IEEE Transactions on Power Systems, Vol. 13, No. 1, February 1998 40

A Knowledge Based Tutoring System for Teaching Fault Analysis Michael Negnevitsky

Member, IEEE University of Tasmania

Australia

ABSTRACT

A knowledge based tutoring system is used to support the education of power engineering students. The aim of this project is to make teaching and learning more productive and ef€icient by employing modern technologies. It seeks to find new methods to teach large numbers of students with no

The knowledge based tutoring system whose development is reported here operates in an active dialogue mode with the student, using examples, and providing immediate explanations and feedback to students. It also allows the students to access to the learning facility at any time that is convenient for them.

2. COMPUTERAIDEDEDUCATION increase in staff. The tutoring system is based on an expert system shell. It provides a functionally interacting set of theory and problems, and supports student progress through monitoring and assessment. This paper describes the development of the tutoring system for teaching electrical engineering subjects, and in particular, fault analysis in power systems. The expert system based software has been successfully used by power engineering students. They found this software easy to use and understand, and it has become an extra teaching tool.

Keywords: Power Engineering Education, Tutoring System, Expert System Shell, Fault Analysis.

1. INTRODUCTION

Rapid growth of quality and quantity demands in engineering education of technologically competitive societies makes computer-based multi-media education crucial [ 11. During the last few years computer-based education has been improved by the introduction of intelligent tutoring systems. Such systems can be based on Expert System (ES) technology.

An expert system provides built-in support facilities such as debugging aids and knowledge base editors, built-in inpuiloutput and explanation mechanisms. These facilities help to make the tutoring system development much easier and faster.

The proposed teaching tool is intended to supplement, but not replace, traditional teaching techniques such as lectures and laboratory sessions. Variation in educational techniques and materials promotes better understanding of a subject. Simply applying theory is ineffective [2]. Examples can influence learning process much more then the presentation of concepts and even rules [3].

Current tertiary teaching is predominantly conducted through lectures, tutorials, laboratory sessions and workshops. However, lectures are not suitable for individualised teaching. Lecturers usually target average students and cannot give adequate attention to weak and bright ones. Tutorials and laboratory sessions used to be an opportunity for individualised teaching. However, the steadily increasing ratios of student to staff and a reduction in laboratory work make it difficult to provide reasonable feedback to students on their performance and understanding of the subject materials. Meanwhile, obtaining feedback is of major importance for both teaching and learning. In order to make feedback effective, it must be provided promptly and include sufficient explanations.

Computer aided education is able to solve most of these problems [4]. It can introduce a wider range of materials, assess the knowledge level of a student and, if it is not sufficient, present additional information, provide prompt feedback to the student with explanations of his or her mistakes and, if necessary, change the learning format, provide feedback to the lecturer on the material students are having difficulties with and, of course, allow the students access to computer- based learning at any time. Modern computer-based systems can also allow the incorporation of sophisticated multimedia representations of learning materials.

In order to be successful, computer aided tutorials should PI: 0

e

Determine the prior knowledge of a student; Gain and maintain attention and reinforce the motivational state of the student; Present examples with clear explanations of key points; Provide interactive mode of learning with prompt response to any student answer;

*

e

Provide sufficient and clear explanations of a student's mistake during tutorial sessions; Provide assessment based on the results achieved and time spent;

PE-763-PWRS-0-05-1997 A paper recommended and approved by the lEEE power Engineering Education Committee of the lEEE power Engineering Society for publication in the IEEE Transactions on Power Systems. Manuscript submitted December 16, 1996; made available for printing May 23, 1997.

0 Identify gaps in the student's knowledge and, if necessary, present additional materials;

0 Provide assessment information to the lecturer.

0885-8950/98/$10.00 0 1997 IEEE

41

Leonardo Expert System Shell ~- User - Knowledge Base - Inference -

Engine Interface Rules

Objects Development Developer E - - Interface _. -

-

These principles were used to develop the knowledge based tutoring system.

