ARTIFICIAL INTELLIGENCE

SUBSTATION CONTROL
 ABSTRACT
Controlling a substation by a fuzzy controller speeds up
response time diminishes up the possibility of risks normally
related to human operations. The automation of electric
substation is an area under constant development Our research
has focused on, the Selection of the magnitude to be controlled,
Definition and implementation of the soft techniques,
Elaboration of a programming tool to execute the control
operations. it is possible to control the desired status while
supervising some important magnitudes as the voltage, power
factor, and harmonic distortion, as well as the present status. The
status of the circuit breakers can be control by using a knowledge
base that relates some of the operation magnitudes, mixing status
variables with time variables and fuzzy sets .The number of
necessary magnitudes to a supervise and to control a substation
can be very high in the present research work, many magnitudes
were not included .To avoid the extensive number of required
rules nevertheless , controlling a substation by a fuzzy controller
has the advantage that it can speed up the response time and
diminish the possibility of risks normally related to human
operations.
 CONTENTS

 Introduction

 Plant Description

 Controller Design

 Experimental Results

 Fast response and diminished risk

 Conclusion
 AI AUTOMATES SUBSTATION CONTROL
Introduction
Electric substations are facilities in charge of the voltage transformation
to provide safe and effective energy to the consumers. This energy supply has to
be carried out with sufficient quality and should guarantee the equipment
security. The associated cost to ensure quality and security during the supply in
substations is high. Automatic mechanisms are generally used in greater or lesser
scale, although they mostly operate according to an individual control and
protection logic related with the equipment itself and not with the topology of the
whole substation in a given moment.
The automation of electric substation is an area under constant
development. Nevertheless, the control of a substation is a very complex task due
to the great number of related problems and, therefore, the decision variables
that can influence the substation performance. Under such circumstances, the
use of learning control systems can be very useful.
Many papers on applications of artificial intelligence (AI) techniques to
power system have been published in the last year. The difficulties associated
with the application of this technique include:
 Selection of the magnitude to be controlled
 Definition and implementation of the soft techniques
 Elaboration of a programming tool to execute the control operations
 Selection, acquisition and installation of the measurement and control
equipment
 Interface with this equipment and
 Applications of the controlling technique in existent substations.
Our research has focused on the first three points, and the interest of
the present work is to expose the obtained result and to present them for
discussion. The objective is to show that it is possible to control the status of
circuit breakers (CB) in a substation making use of a knowledge, mixing status
variables with time variables and fuzzy sets.
Even when all the magnitudes to be controlled cannot be included in
the analysis (mostly due to the great number of measurements and status
variables of the substation and, therefore, to the rules that would be required by
the controller), it is possible to control the desired status while supervising some
 important magnitudes as the voltage, power factor, and harmonic distortion, as
well as the present status.
PLANT DESCRIPTION
The system under study represents a test substation with two 30KVA
three-phase transformers, two CBs, two switches, three current transformers,
and two potential transformers. It also contains an auto transformer (to regulate
the input voltage) as well as impedance to simulate the existence of a transmission
line. The input voltage are the same (220V), this characteristic was selected in

order to analyze the operation of the controller in a laboratory scale in a second

stage of the development of the present work.
Therefore, the first transformer increases the voltage to a value of 13.2KV, while
the second lowers it again to 220V. fixed filter, an automatic filter for the control
of the power factor and the regulation of the voltage, and three feeding lines with
diverse type of loads of different nature (including nonlinear loads) are
connected through CBs to the output bar. Figure. 1 shows the proposed outline
Since the control elements in substations are the CBs and switches, the
goal is to allow the control of the five selected CBs (in the output bar) according
to some configurations and measurements of the observation variables. Initially,
the computer-aided system will try to control the plant and will send alarm
signals when it can not find a solution, waiting for the human intervention. So it
will learn how the human operator reacts to the inputs and will generate the
corresponding behavior rules. In this way, the system will replace gradually the
human operator.

