PSAT
To implement a newly developed model in PSAT, rstly, the following common properties regarding the model (such as
the values of parameters, maximum amount of parameters) have to be determined [15].
1. con: device data in the form of a matrix. Each row denes a new instance of the device, while each column denes a
parameter. The content of the matrix con varies with each device.
2. format: a string containing the format of each row of the con matrix. Used for printing PSAT data to les.
3. n: total number of devices. This is also the amount of rows in con matrix.
4. ncol: maximum number of columns of the matrix con.
5. store: data backup. This is a copy of the con matrix. This attribute can be used for automatically modifying some input
data and can be useful for command line usage.
6. u: vector containing the status of the devices. 1 means on-line, 0 means off-line. Some devices can lack this attribute (i.e.
buses).
Then, for the model implementation, some common and particular methods are utilized to describe each model. As
shown below, the common methods include Jacobian matrix computation, initialization determination, differential equations implementation and so on [15].
1. add: adds one or more instances of the device.
2. base: converts device parameters to system power and voltage bases.
3. block: denes special operations of the device mask (used only for Simulink models).
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
91
A owchart illustrating a general approach for implementing models in PSAT and integrating them in power systems for
simulation is shown in Fig. 8.
3.2. Hydro turbine and governor models implementation
This section presents a hydro turbine and governor implementation example with Model 3. The parameters arranged in
Tg.con for Model 3 are illustrated in Table 2.
To add new turbine and governor models in the TGclass, the following les need to be modied. Some of these les are
particularly necessary only for turbine and governors, rather than common devices:
1. base: in PSAT, when dening the TG class data, the model droop (or gain) is given in p.u. with respect to the synchronous
machine power rating. During initialization, the droop (or gain) is converted to the system power base in order to keep
the same base with other variables in whole system, like electric and mechanical power.
It is quite easy to see why this conversion is necessary when considering two machines, called A and B. SA = 1 MVA and
SB = 100 MVA are the power rates of machine A and B, respectively. When assuming the droops are RA = RB on system base,
Fig. 8. A general approach for implementing models in PSAT and system simulation.
92
Variable
Description
Unit
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Generator number
Turbine governor type
Referenced speed
Pilot valve droop
Maximum gate opening
Minimum gate opening
Maximum gate opening rate
Minimum gate opening rate
Pilot valve time constant
Dashpot time constant
Permanent speed droop
Transient speed droop
Water starting time
@q/@h
@q/@g
@m/@h
@m/@g
Connection status
int
int
p.u.
p.u.
p.u.
p.u.
p.u.
p.u.
s
s
p.u.
p.u.
s
p.u.
p.u.
p.u.
p.u.
0.1
xref
Tg
Gmax
Gmin
vmax
vmin
Tp
Tr
r
d
Tw
a11
a13
a21
a23
u
the two machines will vary their power production by the same amount following a frequency variation. Clearly, this is
not what is requited. What is needed is that, for example, machine B provides 100 times the power variation provided by
machine A. Hence, the droop of the two machines, must be RA = RB on the machine base, and RA = 100RB on the system
base. Eq. (16) shows the droop and gain conversions.
1
Smachine
1
;
Rsystem
Rmachine
Ssystem
or K system
Smachine
K machine
Ssystem
16
The gain (droop) is dened as (the inverse of) the limit for s ) 0 of F(s), where F(s) is the transfer function of hydro turbine
governor. In this case, the rst mission is to calculate the transfer function and determine the gain (or droop). The transfer
function of Model 3 is given by
F 3 s
DG
xref x
1 Trs
T g T r T p s3 T g T r T p s2 T g rT r dT r s r
17
Parameter r is the droop of F3(s) when s ) 0, which is dened as the 11th parameter in Tg.con as shown in Table 2. The
following code transfers it from machine base to system base according to (16).
p.con (i,11) = p.con (i,11).Settings.mva./getvar (Syn,p.syn (i),mva);
The gate position limits and rate limits also need to be converted into system base because the whole loop of the governor
will operate on the system base when performing a simulation.
2. block: this le denes the prompt strings in parameter box for turbine and governor models graphic mask in Simulink
environment. All parameters in Tg.con have to be dened here. The following is an example for Model 3.1
prompts (idx5)=Maximum gate opening rate vmax [p.u.];
Minimum gate opening rate vmin [p.u.];
Pilot valve time constant Tp [s];. . .
