VISVESVARAYA TECHNOLOGICAL UNIVERSITY

JNANA SANGAMA, BELAGAVI – 590 018

A
Technical Seminar Report on
(18TES84)

“APPLICATIONS OF ANN IN POWER SYSTEMS”

Submitted in partial fulfillment of the requirements of the degree of

BACHELOR OF ENGINEERING
in
ELECTRONICS AND TELECOMMUNICATION ENGINEERING
For the academic year
2022 – 2023
By

DIVYA T A
(1MV19ET014)

Under the guidance of

Mrs, SAVITA A. HARKUDE,
Associate. Professor,
Dept. of ETE, SIR MVIT, Bengaluru

DEPARTMENT OF
ELECTRONICS AND TELECOMMUNICATION ENGINEERING

Sir M VISVESVARAYA INSTITUTE OF TECHNOLOGY

(Approved by AICTE New Delhi, Affiliated to VTU, Belagavi, ISO 9001:2008 Certified)
Off International Airport Road, Krishnadevaraya Nagar, Bengaluru – 562157
 Sir M VISVESVARAYA INSTITUTE OF
TECHNOLOGY
Off International Airport Road, Krishnadevaraya Nagar, Bengaluru – 562157

Department of Electronics and Telecommunication Engineering

CERTIFICATE

This is to certify that the Technical Seminar (18TES84) entitled “APPLICATIONS OF ANN IN
POWER SYSTEMS” is a bonafide work carried out by DIVYA T A (1MV19ET014) of SIR M.
VISVESVARAYA INSTITUTE OF TECHNOLOGY.The report is in partial fulfillment of the
requirements for the award of the degree of Bachelor of Engineering in Electronics Telecommunication
Engineering of the Visvesvaraya Technological University, Belagavi, Karnataka State, India, during the
academic year 2022-2023. It is certified that all corrections/suggestions indicated for Internal Assessment
have been incorporated in the report submitted to the Department. The Technical Seminar report has been
approved as it satisfies the academic requirement in respect to the work prescribed for the said Degree.

Signature of the Signature of the Signature of the

Guide Head of the Department Principal

Mrs, Savita A. Harkude Dr. E. Kavitha Prof. Rakesh S. G.

Associate. Professor, Dept. of ETE Prof. & Head, Dept of ETE, Principal,
Sir. MVIT, Bengaluru. Sir. MVIT, Bengaluru. Sir. MVIT, Bengaluru.

Name of the Examiners Signature with date

1.

2.
 DECLARATION

I DIVYA T A (1MV19ET014) student of eighth semester, Department of Electronics and

Telecommunication Engineering, Sir M Visvesvaraya Institute of Technology, Bangalore, declare that the
technical seminar entitled “APPLICATIONS OF ANN IN POWER SYSTEMS” has been carried out by
me and submitted in the partial fulfillment of the course requirements for the award of the degree of
Bachelor of Engineering in Telecommunication Engineering of Visvesvaraya Technological university,
Belagavi during the academic year 2022-2023.

Date: DIVYA T A
Place: Bengaluru (1MV18ET014)
 ACKNOWLEDGEMENT

We are grateful to the Founder and Board of Trustees of Sri. Krishnadevaraya

Educational Trust for having provided us with excellent facilities in the college
during the course to emerge as responsible citizens with Professional Engineering
Skills and moral ethics.

We are indebted to the Principal of our college, Prof. Rakesh S. G., for his
constant support, motivation and encouragement to excel in academics and carryout
project work.

We are thankful to our Head of the Department, Dr. E. Kavitha, for her kind
support, guidance and motivation during the B.E. degree course and especially
during the course of my project work.

We thank our Guide, Mrs, Savita A. Harkude , for her valuable guidance,
suggestions and Encouragement throughout my project work.

We are also thankful to all the staff members of the Department of Electronics and
Telecommunication Engineering and all those who have directly or indirectly helped
with their valuable suggestions in the successful completion of this project report.