CONTROLLER DESIGN
DEFINITION OF N THE INPUT AND OUTPUT VARIABLES
There are a great number of variables that can be chosen to control a
substation. Nevertheless, a limited number of variables were selected for this
study.
The following input variables have been defined:
 Vout: Voltage at output bus, phase A (V).
 PF: power factor at output bus, phase A
 THDv(%):Total voltage harmonic distortion at output bus(%)
 Tv, t(s): Amount of time the voltage Vout is in range
[114.3V;119.5V](tolerance zone)
 Tv,nt(s): Amount of time the voltage Vout is below 114.3V(alert zone).
In selecting the variables, several aspects were kept in mind. For
example, the voltage influences in the connection and disconnection of loads
when its value leaves some ranges during certain time. These ranges are
represented in figure.2

A decrease of the voltage 7.5% below the nominal value is allowed

during a certain time (for example, 10 min), while from 10% on, the
maximum allowed time is much smaller (for example, 20 s). In case these
limits are exceeded, loads will be disconnected in an established preference
order. The reconnection of loads will occur when the voltage arrives at
values above the tolerance interval, i.e., and the normal interval.
On the other hand, the power factor and the total voltage harmonic
distortion influence the connection and disconnection of capacitor and
filters. These variables can be read by means of sensors and/or
transducers, including signal conditional accessories and directed to the
data acquisition card (DAC) by means of analog lines. In this example, the
currents and voltages in each phases are not kept in mind due to the
great number of rules that would be required by the controller.
The CBs were defined as status variables. The switches were not
included in this study, because their status is only important for switching
and not for control purposes.
The following status variables were defined:
 Df: Status of the CB connecting the fixed filter
 Dc: Status of the CB connecting the controlled filter
 D11,D12,D13: Status of the CBs connecting the load feeders 1,2, and3.
Each combination of status variables defines a topology and is an
input for the controller. Possible values for each status variable are
0(open) and 1 (closed).
Thus, what is intended to control is the moment when the filters
and the loads should be connected or disconnected by means of signals
that are sent to the CBs. The controlled filter, once connected, will
maintain its functionality as an automatic filter in dependence of the
present harmonic distortion over the time. In case of a disconnection of
some filter due to over currents, the controller can activate connection rules
after some time that can be freely defined before the controller starts.
The actions to carry out as a response to disturbances in the
measurements (values out of normal ranges) are dependent not only on the
present topology of the substation, i.e., on the values of the present status
variables. For example; the connection of the controlled filter can only
occur when the fixed filter is connected. Similarly, the disconnection of the
fixed filter can only occur after the controlled filter has been disconnected.
The status variables are input and also output variable. Each CB
will maintain its standard protection function against over currents. Since
 signal to these devices are sent by means of additional relays to have the
ability to be controlled in parallel.
The status variables are read by sensors and/or transducer and
connected to digital lines to a DAC. Digital lines to the CB relays also send
the outputs.
For the definition of sets, triangular and trapezoidal shape
functions were used.
The status variables were not fuzzified because their
measurement are not provided with uncertainty. They can only accept two
values:0 and 1.
The time the voltage is in tolerable and not tolerable ranges is
supervised through the event counters.
RULE SYSTEM
The syntax of each rule can be expressed for example follows:
IF
(V is Tolerable ) and (PF is Low) and(THD is Tolerable)and
(Tv,t is acceptable) and (Tv, nt is zero) and (present topology is 00110)
THEN
(Desired topology is 10110)
The topology is expressed as a five-digit binary number that refers to
the five CBs in the following order:
 First digit: CB of the fixed filter
 Second digit: CB of the controller filter
 Third digit: CB of the first load (priority one)
 Fourth digit: CB of the second load (priority two)
 Fifth digit: CB of the third load(priority three).
In simple words, the rule expressed above means: If the present
topology is 00110(i.e., both filter are disconnected and only load 1 and 2
are connected) and the following situation is found:
 Voltage is tolerable
 Power factor is low
 Total harmonic distortion is tolerable
 Voltage has been in a tolerable zone for an acceptable time
 Voltage has not yet entered a not acceptable zone then the desirable
topology is 10110, which means that we must switch in the fixed filter.
To establish the connection and disconnection rules of the loads, it
was attributed to load 1 the biggest preference and to load 3 the smallest.
The definition of the analog variables was carried out using the
following terms:
 For Vout, the fuzzy sets: NT(not tolerable), T(tolerable), H(height),
N(normal)
 For PF, the fuzzy sets: L(low), T(tolerable), H(height)
 For THDv, the fuzzy sets:L(low),t(tolerable), H(height)
 For Tv,t, the crisp sets:Z(zero),A(acceptable),NA(not acceptable)
 Figure 3 shows the defined sets for each input variable. In the present work a
preliminary calculation of the Maximum number of rules yields that 7,776 rules
are necessary in order to start the controller .However, this number could be
reduced notably by keeping in mind that, among the 32 available states , not all
can
Be considered as possible. This way, only 9 topologies have been found as
possible, decreasing to 2,187 the Maximum number of necessary rules so that the
controller can totally replace the human operator . Nevertheless , this number
was reduced with the inclusion of some initials rules .
INITIAL RULE BASE
In case there exists a knowledge base on the plant to be controlled , some rules
can be included as a starting point . The initial knowledge base can be defined
in such
a way that diminish the necessary number of rules for the controller to work
properly.
V PF TH Tv,t Tv,nt Present Desired
Dv state state
NT T U NA Z Z Z
NT U H U A A A
NT U U U NA NA NA
Fig. 4. Some of the first 144 rules as they are presented to the software

