You are on page 1of 4

Artificial Neural Networks applied to Measurements Rotor Angle

in a Synchronous Generator

ELEONORA DARIE
Electrotechnical Department
Technical University of Civil Engineering
B-dul Pache Protopopescu 66, Sector 2, Bucharest
ROMANIA
eleonora_darie@yahoo.com

COSTIN CEPIŞCĂ
Elcos Proiect SRL Bucharest
ROMANIA
costin@wing.ro

EMANUEL DARIE
Engineering Department
Police Academy
Aleea Privighetorilor 1A, Sector 1, Bucharest
ROMANIA
edarie@yahoo.com


Abstract: - This work analyzes the advantages of Artificial Neural Networks applied in development of a technique for
estimation rotor angle and speed, transient stability assessment and control in real-time in Synchronous Generator. Two
multi layered feed-forward Artificial Neural Networks are used for this technique. One is used for the estimation of
rotor angle and another for the estimation of rotor speed. The validation has been made by simulation in a power
system because techniques for the direct measurement were not available.

Key-Words: - Artificial Neural Networks, Control, Modeling, Multilayer Perceptron, Phasor Measurements Units, Synchronous
Generators, Transient Stability.

1 Introduction
Power system security assessment consists of evaluating
the ability of the system to face various disturbances and
of proposing appropriate remedial actions able to counter
its main weaknesses, whenever deemed necessary [3].
Power system security covers a wide range of aspects,
usually subdivided into static and dynamic phenomena.
Power system stability currently refers to the dynamic
part of security.
The rotor angle and speed of the synchronous generator
are the most important reference quantities in power
system dynamic security assessment and control. As
economic considerations continue to demand the
operation of power systems closer to their stability
limits, there is an increasing need for reliable and
accurate means to determine limiting operating
conditions. There are obvious differences between the
real-time stability prediction problem and offline
stability assessment. In conventional offline transient
stability assessments, the critical clearing time (CCT) is
to be found; in the prediction problem, the CCT is not of
interest. Instead, one can monitor the progress of the
transient in real-time thanks to the technique of phasor
measurements [4].
The current and potential applications of Phasor
Measurement Units (PMU) have been well documented
in [1], [4], [6]. An emerging application of this
technology is to track the state of the system
immediately following a transient event to select an
appropriate remedial control action [2].
A possible use of PMU measurements can be made to
predict a developing transient and initiating important
relays, or other control actions such as generation
tripping [2], load shedding [3], and Flexible AC
transmission system (FACTS) devices [1].


2 Problem Formulation

This work indicates Artificial Neural Networks (ANN)
in order to estimate rotor angles and speeds based on
real-time phasor measurements.
COMPUTING and COMPUTATIONAL INTELLIGENCE
ISSN: 1790-5117 193 ISBN: 978-960-474-088-8
Firstly the technology of phasor measurements to
estimate rotor angles and speeds following by the
development of the neural networks for angle and speed
estimation.


2.1 Estimation rotor angles and speeds from
Phasor Measurement Units (PMU)
PMU are power system devices that provide
measurements of real-time phasors of bus voltage and
line currents. A number of PMUs are already installed in
several utilities around the world for various applications
such as monitoring, control, protection, and state
estimation. The capabilities of a PMU are indicates in
Figure 1.













The measurement set is composed of the bus voltage
magnitude V
B
and angle θ
B
, as well as the line and
injection currents magnitude and angles, I
1
,I
2
, I
3
, I
L
, θ
1
,
θ
2
, θ
3
, θ
L
.
The rotor angles and speeds of the synchronous
generators are the most important quantities in power
system transient stability assessment and control.
PMU measured quantities are electrical variables that
may experience fast changes unlike rotor angle which is
a mechanical variable. PMU measured quantities can
experience discontinuity under switching in the electrical
network.
Wrong or noisy rotor angles and speeds may result in
wrong transient stability prediction and wrong
determination of control actions [5].
The way to compute rotor angles from phasor
measurements is to rely on the classical generator model
and relate phasors to reactances (step-up transformer,
generator) to get rotor angles [7], [8].