Patical derivative of turbine torque with respect to gate position a23;
enables (idx6)=off; off;
idx5 and idx6 are indexes indicating the position of each parameter located in the parameter box. For this instance, all
the listed parameters are located at positions 616 in parameter box, while positions 17 and 18 are vacant for other models utilizing more parameters.
idx5 = [6 7 8 9 10 11 12 13 14 15 16]; idx6 = [17 18];
The block.m le in the TGclass programs the parameter boxes for turbine and governor models graphic mask in Simulink
environment. While fm_lib designs more general settings for the whole simulation library, which also contains some turbine and governor models specications. This fm_lib le does not belong to TGclass, but acts as an independent le in the
PSAT package. Within this le it is possible to nd the Tg block denition region.
1
. . . indicates some code are omitted as space limited. The same cases are applied for the following code fragments.
93
As the amount of models changes, relevant variables from MaskStyleString until MaskVariables need to be modied.
More strings should be added in MaskPromptString to make sure that the maximum parameters amount is the same
with that in block.m. These strings are only default, and await for specic strings set in the block.m le to cover them.
ModelHelp is used to give a simple description of the model.
3. dynidx: indexes for state variables in newly developed models are assigned. For Model 3, there are four state variables, so
a.tg1(i) = DAE.n + 1;. . .; a.tg4(i) = DAE.n + 4; DAE.n = DAE.n + 4;
4. fcall: differential equations introduced in Section 2 for Model 3 are revealed here.
% define the names of state variables
tg1 = DAE.x (p.dat3(:,1));. . . tg4 = DAE.x (p.dat3(:,4)); wref = DAE.y (p.wref (p.ty3));
% gate position limits. . .
%rate limits. . .
%differential equations
DAE.f (p.dat3(:,1)) = p.u (p.ty3).(p.dat3(:,5). (wref-DAE.x (p.dat3(:,17)) -(p.dat3(:,15)+p.dat3(:,14)).delta_G + p.dat3(:,15).p.dat3(:,8).tg3)- p.dat3(:,7).tg1);. . .
DAE.f (p.dat3(:,4)) = p.u (p.ty3).(p.dat3(:,11).p.dat3(:,12).(p.dat3(:,6)+delta_G)-tg4.
p.dat3(:,11));
p.dat3 is the index for Tg.con of Model 3 and will be stated in the setx0 le. DAE.x (p.dat3(:,1)) represents a state variable, which is the rst element in the matrix p.dat3, namely xg1 in Model 3s description in Section 2 (here is dened the
name as tg1). Similarly, other parameters index can be found in p.dat3 as well. While, DAE.f (p.dat3(:,1)) stands for the
derivative of state variable xg1, p.u (p.ty3) is a scale factor to change the parameters in per unit.
5. Fxcall: this le computes the Jacobian matrix fx, fy and gx (refer to (18) in Section 4). As (18) indicates, fx represents the
partial derivatives of function f with respect to state variables x, and fy the partial derivative of function f with respect to
algebraic variables, and gx the partial derivative of algebraic function g with respect to state variables x.
The following matrices indicate the parts of the Jacobian matrix fx which are related to the state variables in Model 3.
The rst element T p 1T g represents the partial derivative of x_ g1 with respect to x. These matrices are used for writing the
Fxcall le. The same approach is applicable for writing the Jacobian matrix fy and gx related to turbine and governor
models.
The Jacobian matrix is populated primarily with zeros, which coincidences with sparse matrices. Sparse data is by nature
easily compressed, and this compression almost always results in signicantly less computer data storage usage. PSAT uses
sparse matrices for creating the Jacobians, and therefore the Jacobian elements for new devices need to be implemented
using sparsity. For example, the following code
sparse (tg1, p.dat3 (:,17), u3.p.dat3 (:,5), DAE.n, DAE.n);
determines the partial derivative of x_ g1 with respect to state variables x. The Jacobian matrix fx is represented as DAE.Fx,
whose size is DAE.nDAE.n. The value in the xg1 row, x column is u3.p.dat3 (:,5), which is the partial derivative of x_ g1 with
respect to state variables x according to the DAEs of Model 3.
uG and uv are logically true (=1) only when the gate position and gate rate do not reach limits. For example, when uv is
logically true, the partial derivative x_ g2 with respect to xg1 is 1, otherwise it is zero; this is implemented with the following code.
u_G = delta_G < (p.con (p.ty3,5)-p.dat3(:,6)) & delta_G >(p.con (p.ty3,6)-p.dat3(:,6));
u_v = v < p.con (p.ty3,7) & v > p.con (p.ty3,8);. . .