DIVYA T A
(1MV19ET014)
 ABSTRACT

—The Artificial Neural Network (ANN) has become popular in the past few years for its
great performance. It is suitable for the nonlinear nature of dynamic systems. This report contains the use
of ANN in power applications such as voltage stabilization, fault detection, prediction, maximum power
point tracking, analytical load component, and power quality. ANN is expected to continue evolving in the
next few years, due to the availability of big data for energy applications and the use of deep learning
analysis.Here a recent literary survey is conducted on the use of ANN in power applications. Moreover,
strengths and weak points are shown in order to allow for appropriate selection of the type of ANN which
is as important as the selection of ANN parameters.
Applications of ANN in Power Systems

TABLE OF CONTENTS
CERTIFICATE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .- 1 -
ACKNOWLEDGEMENT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 2 -
ABSTRACT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .- 3 -

CHAPTER 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . -4-

1 INTRODUCTION
1.1 ARTIFICIAL NEURAL NETWORK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .- 6 -
1.2 POWER PLANTS AND CONTROL PROBLEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 9 -

CHAPTER 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .- 11 -

LITERATURE SURVEY

2 NEURAL NETWORKS IN POWER PLANTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 11 -

2.1 Identification and modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . - 11 -
2.2 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 11 -
2.3 Sensor Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 12 -
2.4 Monitoring, Fault Diagnosis and Predictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 12 -

CHAPTER 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .- 13 -

3 A PERSPECTIVE OF POWER SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 13 -

3.1 System Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 15 -
3.2 Transient stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 17 -
3.3 Load Forecasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 19 -
3.4 Fault Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 21 -

4 ADVANTAGES AND DISADVANTAGES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 22 -

5 CONCLUSION AND FUTURE SCOPE . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 23 -

6 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - 26 -

Dept. of ETE 5 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Chapter 1
INTRODUCTION

The objective of the Electric Power Industry is to supply electricity at the least possible cost with a constant
service quality. Among the factors that provoke difficulties in achieving this goal, the inherent variability of
the load and the fast growth of the demand are foremost, followed by requirements of clean environment,
weather, quality fuels, accelerated aging of the plants and fast changes in technology. Recently, promising
Artificial Neural Networks (ANN) approaches have been developed to solve problems in power plants and
power systems --tuning of controllers, process identification, sensor validation, monitoring and fault
diagnosis, in power plants, and security assessment, load identification, load modeling, forecasting and fault
diagnosis, in power systems.

Electrical power systems are among the largest and most complex systems faced by engineers today. They
are composed of many components of varying size and complexity: from low-power appliances to power
turbo generators. Operating, analyzing, and maintaining these systems is a truly arduous task. In the history
of electrical engineering, these have always brought numerous challenges to solve. First and foremost is that
of operations management. The term “operation” is commonly used in electrical power to define the
activities of managing the system in the present time. On the other hand, the term “planning” is used to
describe the actions in the future: experts use the term to refer to the steps needed to manage the system
under certain hypothesis that could occur in the long or medium term.

Planning and operations are always linked because the experiences of yesterday and today allow planning
for future operations. We can summarize that the word planning describes the methods or schemes that
allow any job, undertaking, or process to be organized in advance. Until a few decades ago, electrical
systems planning faced challenges quite different from those of today. These systems had easily manageable
generation equipment (you could define how and when schedule them and with what power) together with
loads that were fixed in space and, in many cases, also repetitive in time cycles.

Dept. of ETE 6 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

1.1 ARTIFICIAL NEURAL NETWORKS(ANN)

The term "Artificial Neural Network" is derived from biological neural networks that develop the structure
of a human brain. Similar to the human brain that has neurons interconnected to one another, artificial neural
networks also have neurons that are interconnected to one another in various layers of the networks. These
neurons are known as nodes.

Fig : 1 - The typical Artificial Neural Network looks something like the given figure.

An Artificial Neural Network in the field of Artificial intelligence where it attempts to mimic the network
of neurons that makes up a human brain so that computers will have an option to understand things and
make decisions in a human-like manner. The artificial neural network is designed by programming
computers to behave simply like interconnected brain cells.