I t t V t t
I X j V E θ θ δ < − < = < ' ' . (1)
where E′ is the constant voltage, V
t
is generator
terminal voltage, X′ transient reactance, and I
t

generator terminal current.
The rotor speed can be approximated as:

t
t t
t

− +
=
) ( ) 1 (
) (
δ δ
ω . (2)

All these provided that MV (medium voltage) generator
voltage and current phasors are available. In more
general situations phasor measurements are not taken
directly from generator buses. In this case, for algebraic
relation of measured voltages V
m
and the generator
(internal) voltages and currents, the reduced admittance
matrix Y
bus
can be solved for the generator internal
voltages,

=

m
g g
V
V
Y Y
Y Y I
22 21
12 11
0
, (3)

where V
m
are the measured voltages, V
g
are the generator
internal voltages and I
g
are the generator internal currents.
The rotor angle is a nonlinear function of the machine
terminal variables and the main idea is to employ a pattern
recognition scheme to map the patterns of inputs (terminal
variables measured by a PMU) to the required rotor angle.
To realize the mapping of the machine terminal variables
measured by a PMU to the rotor angle we use the multi-
layer feed-forward ANN. Multi-layer feed-forward
ANNs with back propagation supervised learning have
several advantages over conventional computing
methods. Those advantages are robustness to input and
system noise, learning from examples, ability to
memorize, handling situations of incomplete information
and corrupted data, and performing in real-time.


2.2 Artificial Neural Networks
An ANN is characterized by its architecture, training or
learning algorithms and activation functions.
The architecture describes the connections between the
neurons. It consists of an input layer, an output layer and
generally, one or more hidden layers in-between.
Figure 2 illustrates one of the commonly used networks,
namely, the layered feed-forward ANN with one hidden
layer.
The layers in these networks are interconnected by
communication links that are associated with weights
that dictate the effect on the information passing through
them. These weights are determined by the learning
algorithm.








Fig. 1. Phasor measurements from a PMU.

COMPUTING and COMPUTATIONAL INTELLIGENCE
ISSN: 1790-5117 194 ISBN: 978-960-474-088-8

Fig. 2. Layered feed-forward ANN.

The output of node j in the hidden layer is given by the
following equation:

|
¹
|

\
|
+ ⋅ =

=
n
i
j i j i j
b x w g h
1
. (4)

and the output of the network is given by equation:

i
k
i
i
h o w y ⋅ =

=1
. (5)

where w
ij
, are the weights connecting the inputs to node j
in the hidden layer, b
j
is the bias to the node, and wo
i
are
the weights from the hidden to the output layer.
Depending on the learning algorithm the ANNs can be
categorized as:
Fixed Weight ANNs: these do not need any kind of
learning.
Unsupervised ANNs: These networks are trained
(weights are adjusted) based on input data only. The
networks learn to adapt using experience gained from
previous input.
Supervised ANNs: These are the most commonly used
ANNs. In these networks, the system makes use of both input
and output data. The weights and biases are updated for
every set of input/output data. The Multi-Layer
Perceptron (MLP) falls into this category.
The activation function relates the output of a neuron to
its input based on the neuron's input activity level. Some
of the commonly used functions include: the threshold,
piece-wise linear, sigmoid, tangent hyperbolic, and the
Gaussian function.
The learning process of the Multilayer Perceptron (MLP)
network involves using the input-output data to
determine the weights and biases.
One of the techniques used to obtain these parameters is
the back-propagation algorithm. In this method, the
weights and biases are adjusted to achieve a minimum
mean square error between the network output and target
value.
There are two different ways in which this algorithm can
be used: incremental mode and batch mode. In the
incremental mode the weights and biases are updated
after each input is applied to the network. In the batch
mode the weights and biases of the network are updated
only after the entire training set has been applied to the
network. The batch mode is used in this work.


3 Problem Solution
The purpose of the ANNs is to estimate the rotor angle
and speed of a synchronous machine using voltage and
current measurements, which are obtained by PMUs. On
use two different neural networks: one to estimate the
rotor angle (ANN1) and another to estimate the rotor
speed (ANN2).