DAE.Fx = DAE.Fx- sparse (tg1,tg2,u_G.u3.(p.dat3(:,14)+p.dat3(:,15)). p.dat3(:,5),DAE.n,DAE.n);. . .
DAE.Fx = DAE.Fx + sparse (tg2,tg1,u_v.u3,DAE.n,DAE.n);
6.
7.
8.
9.
94
a.dat3 = [a.tg1(a.ty3), % 1
a.tg2(a.ty3), % 2 . . .
Porder3, % 6
ap, % 7 . . .
Syn.omega (a.syn (a.ty3)), % 17
zeros (length (a.ty3),1), % 18
zeros (length (a.ty3),1)]; % 19
Moreover, the initial values of the state variables and function f are set here.
DAE.x
DAE.x
DAE.f
DAE.f
(a.dat3(:,1)) = 0;. . .
(a.dat3(:,4)) = a.u (a.ty3).a.dat3(:,12).a.dat3(:,6);
(a.dat3(:,1)) = 0;. . .
(a.dat3(:,4)) = 0;
10. subsasgn: assign turbine and governor properties for the newly developed models.
11. subsref: return turbine and governor properties for the newly developed models.
12. TGclass: add the newly developed models in the TGclass constructor. In addition, change ncol to be equal to maximum
amount of columns in the matrix con and modify the corresponding format.
Bus 1
Bus 3
L1: j0.1
Machine:
991MVA 20kV
Transformer:
20kV/20kV j0.1
L2: j0.1
95
1.002
Model 2
Model 3
Model 4
Model 5
Model 6
1.0006
Model 2
Model 3
Model 4
Model 5
Model 6
1.0015
1.0004
Rotor speed (p.u.)
1.001
1.0002
1
0.9998
1.0005
0.9995
0.9996
0.999
0.9994
0.9992
0
10
15
20
25
30
Simulation time (s)
35
40
45
50
0.9985
0
10
15
20
25
30
Simulation time (s)
35
40
45
50
96
Dx_ fx Dx fy Dy fu Du;
0 g x Dx g y Dy g u Du
18
Extracting Dy from the second equation above and substituting it into the rst one, then we obtain:
Dx_
!
!
fy g x
fy g u
Dx fu
Du;
fx
gy
gy
Dy
gx
g
Dx u Du
gy
gy
19
These equations just correspond to the linearization form of a system with input and output matrices A, B, C, D as:
Dy CDx DDu
20
where
A fx
fy g x
;
gy
B fu
fy g u
;
gy
C
gx
;
gy
D
gu
gy
x_ f x0 0
The eigenvalues ki can be computed from the A matrix according to
detkI A 0
21
k r jx
22
The damping ratio describes the rate of decay the system oscillations after a disturbance. It is given by
r
f p
r2 x2
23
and the frequency (which is called pseudo-frequency in the PSAT manual [15]) that can be observed during the transient is
x
2
24
A lower damping ratio implies a lower decay rate, and inuences the dynamic system behavior more than higher damping ratio. In this case, only two lowest damping modes are considered. Table 3 shows the small signal stability analysis results of the two lowest damping modes in SMIB system.
SMIB system response to a load change in PSAT
Model 2
Model 3
Model 4
Model 5
Model 6
0.9
Mechanical power (p.u.)
0.912
Mechanical power (p.u.)
0.9001
Model 2
Model 3
Model 4
Model 5
Model 6
0.91
0.908
0.906
0.904
0.9
0.8999
0.902
0.8999
0.9
0.898
0
10
12
14
16
18
20
1.9
1.95
2.05
2.1
G, Pm
97
G(t)
Pm(t)
t
Fig. 12. Water hammer effect.