Layer Number of Neurons Type of Activation

function
1 Number of input sets Linear(input=output)
=20
2 2*(number of input Sigmoidal
sets)+1=41
3 Number of state Sigmoidal
variables=5
This happens in most cases where some situations are not possible. In the
presented study, 144 initial rules were included (Figure.4), all in form of
extended rules (including one term U, which indicates that the attribute
can take any value).
 AUTOMATIC RULE EXTRACTION.
The rule base represents the knowledge base of the controller.
The proposed approach is able to start the operation with an empty or
uncompleted rule base.
During the inference process, the membership degree
corresponding to each column in the rule is calculated. The type of the
output of this function in the universe of discourse depends on the variable
type and on the selected term.
The rule extraction takes place each time the controller does not
find any rule in the rule base with a fire degree bigger than zero, and, as
a result, an alarm requiring an operator action is sent.
INFERENCE MODULE
The controller outputs are decided by searching in the rule base.
In this step, called inference, the fire degree of each rule is calculated.
Since the consequence part in each rule only deals with status
variables whose values are crisp numbers(0 and 1), use of defuzzification
method is not necessary. Therefore, the controller output in each case will
be the consequent part of the rule with the biggest fire degree
OPERATON MODULES OF THE CONTROLLER
The controller operation can be carried out in two ways. Following
the typical way the controller run, it can be placed in operation with a
completed rule base, totally replacing the human operator. Nevertheless, it
can be started with an empty or incomplete rule base, which means that
it will activate a leaning mechanism. This way, the controller will complete
step by step the operation rules and at the same time will replace the
human operator. A representation of the controller function can be viewed
in figure 5.

For the disconnection of loads when the voltage diminishes, the

following steps can be processed as an automatic response of some devices:
 Automatic tap change in transformers
 Voltage regulation in the autotransformer.
 If all these possibilities have been tried and the voltage continues
being low, then operator outputs are processed.
To avoid interferance with this response, the execution of
controller outputs (i.e., status changes of the CBs) can be postponed each
time for one sampling interval, provided the inference in the next sampling
time yields the necessity of status changes
In case the system cannot find a satisfactory solution because no
rule could be fired, the system sends an alarm to the operator. He will
have some time to decide what action to take using the switch components
on the PC screen.
If the operator action does not arrive during this time limit (either
due to delays or to operator absence), then the controller will execute a
protection rule previously defined, which could be, for example, the total
disconnection of the substation. After saturation. Of the knowledge base or
after a certain operating time, the system will generate and train an
artificial neural network (ANN), in order to replace the rule base.

SUBSTATION OF THE INFERENCE ENGIN BY AN ARTIFICIAL

NEURAL NETWORK
According to the type and the number of initial rule as well as
the number of existent linguistic variables, the system will calculate the
maximum number of possible rule.
When arriving at this number or time limit, the system will start
a module, where a feed-forward ANN with a hidden layer will be
generated and trained with a codified rule base.
This way, the system will try to diminish the time controller
needs for the inference, avoiding searching in an extensive knowledge base.
Once network has been trained, the interference will be carried out by means of
network and not through the rule base.
Each training pattern is coded with binary digits(0,1). In case of analog variables
, these digits are the codification of the sets in each rule.
The network of the present example is a feed forward network built by
20 neurons in the input layer, 41 neurons in the hidden layer, 5 neurons in the exit
layer (table 1)

The training process is carried out via a back-propagation algorithm.