We can understand the artificial neural network with an example, Consider an example of a digital logic gate
that takes an input and gives an output. "OR" gate, which takes two inputs. If one or both the inputs are
"On," then we get "On" in output. If both the inputs are "Off," then we get "Off" in output. Here the output
depends upon input. Our brain does not perform the same task. The output to inputs relationship keeps
changing because of the neurons in our brain, which are "learning."

Dept. of ETE 7 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Components of Artificial neural network

1) Neurons
ANNs are comprised of artificial neurons which are theoretically derived from biological neurons.Each
artificial neuron has inputs and produces a single output, which can be directed to numerous other neurons.
The inputs can be the feature values of a sample of external data, such as images or documents, or they can
be the outputs of other neurons. The outputs of the final output neurons of the neural net achieve the task,
such as image recognition. To determine the output of the neuron, the weighted sum of all the inputs is
computed, weighted by the weights of the connections from the inputs to the neuron. Consequently, a bias
term is added to this sum. This weighted sum is occasionally called the activation. This weighted sum is then
passed through a (usually non-linear) activation function for output generation.
2) Connections and Weights
“The network consists of connections, each
connection providing the output of one neuron as an input to another neuron. Each connection is assigned a
weight that represents its relative importance. A given neuron can have multiple input and output
connections.
3) Activation Function
Activation functions are functions used in neural networks to compute the weighted sum of input and biases,
which is used to decide whether a neuron can be fired or not. The activation function can be either linear or
non-linear depending on the function it represents.
4) Layers
“The neurons are typically organized into multiple layers, especially in DL. Neurons of one layer connect
only to neurons of the immediately preceding and immediately following layers. The layer that receives
external data is the input layer. The layer that produces the ultimate result is the output layer. In between
them are zero or more hidden layers. Single layer and un-layered networks are also used. Between two
layers, multiple connection patterns are possible. They can be fully connected, with every neuron in one
layer connecting to every neuron in the next layer. They can be pooling, where a group of neurons in one
layer connect to a single neuron in the next layer, thereby reducing the number of neurons in that layer.
Neurons with only such connections form a directed a-cyclic graph and are known as feed-forward networks.
Alternatively, networks that allow connections between neurons in the same or previous layers are known as
recurrent networks.
5) Hyperparameter
A hyper parameter is a constant parameter whose value is set before the learning process begins. The values
of parameters are determined using the process of learning. Examples of hyper parameters include number
of neurons, the number of hidden layers and batch size.
Dept. of ETE 8 Sir.MVIT,Bengaluru
Applications of ANN in Power Systems

1.2 POWER PLANTS AND CONTROL PROBLEMS

A power plant converts energy from nonelectrical to electrical form. Based on the energy transformation, the
plants are classified as fossil, nuclear, solar, geothermal, hydro, etc.

The main goal is to carry out this conversion in the best possible way. Safety, efficiency, reliability and
availability criteria are taken into account as references. A plant consists of several generating units which
work together to meet the electric load demand. For a fossil power plant, each unit consists of three basic
components: the boiler, the turbine and the generator. The complexity of the operation comes from the
variability of the load, and the high efficiency demanded over a wide range of operation. The main
difficulties for the control task then arise by the high coupling among the process variables and the non-
linearity of the process.

Fossil plants at the beginning were designed to operate on a base-load. Single input - single output
controllers were used. Control strategies using feed forward control action proved to satisfy the established
requirements at that time. However, the high growth of demand, the fuel crisis and the new restrictions for
environmental protection demanded different and more efficient types of operations for the electric power
plants.

In more detail, there are three types of operation:

 Base-load
 Load-following
 Cyclic

 Base-load operation: This mode of operation is in general, the most economical for the plants. Plants
operate at fixed demand, at maximum capacity, with variations no bigger than 15-20% of nominal
capacity over long periods of time (months). Plants in this mode basically require regulators or steady-
state controllers.

 Load-following operation: Plants in this operation mode, along with those operating in cyclic operation,
are responsible for maintaining the fixed frequency at 60 Hz. These plants absorb the fast and random
load changes that occur during the day. The Automatic Generation Control (AGC) plays a very
important role in these plants.