Table 3
Summary of the linear analysis results of the two lowest damping modes in SMIB system.
Model in system
Frequency (Hz)
Damping ratio
Model 2
0.86233
6.6960
0.48048
0.12865
Eigenvalues
2.9684 5.4182j
5.4581 42,072j
Model 3
0.89782
6.6960
0.10574
0.12865
0.59983 5.6412j
5.4581 42,072j
Model 4
0.89804
6.6960
0.10394
0.12865
0.58970 5.6426j
5.4581 42,072j
Model 5
0.89116
6.6960
0.11766
0.12865
0.66340 5.5993j
5.4581 42,072j
Model 6
0.82586
6.6960
0.052253
0.12865
0.27151 5.1890j
5.4581 42.072j
As Table 3 illustrates, the SMIB system with Model 6 has a signicantly small damping ratio for the 0.8 Hz mode, which
can probably result in a high risk of instability. This results from its more complex nonlinear structure. Even though its low
level damping does not directly affect the transient stability response in SMIB system, it would have a larger effect when
implemented it in a more complex network. It is important to highlight that proper modeling of hydro turbine and governors
is necessary for small-signal stability analysis; otherwise, simple models not capturing the behavior of hydro mills (such as
Model 2) will result in over-optimistic damping calculations.
Fig. 13 depicts zero-pole map for hydro turbine and governor Model 3 and Model 2 in SMIB systems. It is important to
note from the zero-pole map of Model 2 that this system is a minimum phase system, as its poles and zeros are all in the
left half plane. On the other hand, Model 3 (same cases for Model 4Model 6) are nonminimum phase systems with one right
half plane zero. This is the essential reason which highlights that Model 2 has no water hammer effect. Nonminimum phase
system behaves phase lag compared to minimum phase system, which means they are slower at the start of the time-domain response. While this delay is just the property of water ow like shown by Model 3 (same cases for Model 4Model
6). However Model 2 has no capacity to represent this property of hydro turbine and governor in reality, and should only
be used for thermal or steam units.
4.3. Linear and nonlinear model validation comparison
Linear models can be used to simplify power system analysis and allow control design. This section veries the linear
model response to a step change at hydro turbine and governors speed reference, and nds out how well the linearized
model represents the behavior of nonlinear model in the linear operating region where the nonlinear model has been linearized. Both time-domain simulations run in PSAT, and Model 3 serves as hydro turbine and governor in SMIB system.
As shown in Fig. 14, the behavior of the linearized model is considerably close to the nonlinear model. It is demonstrated
that the linear model does capture the same dominant modes of the nonlinear model in the time-domain simulation. It is
relevant to note that if the Jacobians implemented in Section 3 were incorrect, there would be a discrepancy between the
linear and non-linear model responses. Consequently, we can conclude that linearized model is veried to represent nonlinear models. Moreover, linear analyses, such as small signal analysis are therefore accurate enough to provide the stability
information for nonlinear model which can be used for control design.
5. Implementation of turbine and governor models in large power systems
5.1. KTH-NORDIC32 system introduction and parameters in PSAT
The KTH-NORDIC32 system is depicted in Fig. 15. The overall topology is longitudinal; two large regions are connected
through considerably weak transmission lines. The rst region is formed by the north and the equivalent areas located in
98
0.5
Imaginary Axis
Imaginary Axis
0.5
4
1
6
2
1.5
0.5
0.5
1.5
1.5
0.5
0.5
Real Axis
Real Axis
1.0003
1.0002
1.0001
0.9999
0.9998
0.9997
0
10
15
20
time (s)
the upper part, while the second region is formed by the Central and the South areas located in the bottom part. The system
has 20 generators, 12 of which are hydro generators located in the North and the Equivalent areas, whereas the rest are thermal generators located in the Central and the South areas. In this paper the 12 hydro turbine and governors will utilize Model
2 or Model 3, while 8 thermal turbine and governors only use Model 1. As discussed in the previous section, Model 2 is
unsuitable for representing hydro turbine and governors, and is used here for comparison purpose only. There is more generation in the upper areas while more loads congregate in the bottom areas resulting in a heavy power transfer from the
northern area to the southern area, which easily leads to system oscillations [22].
