seminar presentation on ann in psr

8/14/2019 Seminar Presentation on ANN in PSR

1/35

1

Department Of Information Technology

Guru Tegh Bahadur Institute Of Technology

ARITIFICIAL NEURAL NETWORK IN

POWER SYSTEM RESTORATION

By:By:

VANYAGUPTAVANYAGUPTA

(35/EEE/2006)(35/EEE/2006)EN.NO.0761324906EN.NO.0761324906


2/35

2

CONTENT

IntroductionObjective Of PSRWhat Is Neural Network(NW)

Uses of NWArchitecture of NWNetwork LayersPerceptronsCriteria for RestorationStages in Restoration Process

Guideline Of PSRRestoration StudiesANN Research PaperPrinciple of PSRProposed SchemeProposed Scheme

Conclusion


3/35

3

OBJECTIVE OF POWER SYSTEM

RESTORATION

Minimize power system outages, maintain load flowanalysis between load and generation

To reduce operating and maintenance cost, to maximizeprofit and to work equipment harder and longer.

Power system behavior under emergency conditions

The restoration process is also a function of pre-disturbance conditions, post-disturbance status and post-disturbance target systems.


4/35

4

INTRODUCTION

The question'What is a neural network?

is ill-posed.

A method of computing, based on the interaction ofmultiple connectedprocessing elements


5/35

5

WHAT IS A NEURAL NETWORK?

A Neural Network generally maps a set of inputs to a setof outputs

Number of inputs/outputs is variable. The Network itself

is composed of an arbitrary number of nodes with anarbitrary topology

Neural Network

Input 0 Input 1 Input n...

Output 0 Output 1 Output m...


6/35

6

Example for Neural Network

An ANN is configured for a specificapplication, such as pattern

recognition or data classification,through a learning process.

Learning in biological systems

involves adjustments to the synapticconnections that exist between theneurons. This is true of ANNs as well.


7/35

7


8/35

8

ARCHITECTURE OF NEURALNETWORKS

FEED-FORWARD

NETWORKS

Feed-forward ANNs (figure 1)allow signals to travel one wayonly; from input to output.There is no feedback (loops)

i.e. the output of any layerdoes not affect that samelayer.Feed-forward ANNs tend tobe straight forward networksthat associate inputs with

outputs.They are extensively used inpattern recognition. This typeof organisation is also referredto as bottom-up or top-down.


9/35

9

FEEDBACK NETWORKS

Feedback networks (figure 2) can

have signals travelling in bothdirections by introducing loops inthe network.

Feedback networks are verypowerful and can get extremelycomplicated.

Feedback networks are dynamic;their 'state' is changingcontinuously until they reach anequilibrium point.

They remain at the equilibriumpoint until the input changes and

a new equilibrium needs to befound.

Feedback architectures are alsoreferred to as interactive orrecurrent, although the latterterm is often used to denotefeedback connections in single-

layer organisations


10/35

10

NETWORK LAYERS

The commonest type of artificial neural network consistsof three groups, or layers, of units: a layer of "input"units is connected to a layer of "hidden" units, which is

connected to a layer of "output" units

The activity of the input units represents the rawinformation that is fed into the network.

The activity of each hidden unit is determined by theactivities of the input units and the weights on theconnections between the input and the hidden units.


11/35

11

Mathematical Analysis Of ANN in PSR

The input patterns are represented by a vectorXp=(xp1, xp2, xpN ) simply redistributes them to the following

hidden layer.Each neuron of the following layer receives theweighted signals (signal multiplied by a weight) andgenerates an output signal to the following layer.

This process is repeated until the output layer isreached, where the neurons will generate the output ofthe ANN for the given input vector.

With the output of the ANN obtained, the weightadjustment of the connections will begin in the direction

from output layer to input layer.

The weight adjustments are realized in order tominimize the error functionfor a certain pattern.Equation (1) illustrates the error function where dp is the

desired output for input pattern p and y p is the actualoutput pattern.

Ep

= 1/2Ep

= 1/2


12/35

12

PERCEPTRONS

The most influential work on neuralnets in the 60's went under the

heading of 'Perceptrons' a termcoined by Frank Rosenblatt.

The Perceptrons turns out to be anMLP model(neuron with weighted inputs ) withsome additional, fixed, pre--processing.

Units labelled A1, A2, Aj , Ap arecalled association units and theirtask is to extract specific, localisedfeatured from the input images.

Perceptrons mimic the basic ideabehind the mammalian visual system.

They were mainly used in patternrecognition even though theircapabilities extended a lot more.


13/35

13

CRITERIA FOR RESTORATION

1. RESTORATIONRestoration processes deal with a broad range ofphenomena involving from load flow to electromagneticand electromechanical transients analysis.

Therefore, a large variety of programs and models isneeded and an extensive analysis is demanded to providea set of viable solutions to restore the power after anoutage.