Dept. of ETE 9 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

 Cyclic operation: In this mode of operation, plants follow the slow load variation during the day.
Typically, they are old plants that have lost their economic advantages. Their main characteristic is the
ability to start up and stop at the required times.

In order to achieve the objectives of the cyclic and load-following operation modes, modern control
techniques are needed.

As a perspective and historical point of view, EPRI indicated the importance and necessity to begin to use
multi-input multi-output (MIMO) control techniques in power plants.

Modern multi variable techniques were classified as:

 Modeling and Simulation

 Multivariable Control Systems Design and Analysis
 Multivariable Estimation Design and Analysis

It is interesting to note that these techniques had proved to be useful in process control and aerospace
problems a long time ago.

Some of the reasons for the reluctance of the power industry were:

 Designed control systems have "proved" to be adequate

 Costs associated with simulation and modeling of multi variable control system were high
 Problems in design and implementation of such control systems were anticipated

At present, these arguments are no longer valid, and modern control techniques are providing some answers
to the new operation problems.

Dept. of ETE 10 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Chapter 2

LITERATURE SURVEY
2.1 Identification and modeling

1. One point of view about the use of ANN to deal with the modeling of nuclear power systems. The
special emphasis was placed on the interrelationship among sensor outputs. An ANN is used to predict
one or more of the sensor outputs. If there exist significant difference between the predicted and the
actual outputs, then something in the components, system or instrumentation has changed. Topics
described in the article: transient diagnostics, sensor validation, plant-wide monitoring, check valve
monitoring, and analysis of vibrations. Interesting approach for sensor validation function and detection
of incipient faults. Most of the works mentioned have been realized to demonstrate the feasibility of
some ANN approach.

2. A procedure using ANN to identify the nonlinear empirical model of a steam generator. A hybrid feed
forward/feedback ANN is used. The feed forward portion provides interpolation, while the feedback
portion enables representation of temporal variations in the non linearity's of the system. Simulation
results showed that empirical models can be effectively used to predict transient responses; this
approach appears to produce more accurate results than those reported in the literature based on
conventional nonlinear identification techniques.

2.2 Control

3. An investigation into tuning of power system stabilizers (PSS). The outputs of the neural net were the
desired parameters of the PSS. The inputs were the generator's real power and the power factor. Both
are measured on-line and are representative of the operating conditions. Thus, the PSS parameters can
be adapted in real time. Since this approach doesn’t require model identification, it is more efficient
than the self-tuning controllers, and therefore more suitable for real-time applications (27 references).

4. The results of a project which dealt with the problem of frequency control in power units under load
variations. A feed forward neural network is trained to control the input steam flow to the turbine so the
nominal value of frequency is restored. Two simulated examples are given: 1) a simple generator unit
feeding a power line, and 2) A two-area system. Results showed better performance with the neural
controller than current integral control.

Dept. of ETE 11 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

2.3 Sensor Validation

5. The feasibility of a method to verify the accuracy of the measurement of process variables using an
ANN. The method employs a set of process variables which are measured, related to the target process
variable, and used as inputs to the ANN. The ANN output is an estimate of the target process variable.
This estimate is compared with its actual value. The agreement establishes that the target instrument is
or is not operating properly. Application to the TMI-1 nuclear power plant took into account flow
meters in the two feed water flow loops. Multilayer feed forward and back propagation algorithms were
used. Simulation results showed correct readings of the target variable. Robustness of the system is
investigated.

2.4 Monitoring, Fault Diagnosis and Predictions

6. An approach based on ANNs and rule-based fuzzy expert system to monitor nuclear reactor systems. A
model-reference approach was utilized. The expert system executes the basic interpretation and
performs identification functions. The ANNs provide the model, classify general categories of system
behavior and generate membership functions. These membership functions can be utilized as inputs to a
fuzzy controller without the prerequisite of fuzzifying the measured input values. The system allows for
monitoring of the equipment performance and inferring of process variable values (virtual measuring).
Excellent robustness to noisy and faulty signals was reported. System description is suitable for control
applications.