5.2. Transient stability analysis
A three-phase fault is applied at BUS1011 at t = 5 s and removed at t = 5.02 s. The reason why the fault is set on this bus
is to decrease the fault inuence on critical generators, and at the same time so that the dominant power ow is not disturbed. This allows for a good comparison of the performances of the turbine and governor models implemented in this paper. Fig. 16a and b depict the response of the generators rotor speeds in the KTH-NORDIC32 System with Model 2 and Model
3, respectively.
Comparing these two simulations, we can easily nd out that the system with Model 2 recovers to steady state faster than
that with Model 3. Moreover, in the system with Model 3, there are oscillations with particular frequencies swing around a
steady value 1, even when the system already gets back to steady state. In fact this is a normal and common phenomenon
in large power system called system oscillations [2325].
99
Fig. 15. KTH-NORDIC32 test system (Taken from Ref. [22]). (The generators marked with panes are hydraulic generators, which utilize Model 2 or Model 3
as their hydro turbine and governors in KTH-NORDIC32 test system for simulation; the other generators utilize Model 1 as their steam turbine and
governors.)
Syn 1
Syn 1
1.0008
Syn 2
Syn 2
1.0006
Syn 3
Syn 3
1.0006
Syn 4
Syn 4
Syn 5
1.0004
1.0004
Syn 5
Syn 6
Syn 6
Syn 7
Syn 7
1.0002
Syn 8
Syn 8
1.0002
Syn 9
Syn 9
Syn 10
Syn 11
Syn 10
0.9998
Syn 11
Syn 12
Syn 12
Syn 13
0.9998
Syn 13
0.9996
Syn 14
Syn 14
Syn 15
Syn 15
0.9994
Syn 16
0.9996
Syn 17
Syn 16
Syn 17
0.9992
Syn 18
0.9994
20
40
60
80
100 120
time (s)
140
160
180
200
Syn 19
Syn 18
0.999
0
50
Syn 20
100
time (s)
150
200
Syn 19
Syn 20
100
SingleSided Amplitude Spectrum of Syn18 speed for KTHNORDIC32 system with Model 1&2
0.5
0.45
0.25
0.4
0.35
X= 0.73547
Y= 0.26203
|Y(f)|
|Y(f)|
X= 0.49744
Y= 0.18772
0.2
0.3
0.25
0.15
0.2
0.1
0.15
X= 0.49744
Y= 0.10061
0.1
0.05
0
X= 0.033569
Y= 0.022676
X= 0.030518
Y= 0.022327
0
0
X= 0.72327
Y= 0.044477
0.05
0.1
0.2
0.3
0.4
0.5
0.6
Frequency (Hz)
0.7
0.8
0.9
0.1
0.2
0.3
0.4
0.5
0.6
Frequency (Hz)
0.7
0.8
0.9
(b) FFT on 18
(a) FFT on 6
SingleSided Amplitude Spectrum of Syn18 speed for KTHNORDIC32 system with Model 1&3
12
speed
for KTHNORDIC32 system with Model 1&3
20
2
1.8
10
X= 0.49438
Y= 10.5394
1.6
1.4
X= 0.057983
Y= 1.3078
X= 0.49438
Y= 1.2849
X= 0.77515
Y= 1.0952
|Y(f)|
|Y(f)|
1.2
6
1
0.8
0.6
0.4
X= 0.057983
Y= 1.2773
X= 0.77515
Y= 0.36492
0.1
0.2
0.3
0.4
0.5
0.6
Frequency (Hz)
0.7
0.2
0.8
0.9
0
0
0.1
0.2
0.3
(a) FFT on 18
0.4
0.5
0.6
Frequency (Hz)
0.7
0.8
0.9
(b) FFT on 20
SingleSided Amplitude Spectrum of Syn6 speed for KTHNORDIC32 system with Model 1&3
6
X= 0.77515
Y= 5.5684
X= 0.49438
Y= 5.5338
|Y(f)|
2
X= 0.057983
Y= 1.3138
0
0
0.1
0.2
0.3
0.4
0.5
0.6
Frequency (Hz)
0.7
0.8
0.9
(c) FFT on 6
Fig. 18. FFT on rotor speed signals in KTH-NORDIC32 including Model 3.