In order to investigate the scope for the application of a computer aided teaching system, analysis of balanced and unbalanced fault conditions in power systems was accepted.

User

Developer

3. CALCULATION OF BALANCED AND UNBALANCED FAULT CONDITIONS

Fault analysis is one of the basic elements of power system analysis and operation and is included in Power System undergraduate courses in electrical engineering. However, a lecturer normally has very few hours in which to present fault calculation procedures. For example, at the University of Tasmania, a lecturer can allocate to balanced fault analysis only one lecture and one tutorial in Power System 1, and two lectures and two tutorials in Power System 2 to represent unbalaxed fault conditions. Clearly, in such a short introduction some challenging problems cannot be considered.

In general, the analysis of any fault condition is performed in the following order [6]:

1. Represent the given power system by its positive, negative and zero-sequence networks (the zero-sequence network is omitted for faults without earth, and both the negative and zero-sequence networks are omitted for the balanced three- phase fault condition). This representation requires the calculation of per unit (P.u.) impedances for generators, transformers, lines, cables and other elements of the power system.

2. Reduce each of the sequence networks to its simplest form. The equivalent positive, negative and zero-sequence networks are represented as a series and series-parallel combinations of the p.u. impedances. These are replaced by the single equivalent impedance for each sequence network. It may also involive the use of the delta-star or star-delta transformations.

3. Use the appropriate symmetrical-component equations to find the phase-sequence components of the current in fault under the particular short-circuit condition.

4. Determine the required p.u. phase-current values at the point of fault.

5. Finally, calculate the actual values of the phase-currents by multiplying obtained p.u. values by the base current at the point of fault.

The procedure outlined above provides a complete analysis of the given power system for the specified fault condition and can be easily implemented in computer aided tutorials.

4. LEONARD0 EXPERT SYSTEM SHELL

The Leonardo expert system shell has been chosen from among several packages for development of the prototype

expert system applications. The Leonardo Development System provides facilities for building and testing expert

application. Leonardo is an object oricntcd tool for developing

Fig. 1. Basic components of the Leonardo expert system shell.

systems, the Leonardo Run System allows the delivery of complete applications to the user and the Leonardo Productivity Toolkit provides efficient aids for tailoring the application. Fig. 1 shows the basic parts of the Leonardo expert system shell.

The knowledge base comprises the rules and objects which Leonardo uses for representation of the expertise. The know- ledge base can store objects with frames which hold additional information, for example, procedures and screen design.

The inference engine provides a production rule interpreter, a method for finding values of objects, and How? and Why? explanation facilities.

The development engine provides the facilities necessary to create and edit rules for the knowledge base and to edit information into object frames.

There are three types of rules: normal, assertive and quantified. A normal rule tests the values of some objects in its antecedent and sets the values of some objects in the consequent. An assertive rule is a simple command which defines the goal object (seek), collects initial data (ask) or prints information (say). A quantified rule can generate an exhaustive search over all members of the class objects. Class objects may be created with several member objects inheriting certain slots. Multiple inheritance is also permitted for class objects.

Leonardo automaticallly generates screens for the user input and output. However, a screen design package is also available to create tailored input and output screens. The graphics package provides built-in procedures to write a wide range of graphic images to the screen.

Leonardo has its own procedural programming language which is used to perform complex computations, access external databases and programs developed in C, FORTRAN and Pascal, print reports, manipulate the screen and so on.

All these features make Leonardo a very inviting shell for the development of expert systems, especially in the idea and prototype stages.

5. DESIGN OF THE TUTORING SYSTEM

Power systems are subject to the following principal types of faults:

phase-to-earth (single-phase); double phase-to-earth1 (phase-phase-earth).

three-phase with and without earth connection; - phase-to-phase (two-phase);

42

LEONARD0 (c)1986-1990 Creatlve Logc Ltd Knowledge Base FAULT

INTELLIGENT TUTORING SYSTEM AEA 441 Power Systems 2

Introduction Fault Analysis in Power Systems

POWER SYSTEMS are subject to many mfferent lands of faults A fault can be defined as any abnormal condmon which causes a reductlon in the basic insulahon strength between phase conductors and earth The pnncipal types of faults are as follows

1 2 Phase-to-phase faults 3 Double phase-to-earth faults 4 Three-phase with and without earth connectlon

Short circuit of a single conductor to earth

[NEXT screen will take you to the illustrative example which gives a step by step guide on how to deal with fault analysis problems]

Type any key to see the next screen

Fig. 2. Introduction screen.