7. The application of ANN methodology for recognizing transient events in a nuclear power plants.
Classifying events attempts to cover anticipated accidents. Eleven transients/accident events and steady
state operation for a Combustion Engineering System 80 pressurized water reactor (PWR) were studied.
Simulations were used to obtain the plant process parameters. Feedforward ANN and the back
propagation algorithms were used. The network was able to classify all transient events after 5 minutes
following event initiation. Reliable results were obtained for up to 5% sensor input noise.

Dept. of ETE 12 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Chapter 3

A PERSPECTIVE OF POWER SYSTEMS

While Power Plant topics mainly concern electric power generation, those corresponding to Power Systems
concern transmission, sub-transmission and distribution of electric power.

An Electrical Power System (EPS) can be seen as a network of interconnected components with the
purpose of:

* converting nonelectrical energy into electrical form

* transforming electrical energy into a specific form subject to strict requirements
* transporting electrical energy over long distances
* converting electrical energy into another energy form.
The (EPS) can be classified in five subareas:
1. - Generation
2. - Transmission
3. - Sub-transmission
4. - Distribution
5. - Use

Fig : 2 - Basic Electric Power and Structure of Power System

Dept. of ETE 13 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

1) Generation: represents the sources of Electrical Energy. It implies conversion of the energy from
nonelectrical form into electrical form. Power plants such as nuclear, geothermal, solar, are used for this
purpose.

Generation or Generating Station

The place where electric power produced by the parallel connected three phase alternators/generators is
called Generating Station (i.e. power plant).

2) Transmission: Considers the movement of the power from generation sites to the areas of use
(voltage range: 115-765 KV. and capacities: 100-4000 MVA).

3) Subtransmission: Saying " yesterday’s transmission is today’s sub transmission" (voltage range: 12-
69 KV. and capacities: less 100 MVA).

4) Distribution: Considers the individual circuits from the substation to the customer’s location (voltage
range for primary distribution: 2.4-12 KV., for secondary distribution: 120, 240, 480 V).

5) Use: All electrical devices which convert electrical energy into more usable forms such as light, heat,
sound, motion of mechanical force, etc.
Fig 3 : Transmission and distribution system key characteristics

Fig : 3 - Transmission and distribution system key characteristics

Among the applications of ANNs in power systems, alternative solutions to problems related with security
assessment, transient stability, load modelling, forecast and fault diagnosis have been proposed. To have a
complete idea of the kind of problems, the applications areas are outlined in the subsections below.
Dept. of ETE 14 Sir.MVIT,Bengaluru
Applications of ANN in Power Systems

3.1 SYSTEM SECURITY

"System security corresponds to the ability of the power system to withstand some unforseen but probable
disturbances with the minimal disruption of service or its quality.
Security assessment is a function that predicts the vulnerability of the system to possible events on a
realtime basis. Note the system being operated is different from that which was planned. Maintenance
requirements, forced outages, and different pattern loads make the differences. Thus the system security
levels are constantly changing.
Security assessment can be seen as an algorithm that forecasts the future evolution of the system, assesses
the probability of a security violation into determined period of time and determines if a preventive control
action should be taken to prevent hazardous operating conditions.

The security assessment is based on:

1. Knowledge of system dynamics
2. Measurement of system variables
3. Models for measurement uncertainty and system disturbances
4. Well-defined criterion for security.
Maximal admissible currents of the transmission gives the boundaries of the secure domain of the state
space (operating point are defined by a vector whose components are active and reactive power).

In the field of power system operation, system security can be seen as the counterpart of system reliability
in the field of system planning. In reliability assessment, the configuration of the system and the probability
distributions of individual component failures are given. The process of planning then consists of adding
components (e.g. generating units) or re-configuring the system in order to meet accepted system reliability.
Since this analysis is carried out via a planning mode, the chosen options can be implemented in due time.

Fig : 4 - EPS Operation States

Dept. of ETE 15 Sir.MVIT,Bengaluru
Applications of ANN in Power Systems

Operating States (see figure 1):

Normal: Voltage and frequency specifications are met without violating limits on any power device for all
system loads.