3.1 Input Selection
The inputs to the neural network ANN1 are the voltage,
current, angle of voltage and angle of current at the extra
high voltage bus, at time instants t, t-1 and t-2, totaling
12 inputs.
The output of the neural network model consists of one
neuron representing the rotor angle for a specific
operating condition,

{
} ) 2 ( ), 1 ( ), ( ), 2 ( ), 1 (
), ( ), 2 ( ), 1 ( ), ( ), 2 ( ), 1 ( ), ( ) (
− − − −
− − − − =
t t t t t
t t i t i t i t v t v t v f t
i i i v v
v
θ θ θ θ θ
θ δ
. (6)

where v(t) and i(t) are the positive sequence terminal
voltage and current at the time t, v(t-1), v(t-2), i(t-l) and
i(t-2) are the voltage and current at the time t-1 and t-2,
θ
v
and θ
i
are the voltage and current angles at the same
time instants.
For ANN2, on used the same inputs as with ANNl, with
three inputs added, the rotor angle obtained from the
output of ANNl at time instants t, t-1 and t-2.
For this reason the number of inputs for ANN2 is 15.
The output of the ANN2 consists of one neuron
representing the rotor speed as illustrated in Figure 3.






COMPUTING and COMPUTATIONAL INTELLIGENCE
ISSN: 1790-5117 195 ISBN: 978-960-474-088-8


Fig. 3. The output of the ANN2 consists of one neuron
representing the rotor speed.


3.2 Selection of ANN
The ANN used are of the multi-layer feed-forward type, with
one hidden layer.
Figure 4 represents the multi-layer feed-forward network
























The number of units in the hidden layer is numerical
determined, from studying the network behavior during
the training process taking into consideration some
factors like convergence rate, error criteria, etc.
In this regard, different configurations were tested and
the best suitable configuration was selected based on
the accuracy level required.
The number of hidden units for the ANN1 is 40 and the
number of hidden units for ANN2 is 35.
Tangent hyperbolic activation function is used for these
units, while linear activation function is used for output
neurons for both of ANNs.
The neural networks were trained off-line.


4 Conclusion
This work analyzes the advantages of the Artificial
Neural Networks applied in development of a technique
for estimation rotor angle and speed for a synchronous
machine.
This analysis includes two ANNs, one to estimate rotor
angle and another, that include estimated angle.
The use of ANNs for this purpose offers attractive way
to cope with these new requirements.


References:
[1] A. Diu, L. Wehenkel, Examine - Experimentation
of a Monitoring and Control System for Managing
Vulnerabilities of the European Infrastructure for
Electric Power Exchange, IEEE/PES Summer
Meeting, 2002, Chicago, USA.
[2] M. Pavella, D. Ernst, Transient Stability of Power
Systems: A Unified Approach to Assessment and
Control, Kluwer Academic Publishers, 2000.
[3] M. Dicorato, A Corrective Control for Angle and
Voltage Stability Enhancement on the Transient
Time-Scale, IEEE Trans. on Power Systems, Vol. 15,
No. 4, 2000, pp. 1345-1353.
[4] A. G. Phadke, Synchronized Phasor measurements in
Power Systems, IEEE Computer Applications in
Power, Vol. 6, No. 2, 1993, pp. 10-15.
[5] C.H. Liu, Application of a Novel Fuzzy Neural
network to Real-Time transient Stability Swings
Prediction Based on synchronized Phasor
measurements, IEEE Trans. On Power Systems, Vol.
14, No. 2, 1999, pp. 685-692.
[6] V. Centeno, J. de la Ree, Adaptive out-of-step
relaying using phasor measurement techniques",
IEEE Computer Applications in Power, Vol. 6, No.
4, 1995, pp. 12-17.
[7] C.W. Liu, J. Thorp, Application of synchronized
phasor measurements to real-time transient
stability prediction, IEE Proc. Gener. Transm.
Distrib., Vol. 142, No. 4, 1995, pp. 355-360.
[8] S. Rovnyak, Predicting future behavior of transient
Events Rapidly Enough to Evaluate remedial
Control Options in Real time, IEEE Transactions on
Power Systems, Vol. 10, No. 3, 1995, pp. 1195-
1203.