The tutorial session starts with the introduction screen which provides a start-up message. This message is placed in the Introduction slot of the object targeted by the seek command in the main rule set. When execution begins the corresponding text will be found and displayed on the screen as shown in Fig. 2.

Then the illustrative example may be displayed. It provides step-by-step calculations of the symmetrical and asymmetrical currents in the power system.

Typical screens presented to the student are shown in Fig.3. The solution is easy to follow and understand.

The student is then asked whether or not he or she wants to run this example again. If the example is understood, the user is taken to the standard query screen.

The intelligent system provides three different sets of tutorial problems, and each set consists of a number of problems distinguished by the level of difficulty. The first tutorial set can be used to give some introduction to the field. It contains simple ordinary problems which do not have a difficulty level higher than 8. The second set provides more difficult problems and the last set offers the most challenging problems with a difficulty level of 10. The student is asked to select the difficulty level as shown in Fig. 4. The difficulty levels are then used to calculate the final mark obtained by the student.

The actual text which appears on the query screen shown in Fig. 4 is determined by the corresponding slots of the object in question. The query preface area on the top half of the screen is used to give the user information about the query. If the QueryPrefuce slot has no text in it, the query preface area of the screen will be blank. The QueryPrompt slot contains a text which is displayed on the middle line of the screen. If the QueryPrompt slot is empty, a default prompt will be generated. The input area under the prompt line includes the allowed input values for the object and the explanatory text for each individual value. This area is controlled be the Allowedvalue and AVExpansion slots. When the query screen is displayed, the first option on the input menu is highlighted.

FAULT ANALYSIS: EXAMPLE The power system shown below develops a fault on the 132 kV transmission line Determine the short-circuit currents at the point of fault, assuming the fault condition to be (a) a three-phase fault (b) a phase-to-phase fault (c) a single-phase-to-earth fault

/

Z,* = ~ 9 . 2 %

Type any key to see the next screen

STEP 1: Choose S buse * The power base can be adopted equal to the power rahng of

any power system generator or transformer, or may be chosen as a whole number as 10,100 or 1000 MVA.

S base is chosen as 100 MVA.

STEP 2: Choose V buse * The choice of the voltage base IS deteniiiiied by the voltagt:

rating of the busbar where the fault IS located

V base is chosen as 132 kV.

STEP 3: Culculute I buse

S buse 100 & V buse - & 132

Ibuse = - - - = 0.437kA

Type any key to see the next screen

STEP 6: Reduce each of the sequence networks to its simplest form * The system positive, negative and zero sequencc nctworks are.

Fl

F2

P"

zo=]473 +J0092+]0 125+]0 1 =J5 047pU

Type any key to see the next screen

Fig. 3. The step-by-step solution screens.

When a student has made a selection of the difficulty level, he or she is given a corresponding problem to solve. A screen showing one of the problems is demonstrated in Fig. 5. The values shown in the rectangles are obtained using a random number generator to provide different magnitudes every time for all variables.

If the answer is correct (for example, Z, = 0 as shown in Fig. 5), an encouraging massage is displayed (eg. "That is

43

FAULT ANALYSIS: Level 2 Problem I

Calculate the fault currents in the system shown below if a single-phase-to-earth (a-e) fault occurs at point F

LEONARD0 (c)1986-1990 Creative Logic Ltd. Knowledge Base: FAULT

The intelligent system provides you an apportunity to select one of the three sets of tutorial problems. Each set consists of a number of problems identified by the level of difficulty.