Emergency: Some operating limits are violated (over-voltage, very low frequency, overloaded lines)

Restorative: Portion of the system loads are not met (partial or total blackout) but the operating portion is in
a normal state.

Static security corresponds to those states where the transients following a disturbance have died out. In
these states the system has some limit violations that can only be tolerated for a short period of time (i.e.
overloads line, over-voltage conditions). If a corrective action is not possible, then the pre-disturbance state
must be classified as seriously insecure and should be avoided by preventive measures.

Dynamic security corresponds to the investigation of disturbances which may lead to transient instabilities
(e.g. loss of synchronism among generator). Dynamic security assessment is a very complex and difficult
task. The difficulties arise from a compounding of the following elements:

* The EPS is time-varying, highly non-linear, and very large scale.

* The inputs of interest are multidimensional statistical distributions of random events,
* It requires complex and time-consuming forecasts (since robustness evaluation of the system is carried
out
over a future interval) of the system behavior for all possible random events.
* The assessment must be executed on-line.

In the past several approaches have been proposed using dynamic estimation, statistical modeling of events
and perturbations, stochastic forecasting and stochastic contingency evaluation.

1. Security assessment

Security assessment is analysis performed to determine at what extent the system is safe from serious
interference to its operation.

ALGORITHM FOR CLASSIFICATION USING SOFM

In the classification phase, the network maps a vector with unknown features to the cluster where its closet
neighbors have been mapped to.
The algorithm is:
1. Present the input vector x.
2. Select the neuron with the weight vector closest to the
input vector.

Dept. of ETE 16 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Fig : 5 - Comparison between Conventional and Proposed method

3.2 TRANSIENT STABILITY

An electrical power system is a nonlinear, high order system which is subjected to both predictable and
unpredictable disturbances. These disturbances can also be classified as internal (random load) and external
(lightning strikes, wind).

Transient stability problems are those related to large-scale disturbances which cause the lose of
synchronization in a portion of the system, and in extreme cases, instabilities of the system.

Stability of power systems deals with the character of the electro-mechanical oscillations of synchronous
generators created by disturbances in the power system conditions.

Electromechanical oscillations represent exchange of energy among generator rotors (via the interconnection
network) which is caused by the instantaneous unbalance between generation and consumption of electric
power.

This imbalance is inherent to an AC power system and varies from low levels during normal changes of
system operating conditions, to relatively large levels in the case of major disturbances such as faults. In
both cases, the system stability depends on its capability to efficiently preserve synchronous operation of all
its parts and damp out electro-mechanical oscillations between them.

Dept. of ETE 17 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

When the disturbance is small, the system is confined to a small region around an equilibrium point. Linear
models are used and the stability properties of this equilibrium point are studied. If the disturbance is large,
the subsequent oscillatory transient will be of significant magnitude. And now the stability of the system is
determined following the trajectories related to the attraction region of the equilibrium point. In this case
both nonlinear model and theory are used.

Existing approaches to assess transient stability:

1. Numerical integration
2. The second method of Lyapunov
3. Probabilistic Methods
4. Pattern recognition
5. Neural Networks

1. Numerical methods

The system dynamics are described by a set of first order differential equations. These are solved during the
fault and post-fault period by numerical integration algorithms such as Runge-Kutta and its variations. Real-
time operation constraints make it difficult to apply this method to power systems.

2. Second Method of Lyapunov

The main idea is to evaluate the Lyapunov function at the instant of the last switching in the system. If the
value is smaller than a reference value then the post-fault transient process is stable.

3. Probability method

Probability of stability is defined as the probability that the system remains stable should the considered
disturbance occur. Security is measured in terms of the probability of a transition from normal state to the
emergency or "in extremis" state measured relative to a threshold.

4. Pattern Recognition (PR)

The transient stability studies using PR have focused on the selection of the initial system description,
feature extraction and on the design of classifiers.

Dynamic programming and modified search methods are used to improve the future extraction processes.
Several kinds of classifiers have been studied: adaptive hyperplane nearest neighbor, adaptive sphere and
prototype hyperplane. The last one has the best performance of the three.

On the selection of the initial system description four kind of parameters are used:

- the voltage magnitude and phase angle at each system bus.