It is difcult to identify the frequencies of system oscillations by looking at the time-domain simulation signals. The fast
Fourier transform (FFT) should be used to convert the time-domain signals into frequency domain, so that is possible to
determine the particular frequencies in the signals. Making sure that the signal for FFT is from a xed point time-domain
simulation, and that the smaller time step is, the higher resolution it can be obtained. Here a window from 5 to 200 s is
101
Frequency
Damping ratio
Model 2
0.49866
0.73218
0.035223
0.031801
Model 3
0.49362
0.77442
0.0019950
0.0082036
Eigenvalues
0.11043 3.1331j
0.14637 4.6004j
x18, d18
x6, d6
0.0061875 3.1015j
0.039918 4.8658j
x18, d18
x20, d20
selected form the x6,x18 signals in the system which includes Model 2; and 5200 s from the x18, x20 and x6 signals in the
system which includes Model 3 for FFT operation with 0.01 s time step.
Fig. 17 depicts the frequencies of rotor speed signals. There are two primary frequencies around 0.5 Hz and 0.7 Hz, and an
inconspicuous frequency about 0.03 Hz for both x6 and x18. As the frequency of oscillations caused by the fault and controlled by hydro turbine and governor (here for short we can call it turbine/governor dynamics) normally keep below
0.1 Hz, we can assert that the frequency 0.03 Hz in Fig. 17 belongs to the turbine/governor dynamics. The other frequencies
are due to system oscillations, which are extremely weak and disappear quickly, as shown in Fig. 16a. Similarly, Fig. 18 indicates the frequencies of system oscillations and turbine/governor dynamics. When hydro turbine and governor models are
used, the related turbine/governor dynamics as shown in Fig. 18 become more prominent, indicated by an increase energy
of the related mode at about 0.05 Hz.
Even though the system with Model 2 behaves faster in response to the fault, we cannot assert Model 3 is bad, because
Model 3 represent the real behavior and properties of hydro turbine and governors, as we discussed in Sections 4.1 and 4.2.
5.3. Small signal stability analysis
All eigenvalues for the KTH-NORDIC32 test system, either with Model 2 or Model 3, are located in the left half plane,
which indicates that the systems are stable. The two lowest damping modes in the KTH-NORDIC32 system, as done for
the SMIB system, are provided in Table 4. Moreover, which states are most associated with these two eigenvalues can be
directly gured out from the Eigenvalue Report in the Eigenvalue Analysis routine from PSAT. Looking at these frequencies, they are just the system oscillation frequencies as computed in the FFT operation. This means the two lowest decay rate
are just at the frequencies of system oscillations. It can therefore be asserted that the time-domain simulation results coincide with the small-signal stability analysis results, and thus, there is condence of the implemented models for small-signal
stability analysis.
6. Conclusions
This paper presented four developed hydro turbine and governor models and implemented them in PSATa free and open
source software. Moreover, detailed implementation steps are also provided. To evaluate models features, the performances
of the implemented models in two power systemsa single-machine innite-bus (SMIB) system, and the KTH-NORDIC 32
system were examined. The evaluation of a power systems performance is concerned with the stability of that system by
analyzing transient stability and small signal stability. Model validation demonstrates that the developed models have
the capacity of representing the behaviors of hydro turbine and governors in reality. Future work will focus on tuning parameters of hydro turbine and governors models in order to reduce the system oscillations [24]. Additionally, models performances in real time simulation environment [26] will be explored.
Acknowledgment
The authors gratefully acknowledge the aid of Prof. Federico Milano of the University of Castilla-La Mancha for his valuable comments and suggestions for implementing the new models described in this paper.
This Project was supported in part by the EIT InnoEnergy Collocation Center Sweden within the Smart Power Thematic
Area, Work Package 3.1: Smart Transmission Grids Real-Time Simulation Platform.
Luigi Vanfretti is supported by the STandUP for Energy collaboration initiative and the KTH School of Electrical
Engineering.
References
[1] International Energy Agency, Electricity/Heat in Sweden in 2008. <http://www.iea.org/stats/electricitydata.asp?COUNTRY_CODE=SE>.