Fig. 4. The model of layer feed-forward for angle
estimation.
COMPUTING and COMPUTATIONAL INTELLIGENCE
ISSN: 1790-5117 196 ISBN: 978-960-474-088-8

control. namely. The capabilities of a PMU are indicates in Figure 1.COMPUTING and COMPUTATIONAL INTELLIGENCE Firstly the technology of phasor measurements to estimate rotor angles and speeds following by the development of the neural networks for angle and speed estimation. (1) where E ′ is the constant voltage. A number of PMUs are already installed in several utilities around the world for various applications such as monitoring. The way to compute rotor angles from phasor measurements is to rely on the classical generator model and relate phasors to reactances (step-up transformer. and It generator terminal current. E'< δ =Vt <θt V − j X'It <θt I . θ3. The architecture describes the connections between the neurons. The rotor angles and speeds of the synchronous generators are the most important quantities in power system transient stability assessment and control. The rotor angle is a nonlinear function of the machine terminal variables and the main idea is to employ a pattern recognition scheme to map the patterns of inputs (terminal variables measured by a PMU) to the required rotor angle. [8].I2. and state estimation.2 Artificial Neural Networks An ANN is characterized by its architecture. ω(t) = δ (t +1) −δ (t) ∆t . an output layer and generally. The layers in these networks are interconnected by communication links that are associated with weights that dictate the effect on the information passing through them. Multi-layer feed-forward ANNs with back propagation supervised learning have several advantages over conventional computing methods. the reduced admittance matrix Ybus can be solved for the generator internal voltages. training or learning algorithms and activation functions. Wrong or noisy rotor angles and speeds may result in wrong transient stability prediction and wrong determination of control actions [5]. Vg are the generator internal voltages and Ig are the generator internal currents. the layered feed-forward ANN with one hidden layer. In this case. as well as the line and injection currents magnitude and angles.1 Estimation rotor angles and speeds from Phasor Measurement Units (PMU) PMU are power system devices that provide measurements of real-time phasors of bus voltage and line currents. I3. 1. Vt is generator terminal voltage. one or more hidden layers in-between. PMU measured quantities can experience discontinuity under switching in the electrical network. The measurement set is composed of the bus voltage magnitude VB and angle θB . for algebraic relation of measured voltages Vm and the generator (internal) voltages and currents. Figure 2 illustrates one of the commonly used networks. Ig  Y11 Y12 Vg   0  = Y Y  ⋅V  . The rotor speed can be approximated as: ISSN: 1790-5117 194 ISBN: 978-960-474-088-8 . I1. To realize the mapping of the machine terminal variables measured by a PMU to the rotor angle we use the multilayer feed-forward ANN. In more general situations phasor measurements are not taken directly from generator buses. All these provided that MV (medium voltage) generator voltage and current phasors are available. θ1.    21 22  m  (3) Fig. These weights are determined by the learning algorithm. 2. θ2. Those advantages are robustness to input and system noise. (2) 2. X ′ transient reactance. and performing in real-time. protection. Phasor measurements from a PMU. It consists of an input layer. PMU measured quantities are electrical variables that may experience fast changes unlike rotor angle which is a mechanical variable. generator) to get rotor angles [7]. ability to memorize. where Vm are the measured voltages. handling situations of incomplete information and corrupted data. learning from examples. θL. IL.

the rotor angle obtained from the output of ANNl at time instants t. In the incremental mode the weights and biases are updated after each input is applied to the network. and the Gaussian function. The batch mode is used in this work. The output of the ANN2 consists of one neuron representing the rotor speed as illustrated in Figure 3. which are obtained by PMUs.  i=1  and the output of the network is given by equation: (4) 3. Layered feed-forward ANN. θv and θi are the voltage and current angles at the same time instants. In these networks. angle of voltage and angle of current at the extra high voltage bus.i(t). are the weights connecting the inputs to node j in the hidden layer. with three inputs added. tangent hyperbolic. n  hj = g∑wi j ⋅ xi + bj  . The learning process of the Multilayer Perceptron (MLP) δ(t) = f {v(t). One of the techniques used to obtain these parameters is the back-propagation algorithm. totaling 12 inputs. ISSN: 1790-5117 195 ISBN: 978-960-474-088-8 . v(t-2). and woi are the weights from the hidden to the output layer. The Multi-Layer Perceptron (MLP) falls into this category. On use two different neural networks: one to estimate the rotor angle (ANN1) and another to estimate the rotor speed (ANN2). on used the same inputs as with ANNl. Supervised ANNs: These are the most commonly used ANNs. Unsupervised ANNs: These networks are trained (weights are adjusted) based on input data only. t-1 and t-2. The networks learn to adapt using experience gained from previous input.v(t − 2). Some of the commonly used functions include: the threshold. i(t-l) and i(t-2) are the voltage and current at the time t-1 and t-2. the weights and biases are adjusted to achieve a minimum mean square error between the network output and target value.1 Input Selection The inputs to the neural network ANN1 are the voltage. current. In this method.v(t −1). y = ∑woi ⋅ hi . In the batch mode the weights and biases of the network are updated only after the entire training set has been applied to the network. 3 Problem Solution Fig. at time instants t.θi (t − 2)} (6) where v(t) and i(t) are the positive sequence terminal voltage and current at the time t. The activation function relates the output of a neuron to its input based on the neuron's input activity level. . 2.COMPUTING and COMPUTATIONAL INTELLIGENCE network involves using the input-output data to determine the weights and biases.i(t −1).θv (t). bj is the bias to the node. sigmoid. There are two different ways in which this algorithm can be used: incremental mode and batch mode. the system makes use of both input and output data. θv (t −1). v(t-1).θv (t − 2).i(t − 2). For this reason the number of inputs for ANN2 is 15. t-1 and t-2. The output of the neural network model consists of one neuron representing the rotor angle for a specific operating condition. piece-wise linear.θi (t −1).θi (t). For ANN2. The weights and biases are updated for every set of input/output data. Depending on the learning algorithm the ANNs can be categorized as: Fixed Weight ANNs: these do not need any kind of learning. i=1 k (5) where wij. The output of node j in the hidden layer is given by the following equation: The purpose of the ANNs is to estimate the rotor angle and speed of a synchronous machine using voltage and current measurements.