- the active and reactive power of each generator plant.
- the active and reactive power of each bus load.
- the active and reactive power flow of all the lines.

These parameters refer to static system conditions. It’s mentioned that improvements in the classification
can be obtained using transient measurements such as individual kinetic energy of all generators.

Dept. of ETE 18 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

5. Neural Networks This recent technique is being used as pattern classifier and recognizer in this kind of
problems. Up to now, the literature has reported only simulation results taking into account real data.

3.3 LOAD FORECASTING

Load forecasting is an important task which allows the system operator:

 to schedule spinning reserve allocation;

 to decide for possible energy interchange with other utilities, and
 to assess system security

Accurate long-term and mid term electric load forecasting plays an essential role for electric power system
planning. It corresponds to load forecasting with lead times long enough to plan for long-term and mid term
maintenance, construction scheduling for developing new-generation facilities, purchasing of generating
units, and developing of transmission and distribution systems. The accuracy of the long-term load forecast
has a significant effect on developing future generation and distribution planning. An extensive
overestimation of load demand will result in substantial investment for the construction of excess power
facilities, whereas underestimation will result in customer discontentment.

The time horizon for long-term and mid-term forecasting ranges between a few weeks to several years.
Unfortunately, it is difficult to forecast load demand accurately over a planning period of this length. This
fact is due to the uncertain nature of the forecasting process. A large number of influential factors
characterize and directly or indirectly affect the underlying forecasting process; all of them are uncertain and
uncontrollable. Therefore, any long-term forecast, by nature, is inaccurate!

Two approaches have been traditionally suggested. One treats the load patterns as a time series signal and
predicts the future load based on time series analysis. Another one, regression approach, recognizes the
heavy dependency between the system load pattern and weather variables and finds a functional relationship.
The future load is then forecast by inserting the predicted weather information into the relationship.

Dept. of ETE 19 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Fig : 6 - Load forecasting depending on time series

Purpose of Load Forecasting

 For proper planning of power systems

 For proper planning of transmission and distribution facilities
 For proper power system operation
 For proper Financing
 For proper manpower development
 For proper grid formation
 For proper electrical sales

In the time series approach, inaccuracies of the prediction and numerical instabilities are general problems.
For the regression approach, conventional techniques assume a linear or piece-wise linear relationship
without justification; also the approach considers no variation in the spatio-temporal elements and it reports
averaged results, being in fact a non stationary process.

ANN’s seems to be a useful technique to solve these kinds of problems.

It is a technique used by companies to predict the power/energy needed to meet the demand and supply
equilibrium.
Irregular power supply problems occurs if production is not sufficient.
Excess power generation leads to loss
Neural network technique, back propagation approach is used for Load forecasting.
Factors affecting Load forecasting includes geographical factors, weather conditions, population growth ,
class of customers and so on.

Dept. of ETE 20 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Fig : 6 - Load forecasting flowchart

3.4 FAULT DIAGNOSIS

One of the major causes for power system outages is the failure of electrical equipment. Reliability, security,
and availability of the system can be improved by having good detection and diagnosis systems.

ANNs are being studied as classifiers of failures in electrical apparatus, and successful results have been
reported. For example, for failure diagnosis during transformer testings, ANNs are trained off-line with all
possible signals of simulated faults. A well known model of a transformer is used. After the training period
the ANN is ready to act as experienced evaluator. An extra advantage is that ANN not only can record the
failures but can also reproduce them [43]-1993. Another example is the study of underground oil switches.
ANNs proved to be a successful pattern classification systems to determine the condition (corrosive or non-
corrosive) of a given switch. The main advantage is that the diagnosis is carried out without interrupting the
service or dismantling the switch. Actual data are used to train the ANN.

To date, promising approaches based on ANN are applied in security assessment, load modeling, load
forecasting and fault diagnosis.