Please make v i r selection

mi I Level 1 I Number of problems: Difficulty level: Time recommended: 20 minutes Maximum mark: 80

FKeys: 1 Help 2 Quit 3 Why? 5 Volunteer 6 Backup 7 Expand 8 Review

Fig. 4. Format of the standard query screen. ~~

FAULT ANALYSIS: Level 3 Problem 2

Calculate the fault currents in the system shown below if a two-phase-to-earth (b-c-e) fault occurs at point F

Z,, =a+ ] 0 Ohms Z , = 0 + I a O h m s F

- - f a - . .

Please enter the magnitude for fault current I, in Amps: > O That is correct! Now enter the magnitude for fault current Ib in Amps: > 2 Your answer is accepted. However it is inaccurate! Please be careful next time Please enter the angle for fault current Ib in degrees: >

Fig. 5. Standard screen with the problem.

correct", "Your answer is accepted") and the student is asked to input the magnitude and angle for current Ib and then I,.

If the student has made a mistake, a message to correct or revise the answer occurs. However, if the student makes the same mistake a second time, an explanation is provided and the intelligent system changes its teaching mode. It poses basic questions to the student and allows him or her to find the solution in a step-by-step manner. This approach helps the student to identify mistakes easily and learn how to obtain the right answers. Typical dialogue screens are shown in Fig. 6.

5. ASSESSMENT PROCEDURES

Student assessment is a key element of the developed knowledge based tutoring system. It allows to monitor progress of every atudent in the class and also present information about overall group performance.

Please enter the magnitude for fault current I , in Amps: > 0.71 The fault current is not correct! Please try again: > 0.85 This value is still wrong! This is how the problem is solved. You need first to select the power base, S base = V base = 11321 kV. I base is then calculated. Calculate I base and enter it in Amps: > 0.44 That is correct!

MVA, and the voltage base,

H[it any key to continue

F a

Length=m!m yo ! q ! S , = m M V A V g = a k V S , = m M V A Zi=jllLmOhms/!un Z 1 = . i m m / m I k V ZO=j10.380hmsikm Z,=jlzz.op/, z , = j m S Z , = j l >

To calculate the positive, negative and zero-sequence p.u, reactances for the generator, you need its MVA, Z1,ZZ and 2.0 and S base. Enter the positive-sequence p.u. reactance for the generator: > 26 The p.u. reactance is not correct! This is the formula you should apply:

Zl% S base z =-- El 100 s ,

Calculate the positive-sequence PSI. reactance for the generator and enter the value: > 0.26 That is correct! Enter the negative-sequence p.u. reactance for the generator: > 0.22 You are right! Enter the negative-sequence p.u. reactance for the generator: > 0.086 That is correct!

Hit any key to continue

/ F FQ IqI a

To calculate the p.u. reactances for the transformer, you need its MVA , Z and S base. Enter the positive-sequence p.u. reactance for the transformer: > 0.14 That is correct! Enter the negative-sequence p.u. reactance for the transformer: >0.14 Yes! Now enter the zero-seqquence p.u. reactance for the transformer: >0.14 Excellent!!! The positive, negative and zero-sequence reactances of the transformer have the same values.

Hit any key to continue

Fig. 6. Screen {output of an interactive session.

44

On average, lecturers need to allocate from 10% to 15% of their time to student assessment and examination [7 ] . However, these activities can be now easily automated. Using the knowledge based tutoring system, lecturers are relieved of the tedious tasks of mark allocation and recording, and calculating individual student and group statistics. With the developed system, a student can test his or her own knowledge any time with minimum help from the lecturer.

All tutorial problems concerned with fault analysis in power systems are divided into three groups at different levels of difficulty. Every problem is in turn subdivided into smaller elements and each of the elements is given an appropriate score depending on the difficulty level of the problem.

Students should perform their calculations to an accuracy of 10%. Additional 5% inaccuracy is also allowed but 0.5 of a mark is subtracted from the final score. If the student makes any mistake and his or her input is incorrect, the explanation and guidance will be provided automatically but marks are subtracted accordingly. When the student completes a problem an assessment of his or her mistakes is displayed.