In the industrial region the total amount of interrupted

load is above 10,000 MW. The disturbances, caused bythem can lead to some frequent such as :

Congestion of communication links,

Difficulties in identifying the post disturbance

configuration


14/35

14

2. GEO-ELECTRICAL AREAS

The concept of the geo-electrical areas is not muchdifferent from the power restoration line where there is apredefined partitioning of the network.

Geo-electrical areas are minimum grid configurationscontaining one or more generating units allowing therestoration of priority loads in the fastest and safest way.

They represent the minimum of island configurationwhere the power supply is still attained within the normallimits.


15/35

15

3. RELIABILITY DEGREE

This index is used to define a geo-electrical area, Thisdegree can be classified in the following way:

High Reliability: units capable of black startindependently of any external supply and can start fromstandstill.

Medium Reliability: capable of supplying their auxiliarysystems with terminal voltage in the generating units

Low Reliability: units that require an external supplyfrom auxiliary system


16/35

16

GUIDELINES FOR POWER SYSTEM RESTORATION

The national and regional operating centers coordinatethe load shedding and closing of loops or parallelling ofsystems in distinct geo-electrical areas that wererestored during the fluent restoration.

The geo-electrical areas are first restored in apractically independent manner during the fluent phase.

There are restoration procedures that involve, in thebeginning, only fluent process and later on, start the

coordinated one, while there can be a procedure whereeven the first steps in restoration are taken in acoordinated manner.

The restoration can also be coordinated when there isany sort of impediment that requires the action of a

group of distinct operating centers or a higher


17/35

17

RESTORATION STUDIES

1. Power Availability

The first criteria is the availability of generating unitsequipped with black startThe amount of pickup load cannot exceed the initialactive power capability for each region.

The second criteria is the minimum amount ofgenerating units that can be considered, taking intoaccount maintenance and electrical parameters. If on the one hand, the number of generating unitsmust be such that overload or self-excitation are avoided;on the other hand, the overall restoration time has to be

less than the maximum operating time that any machinecan operate unloaded. Therefore, for (n-1) generating units or the minimumnumber of generating units available, equipped withblackstart, the initial power availability for a geo-electrical area is defined by

Pint=0.8(n-1)Pdisp (1)


18/35

18

2. Voltage Control in Geo-Electrical Areas During

Fluent Restoration

Shunt reactors are used together with the reactivepower of the generating units to provide a reasonablevoltage profile and avoid high overvoltage due to Ferrantieffects in the transmission paths.

The reactive power supply provided by the generatingunits must comply with the machine capability curve.

The amount of priority load to be picked up in any geo-electrical area is from the total available active andreactive power delivery by the generating units.

During all of the stages of the restoration process, thevoltage control is such that the maximum allowed

overvoltage is 110% of nominal voltage and the minimumis 90%.


19/35

19

To analyze the behavior of frequency and voltageoscillations during load pickup and load rejection

Frequency deviations from 55 up to 65 Hz can beaccepted as minimum and maximum limits

This limitation is a question of design, the unitsnowadays operating in INDIA are not capable of large

frequency excursions.

Besides the frequency oscillation, there is another issuerelated to the reactive power capability of thermal units.

Typically, a thermal unit could energize around 150 to

200 km of a 500-kV transmission line.

Restoration Studies


20/35

20

4. Electromagnetic Transients

Define maximum voltage for equipment energization and if it ispossible to energize the equipment without any risk consideringtransient over voltages or system resonance.

They are used to verify short-term conditions such as line,transformer energization, and load rejection.

For the transmission lines, the studies contemplate the lineenergization with or without a fault at the receiving end of thetransmission line.

The load-flow results represent a prior to fault scenario for thetransient case.

A statistical case is done to obtain which is the worst casescenario.

The load rejection studies define the levels of maximum load

pickup during the restoration process as well as the minimumconfi uration of reactors for the s stem under stud .


21/35

21

PRINCIPLES FOR POWER SYSTEM

RESTORATION


22/35

22

SYSTEM

DATA BASESCADA/EMS

IMPLEMENTINGTHE PLANCOMPUTER

AND/OROPERATOR

Communication

Communication

THREE BASIC FUNCTIONS OFRESTORATION

PLANNINGFUTURE STEPS

AND/OR

OPERATOR

Communication


23/35

23

AUTOMATED RESTORATION

In this restoration technique, computer programs areresponsible for the PSR plan development andimplementation.

The PSR techniques based on this principle acquiresystem data from the supervisory control and dataacquisition system (SCADA) and the energy managementsystem (EMS).

Under a wide area disturbance, a PSR program installedin the EMS system will use the acquired system data to

develop a restoration plan for the transmission system.

After developing the restoration plan, a switchingsequence program, which is also a part of the EMS, willbe responsible for the transmission of control signalsthrough SCADA to circuit breakers and switches to

implement the plan. In this technique, the systemoperator plays the role of a supervisor.