Dept. of ETE 21 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Chapter 4

ADVANTAGES AND DISADVANTAGES

Main advantages of using ANNs are:

• Very fast irrespective of the complexity of the problems

• On-line processing and classification
• Capability of handing with stochastic variations of the scheduled operating point with increasing data
• Implicit nonlinear modeling and filtering of system data automatically
• Massively parallel distributed structure and ability to learn

In spite of the plus points there are some negative points which are as under:

• Determination of various parameters like the number of hidden layers

• Number of neurons in the hidden layer
• Finding the optimal configuration of ANNs is a very time consuming process
• In contrast to conventional techniques or expert systems, which attempts to formalize knowledge and
develop partial quality system models, ANN do not provide a formal representation of the relation between
input and output data, it is therefore essential to test the ANN performance by statistical tools
• ANN suffers from the requirement of expert user in their design and implementation. They also suffer from
a lack of the formal model theory and mathematical rigors and so are exposed to the experts’ depth of
knowledge in problem definition.

However, ANNs for power system should be viewed as an additional tool instead of a replacement for
conventional or other AI based power system techniques. Currently ANNs rely on conventional simulations
in order to produce training vectors and analysis the training vectors, especially with noisy data. There are
some major issues to be tackled using ANNs for power system: training time, selection of training vector,
upgrading of trained neural nets and integration of technologies. ANN with its promise of adaptive training
and generalization deserves scope for further study.

Dept. of ETE 22 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Chapter 5
CONCLUSIONS AND FUTURE RESEARCH

Demonstrated success of ANN applications in a broad range of problems and the increasing interest of
researchers, vendors and electric power companies indicate the strength and applicability of the ANN
technology. In fact, power systems computing with neural nets is considered one of the fastest growing
field in power system engineering.

In order to provide an overview of the applications of ANN in Electric Power Industry, a representative
number of research projects have been outlined. These projects were categorized in two groups:
applications for Power Plants and applications for Power Systems.

To summarize all this work, the nature of problems, the type of neural nets used and their main functions
are shown in tables 1, 2 below.

Dept. of ETE 23 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Dept. of ETE 24 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Future Research In fundamental theory, two issues remain to be addressed thoroughly:

* Determination of a theoretical basis to design ANN based on a prior knowledge of the process or system in
question.
* Learning algorithms.

Dept. of ETE 25 Sir.MVIT,Bengaluru

Applications of ANN in Power Systems

Chapter 6

REFERENCE

[1] Jorge Ramirez Orrego; Rajarshi Roychowdhury; Mahesh S. Illindala, “Sensitivity Analysis Based
Identification of Key Parameters in the Dynamic Model” IEEE Transactions on Power
Systems ( Volume: 37): 02 | August 2021.

[2] Ms. Gagana M 1, Dr. Padma M C2 “Ms. Gagana M 1, Dr. Padma M C2 “Identification of Power
Systems Parameters” International Research Journal of Engineering and Technology (IRJET) e-ISSN:
2395-0056 Volume: 08 Issue: 06 | June 2021.

[3] Xin Zhang; Yawen Wang; Penghai Xie; Sajia Lin; Hanxing Luo; Hao Ling, “Power system transient
Stability control methods” IEEE/IAS Industrial and Commercial Power System Asia (I&CPS Asia):
02 | December 2021.

[4] Santoso, N.I. and O.T. Tan, "Neural-Net Based Real-time Control Capacitors Installed on
Distribution Systems," IEEE Transaction on Power Delivery, Vol.5, No.1: January 2019.

[5] Khadem, M., A. Ipakchi, F.J. Alexander, and R.W. Colley, "Sensor Validation In Power Plants Using
Neural Networks," : August 2020.

[6] Uhrig, R.E., "Potential Use Of Neural Networks In Nuclear Power Plants," Neural Network
Computing For the Electric Power Industry: Proceedings Of The 1992 INNS Summer Workshop, Editor
Dejan J. Sobajic, Lawrence Erlbaum Associates, Publishers, Hillsdale, N.J.: 09 | December 2018.

[7] Ostojic, D.R. and G.T. Heydt, "Transient Stability Assessment By Pattern Recognition In The
Frequency Domain," IEEE Transactions on Power Systems, Vol.6, No.1, 01 | February 2021.

Dept. of ETE 26 Sir.MVIT,Bengaluru