IEEE Computer Applications in Power. 685-692. 2000. [5] C. ISSN: 1790-5117 196 ISBN: 978-960-474-088-8 . Gener. Adaptive out-of-step relaying using phasor measurement techniques". 3. In this regard. Vol. Fig. No. 1993. error criteria. 3. The neural networks were trained off-line. 1995. on Power Systems. etc. No. Vol. IEEE Trans. 1995. Diu.H. de la Ree.COMPUTING and COMPUTATIONAL INTELLIGENCE neurons for both of ANNs. pp. 14. Kluwer Academic Publishers. 142. with one hidden layer. pp. IEEE Trans. This analysis includes two ANNs. J. The number of units in the hidden layer is numerical determined. 1999. Liu. pp. 4. No. 10. Centeno. Thorp. On Power Systems. Distrib.. 11951203. while linear activation function is used for output References: [1] A. [2] M. 6. [8] S. Phadke. Application of synchronized phasor measurements to real-time transient stability prediction. different configurations were tested and the best suitable configuration was selected based on the accuracy level required. Vol. L. Dicorato. A Corrective Control for Angle and Voltage Stability Enhancement on the Transient Time-Scale. G. 4. Application of a Novel Fuzzy Neural network to Real-Time transient Stability Swings Prediction Based on synchronized Phasor measurements. from studying the network behavior during the training process taking into consideration some factors like convergence rate.2 Selection of ANN The ANN used are of the multi-layer feed-forward type. No. 12-17. 1345-1353. IEEE Computer Applications in Power. 2. 2000. Figure 4 represents the multi-layer feed-forward network Fig. pp. Ernst. Pavella. Examine . Vol. 4. Rovnyak. 15. IEE Proc. IEEE Transactions on Power Systems. Predicting future behavior of transient Events Rapidly Enough to Evaluate remedial Control Options in Real time. USA. D. [3] M. Liu. pp. 4 Conclusion This work analyzes the advantages of the Artificial Neural Networks applied in development of a technique for estimation rotor angle and speed for a synchronous machine. 2002.W. Tangent hyperbolic activation function is used for these units. J. The output of the ANN2 consists of one neuron representing the rotor speed. Chicago. Vol. The use of ANNs for this purpose offers attractive way to cope with these new requirements. IEEE/PES Summer Meeting.Experimentation of a Monitoring and Control System for Managing Vulnerabilities of the European Infrastructure for Electric Power Exchange. [7] C. [4] A. Vol. pp. The number of hidden units for the ANN1 is 40 and the number of hidden units for ANN2 is 35. 4. 6. 1995. No. Transient Stability of Power Systems: A Unified Approach to Assessment and Control. Wehenkel. 355-360. The model of layer feed-forward for angle estimation. 10-15. 3. [6] V. one to estimate rotor angle and another. Transm. 2. Synchronized Phasor measurements in Power Systems. that include estimated angle. No.