If the student has solved all problems in the tutorial set proposed by the system, a detailed assessment will be provided. An example of the assessment can be seen in Fig. 7. The total score obtained by the student is divided by the maximum possible marks and multiplied by the difficulty level. Then the student is asked to insert in drive A an assessment diskette which must be obtained from the lecturer.

A group's performance can be reviewed by the lecturer using the assessment diskette. The lecturer may analyse the student's work by applying any available option shown on the screen in Fig. 8.

The chronological order option indicates the score obtained by the student, time taken to complete the task, date and the difficulty level. The alphabetical order option provides the same information but lists the students in alphabetical order, as shown in Fig. 9.

FAULT ANALYSIS: Level 3 Numbcr of problems solved : 4

Mistakes committed Number Marks Subtracted 1. Wrong I base 1 2 2. Inaccurate Ibase 2 1 3. Wrong generator reactance 1 2 4. Inaccurate generator reactance 2 0.5 5. Wrong lransforiner reactance 1 2 6. Inaccurate transformer reactance 1 0.5 7. Wrong transmission line reactanse 1 2 8. Inaccurate transmission line reactanse I 0.5 9. Wrong equivalent reactance 1 2 10. Inaccurate equivalent reactance 1 0.5 I 1. wionp rauli current 1 4 12. Inaccurate fault current 2 0.5 13. Wrong fault MVA 1 2 14. Inaccurate fault MVA 1 0.5 15. Assistance required 4

Otley, Martin Kenneth, your mark is 46.0 / 7 0 x 100 = 65.7% Saving is complete. Return the assessment diskette to your lecturer.

Hit any key to continue

Fig. 7. Assessment of the student performance.

e students You

Please make your selectlon

mark obtamed by the student, tlme taken to

FKeys 1 Help 2 Quit 3 Why? 5 Volunteer 6 Backup 7 Expand 8 Revieu

Fig. 8. The main menu for group performance analysis.

FAULT ANALYSIS Group Performance in the Alphubeticul Order

Name

Anas, Mohd Elmi Foo, Cheng Zui Lones, Gareth Mark Koh, Kok We1 Kumar, Manoj Lim, Chang Thiaw Loh, Hak Leong Ngo, Shing Haw Otley, Martin Kenneth Sia, Yn Fu Henry Tal, Lip Fatt Tan, Kwang Meng Tan, Lye Teik Roland Veimey, Darman Michael

ID No.

913726 936384 913005 938782 926055 925932 939445 938604 920717 926041 939173 938766 926010 911968

Mark Time

52 48 78 37 55 51 75 36 73 39 72 42 78 41 62 49 14 34 44 52 58 45 62 47 57 53 58 38

Hit any hzy to continue

Date Level

11-Sep-95 3 11-Sep-95 3 11-Sep-95 3 14-Sep-95 3 11-Sep-95 3 14-Sep-95 3 11-Sep-95 3 11-Sep-95 3 11-Sep-95 3 14-Sep-95 3 11-Sep-95 3 11-Sep-95 3 14-Sep-95 3 11-Sep-95 3

Fig. 9. Group performance analysis in the alphabetical order.

The descending mark order option allows assessment returns to be viewed so that the best students are at the top of the list. The detailed performance option offers analysis and gives detailed information about individual mistakes made by any student in the group. Finally, the group performance option provides the total number of different mistakes made by the students during the test, as shown in Fig. 10. This option helps a lecturer to identify the most common mistakes and bottlenecks, and to decide whether it is necessary to revise some of the tutorial problems.