Stop criteria for the training is the total network error as well as a time limit. In
fact , interference by means of the rule base will finish when the network is
completely trained.
During the interference through the neural network, the analog
measurements will have to be processed initially through a membership function
module before being presented to the network. This step is necessary inn order to
codify the rules with membership degrees of each sets for all magnitudes, i.e.,
instead of having binary digits (0,1), the input pattern will be vectors of real
numbers in the range [0,1] that represent the membership degrees of the inputs
in each set.
 The results of the output layer are real numbers between zero and one; thus, the
controllers will round these results to an integer-type value in order to find the
proposed status for the CBs.

Experimental results.
To carry out the experiment, a software for the platform windows
9x/2000 using Delphi was elaborated. Signal generators for the analog input
variables were used.
The experiment was starting from the status 0011. During the first
400 measurments,243 actions could not be determined by the controller ,an
expert gave the answers i.e., ,and as a result,243 new controller rules were
extracted.

Figure 6 shows the status behavior of the plant during the first 400
sampling times with extreme simulated variations in the inputs. The status
frequency is shown in table2.
Table 2. Frequency of topologies obtained during the 400 system operations
Topology Frequency
0(=00000) 0.019
4(=00100) 0.009
6(=00110) 0.102
7(=00111) 0.046
20(=10100) 0.93
22(=10110) 0.056
23(=10110) 0.009

The following can be said about the comparison of the plant

behavior with and without the controller. This control system has been
designed for human operator replacement, i.e.,
 To decide about actions beyond the conventional automatic control
procedure, i.e., action for which a human operator is always needed. Since
the implemented controller responds according to a knowledge base
extracted by human operations, the topology change will always be the
same (with or without the controller). Nevertheless, the difference and
advantage due to the use of this controller reside in:
 Speeding up response time
 Avoiding operation mistakes
 Gradually replacing a human operator.
The network was trained with a group of 1,377 rule. It was
found that, for activation functions of lineal type (first layer) and sigmoidal
type (hidden and last layer), an approximate time of 2 hours a 455 MHz
Pentium-based PC was sufficient to obtain approaches with errors smaller
than 0.1 for each output neuron. Since the final result of the
approximation by means of the network is based on whole rounds to zero
or one, this level of accuracy was found acceptable.
FAST RESPONSE AND DIMINISHED RISK
For the control of a substation by means of the connection
of devices for improving its performance, it is necessary to keep in mind
not only the measurement of the electric magnitudes but also the status of
some control devices that define their topology.
Control and protection devices governed by individual
decision methods that allow the substation operator when needed can be
coordinated from an upper supervisory level. It is possible to follow a
preference criterion for the disconnection or connection of CBs (in this
case for load feeders and filters) when using a controller based system.
Controllers can be used for rule extraction in a first
working stage. Control systems that do not require instantaneous responses
in an initial stage, can be designed for learning the operator action and
constructing a decision table for total replacement of the human operator
in the future.
Even when the number of rules for controlling the
substation is very high, it is possible to obtain these rules automatically by
means of a fuzzy controller. After a certain operating time (which depends
on the initial knowledge base as well as on the status variables and the
fuzzification of the inputs), the inference process through a rule base can
be replaced by an approximation via an ANN to diminish response time.
The number of necessary magnitudes to supervise and to
control a substation can be very high. In the present research work, many
magnitudes were not included to avoid the extensive number of required
rules.
Nevertheless, controlling a substation by a fuzzy controller
has the advantage that it can speed up the response time and diminish the
possibility of risks normally related to human operations.
Bibliography
 A,g.Kima, et l., “Cost of power quality problems in large industrial customers “,
PQA’93.
 K.S. Fu, “learning control systems and intelligence system” IEEE Trans .
Automat. Cont.,
 K.Y.Lee,”Current trend and the state of the art in intelligent system application
to power systems ”, 1999 international Conference on Intelligent systems
Application to power systems (ISAP), Rio de janeiro, brazil, Apr,1999.
 J.A.Jardhini,”Automation in power plants and high voltage
substations”,Editorial EB/USP, 1st edition.