24/35

24

SYSTEM

DATA BASESCADA/EMS

OPERATORFOLLOWINGPROCEDURE

S

COMPUTERINTERFACNGALGORITHMS

Operator action

communication

Response

Inquiring

Communication

COMPUTER AIDED RESTORATION


25/35

25

SYSTEM

DATA BASESCADA/EMS

RESTORATIONSCHEDULEPLAN

NED ON LINE

Implemenation

through

communication

Communication

COOPERATIVE RESTORATION

COMPUTER


26/35

26

PROPOSED SCHEMEEach IRS is composed of two ANNsand a switching sequence program(SSP).

Input to 1st ANN will be anormalized vector composed of thepredisturbance load. Responsible forload forecast.

The second ANN of each IRS is

responsible for the determination ofthe final island configuration and theassociated forecast restoration loadpick up percentage that will generatea feasible operational condition.The final element of each IRS is theSSP. The SSP will determine the

energizing sequence of transmissionpaths that will lead to the finalconfiguration chosen by the secondANN.

The energizing sequence databaseof each IRS is composed of

transmission path sequencesconnecting island generators to


27/35

27

Study System

The transmission system chosen for the application ofthe technique is the IEEE 162-bus 17-generator system.

The operating conditions of a transmission system can

vary significantly during a year and even during a week.

Therefore, more than one operating condition wassimulated to train the IRSs. Due to the size of thetransmission system, it was assumed that up to tenislands could be formed following the occurrence of a

wide area disturbance.

Each islandwas predetermined based upon offlinestudies satisfying the requirement of approximate loadgeneration balance within each island.

The all open switching strategy was assumed to be in


28/35

28

Restoration Constraints

In order to generate a feasible restoration plan to beused as a training pattern by the IRSs, certain operationalconstraints must be considered.

Thermal limits of transmission lines, stability limits,number of lines used in the restoration plan, allowableover and under voltages, and recognition of locked-outcircuit breakers are some of the constraints considered.

The thermal rating of normally designed transmission

lines depends mainly on the voltage level at which theyoperate, the line lengths, and reactances. Power systemstability is a subject of major concern in PSR .

The restored system generated by the PSR scheme hasto be able to allow for sufficiently large load and

generation variations without encountering undesirable


29/35

29

Tests


30/35

30

However, sustained power frequency over-voltage iscaused during this process by charging currents ofunloaded or lightly loaded transmission lines . If excessive and not controlled, sustained over-voltage

can overexcite transformers, cause transformeroverheating, more severe switching transients, and evengenerator self excitation and runaway voltage rise thatmay make the system collapse again. Therefore, sustained over-voltage is of great concernduring the early stages of power system restoration

A simple and approximate approach was presented forevaluating the sending- and receiving- sustained andtransient over-voltages in [5]The system is weaker; therefore, adjustments to controlvariables for one target bus should not cause new voltageviolations

Control of overvoltage during the early stage of

power system restoration


31/35

31

HIERARCHAL TIME EXTENDED

PETRI NET SCHEME

In recent years, the Petri Net technique has receivedextensive attention in generic restoration schedule ofpower system.

The graphical and mathematical modeling tool PN is apowerful inference mechanism and has beensuccessfully applied in the areas of service restorationscheduling.

PN simulates the operations of a system with visual

representations. In order to improve the limitation ofclassical PNs, Hierarchical Time-Extended Petri Net (H-EPN) has been proposed .

H-EPN is a reasoning tool for modeling of discreteevent systems. It has a superior performance in terms

of modeling and decision power


32/35

32

CONCLUSION

transmission distance is high, over 400 km, careful analysis ofthe performance concerning not only steady-state and stabilitybut also electromagnetic transient studies.The load shedding was defined as a function of safety of thevoltage and the level of reactive compensation in the LIGHTarea. An alternative restoration process can only be chosen if itpresents a minimum interference with the other processes.

The load pickup capability in power station does not increasewith a higher amount of generation nor does this imply a fasterrestoration procedure for this area.The fastest restoration procedure is the one carried out only bythe Marimbondo power station with five machines withoutinterference from any other geo-electrical area


33/35

33

REFERENCES

www.doc.ic.ac.uk/~nd/surprise_96/journal/vol4/cs11/ report.html[2] Teo, C.Y.; Wei Shen, Development of an interactive rule-based system for bulk power system restoration, PowerSystems, IEEE Transactions on , Vol.15, Issue:2 , May 2000 ,pp.646 653.[3] Bretas, A.S.; Phadke, A.G.; Artificial neural networks inpower system restoration Power Delivery, IEEE Transactionson , Volume: 18 , Issue: 4,pp.1181 - 1186 , Oct. 2003.[4] Authorized licensed use limited to: INDIAN INSTITUTE OFTECHNOLOGY DELHI. Downloaded on September 24, 2009 at06:40 from IEEE Xplore. Restrictions apply.[5] Backtracking Based Algorithm in Hierarchical Time-extended Petri Net Model for Power System Restoration DongLiuYunping Chen, Rongxiang Yuan, Guang Shen, School of

Electric Engineering, Wuhan University,IEEE Transactions OnVol.15,August 2004.


34/35

34

QUESTIONSQUESTIONS


35/35

35

THANK YOU

seminar presentation on ann in psr

Documents