Finally, a distribution of grades obtained by the students can be also displayed as shown in Fig. 1 1,

6. DISCUSSION

The Leonard0 based tutoring system is a valuable tool for teaching fault analysis in power systems. In October 1994, the system was installed in a computer network and it can now be accessed from any computer in the Department of Electrical and Electronic Engineering. It has been found that network delivery of computer-based tutorials is the most cost effective

45

FAULT ANALYSIS: Level 1

Group Performance: Summary of Mistakes Mistakes Committed 1. Wrong I base 2. Inaccurate Iha9e 3. Wrong generator reactance 4. Inaccurate generator reactance 5 . Wrong transformer reactance 6 . Inaccurate transformer reactance 7. Wrong transmission line reactanse 8. Inaccurate transmission line reactanse 9. Wrong equivalent reactance 10. Inaccurate equivalent reactance 11. Wrong fault current 12. Inaccurate fault current 13. Wrong fault MVA 14. Inaccurate fault MVA

Hit any key to continue

Number 2 5 1 0 2 3 6 2 4 0 5 2 0 1

Fig. 10. Summary of mistakes made by the group of students.

FAULT ANALYSIS: Level 3

Group Performance: Distribution of Grades

The number of students

5

NS PP CR DN HD Grade

Hit any key to continue

Fig. 11. Distribution of grades obtained by the group.

Updating the tutorial problems is a relatively simple procedure. When the material is delivered via the computer network it can be updated only on the network server.

The comments provided by the students during the testing period were mostly very positive. Examples of comments are listed below:

Tutorial sessions benefit by including the knowledge based tutoring system. Examples and explanations provided by the system help me to understand the material. I like using the knowledge based tutoring system. It is always available. It is more interesting than traditional tutorial sessions. The program runs very smoothly and it is well written. This is a good study for the exam. I found the problems were easier as I went. The first few I got wrong because I did not understand the basics. After I understood the basic methods the last ones wcrc easy, m e n problems in Level 2. There should have been a few harder problems.

Based on the student comments and recommendations, a few new more challenging problems have been added into Level 3.

It should also be noted that materials used in computer aided teaching must be prepared very carefully, keeping in mind that the lecturer cannot be accessed during every tutorial session to explain or clarify difficult points.

7. CONCLUSION

A knowledge based tutoring system for teaching fault analysis in power systems has been developed and successfully used by tlne power engineering students.

0 The students have found the tutoring system easy to use and understand and have received a good introduction to fault analysis in power systems. The system provides an automatic student assessment and helps lecturers to identify the most common mistakes and bottlenecks in the tutorial problems.

8. REFERENCES [ 11 P. Darvall, "Towards world-best practice in engineering

education", Australian Journal of Public Administration,

[2] G. Gibbs and T. Habeshaw, Preparing to Teach - An

Vol. 52, NO. 1, pp. 53-64, March 1993.

Introduction to Effective Teaching in Higher Education, Technical and Educational Services Ltd, 1989.

A. Lesgold, J. Pellegrino, S. Fokkema and R. Glasser, Cognitive Psychology and Instruction, Plenum Press, New York, 1978.

D.Luketina, "Towards effective computer based education", Proceedings of the IEEE First International Conference on Multimedia Engineering Education, Melbourne, July, 1994, pp. 36-41.

M. Negnevitsky, "An Expert System Based Teaching Tool for Electrical Engineering Subjects", Proceedings of the Pacific Region Conference on Electrical Engineering Education, Marysville, Victoria, Australia, February 23-24,

J.J Grainger and W.D. Stevenson, Power System Analysis, McGraw-Hill, Inc. New York, 1994.

1995, pp. 170-173.

[7] K.E.Green, Educational Testing: Issues and Applications, Garland Publishing, Inc. 1991, pp. 249-260.

Michael Nemevitsky received his BSEE and Ph.D degrees from the Byelorussian University <of Technology in 1978 and 1983 respectively. From 1984 to 1991 he worked there as a Senior Research Fellow and Senior Lecturer in the Department of Electrical Engineering. After his arrival in Australia, Dr M. Negnevitsky worked at Monash University, Melbourne. Currently, he is a Senior Lecturer in the Department of Electrical and Electronic Engineering at the University of Tasmania, Australia. His major interests are power system analysis, and cnpert system applications in power systems. He has published over 90 papers. Dr M. Negnevitsky is a Senior Member of IEAust and a Member of IEEE.