[2] IEEE Committee Report, Dynamic models for steam and hydro turbines in power system studies, IEEE Transactions on Power Apparatus and Systems
(1973) 19041915.
[3] J. Machowski, J. Bialek, J. Bumby, Power System Dynamics: Stability and Control, second ed., John Wiley & Sons, 1997.
[4] P. Kundur, Power System Stability and Control, McGraw-Hill, 1993.
[5] F. Milano, An open source power system analysis toolbox, IEEE Transactions on Power Systems 20 (2004) 11991206.
[6] F. Milano, M. Zhou, G. Hou, Open model for exchanging power system data, Power & Energy Society General Meeting, 2009. PES 09. IEEE, 2009, pp. 17.
102
[7] R.M. Stallman, Free Software, Free Society: Selected Essays of Richard M. Stallman, second ed., Free Software Foundation, Inc., 2010.
[8] D. Ramey, J. Skooglund, Detailed hydrogovernor representation for system stability studies, Power Apparatus and Systems, IEEE Transactions on PAS89 (1970) 106112.
[9] W. Li, Hydro Turbine and Governor Modeling and Scripting for Small-signal and Transient Stability Analysis of Power Systems, Masters Thesis, KTH
Royal Institute of Technology, Stockholm, Sweden, September, 2011.
[10] I.C. Report, Hydraulic turbine and turbine control models for system dynamic studies, Power Systems, IEEE Transactions on 7 (1992) 167179.
[11] L.N. Hannett, J.W. Feltes, B. Fardanesh, Field tests to validate hydro turbine-governor model structure and parameters, Power Systems, IEEE
Transactions on 9 (1994) 17441751.
[12] T.V. Cutsem, Description, Modeling and Simulation Results of a Test System for Voltage Stability Analysis, Version 2, November, 2010.
[13] I. The MathWorks, SimPowerSystems 5.4, 2011. <http://www.mathworks.com/products/simpower/index.html?ref=productlink>.
[14] The
MathWorks,
Inc.,
Hydraulic
Turbine
and
Governor.
<http://www.mathworks.com/help/toolbox/physmod/powersys/ref/
hydraulicturbineandgovernor.html>.
[15] F. Milano, Power System Analysis Toolbox Documentation for PSAT, Version 2.1.6, May, 2010.
[16] P.M. DeRusso, R.J. Roy, C.M. Close, A.A. Desrochers, State Variables for Engineers, second ed., John Wiley & Sons, 1997.
[17] F. Milano, L. Vanfretti, J. Morataya, An open source power system virtual laboratory: the PSAT case and experience, Education, IEEE Transactions on 51
(2008) 1723.
[18] L. Vanfretti, F. Milano, Application of the PSAT, an open source software for educational and research purposes, in: Power Engineering Society General
Meeting, 2007, IEEE, 2007.
[19] J.W. Eaton, GNU Octave Manual, Network Theory Ltd., Bristol, UK, 1997.
[20] M.H. Chaudhry, D. Arsenie, P. Vlase, Applied Hydraulic Transients, second ed., Taylor & Francis, 1987.
[21] B.B. Sharp, D.B. Sharp, Water Hammer: Practical Solutions, Butterworth-Heinemann, 1996.
[22] Y. Chompoobutrgool, L. Vanfretti, Linear Analysis of the kth-nordic32 System. Internal Report, Tech. Rep., KTH Royal Institute of Technology, March,
2011. <http://www.vanfretti.com/Dr._Luigi_Vanfrettis_Website/Welcome.html>.
[23] F. Schleif, G. Martin, R. Angell, Damping of system oscillations with a hydrogenerating unit, Power Apparatus and Systems, IEEE Transactions on PAS-86
(1967) 438442.
[24] F. Schleif, A. Wilbor, The coordination of hydraulic turbine governors for power system operation, Power Apparatus and Systems, IEEE Transactions on
PAS-85 (1966) 750758.
[25] D.N. Kosterev, C.W. Taylor, Model validation for the august 10,1996 WSCC system outage, Power Systems, IEEE Transactions on 14 (1999) 967979.
[26] C. Dufour, J. Belanger, A pc-based real-time parallel simulator of electric systems and drives, in: International Conference on Parallel Computing in
Electrical Engineering, 2004. PARELEC 2004, pp. 105113.