Abstract:

The frequent failures and repairs of electrical transformers in a typical large electrical network leaves behind lots of data that can prove to be very useful. In this paper, we investigate data structures used to trace backward old information that is normally missing from the latest version of the system database. The database involved is spatio-temporal in nature; it involves spatio-temporal attributes related to the various transformers on the network. The problem we are interested in is to be able to reconstruct the status of the network at any given time in the past.
 

Introduction

People often associate a number of definitions with a Geographic Information System (GIS). There are those who simply describe a GIS as a spatial database system.  Others like to think of their GIS as an intelligent spatial database system. Still, the definition that we think is most engaging is the one that thinks of a GIS as a modeling tool capable of bringing the real world into our computer. Today, and after about three decades of GIS-related research, existing GISs are very successful at modeling the static part of our world; that is, modeling real-world phenomena that are dependable from space or time, but not both. After all, a GIS was originally designed to process, manage and analyze spatial data and the great success of GISs today has proven the original design to be a big success.

However, GISs' ultimate challenge remains unanswered. That is, can a GIS model real-world dynamic phenomena as well? Obviously dynamic phenomena are more difficult to model because they introduce the new element of change through time into the GIS.  To deal with dynamic phenomena, one has to understand the underlying components of change and how people reason about change in general.   The attempts to model dynamic phenomena have resulted in a new field of research named spatio-temporal database systems which are basically the outcome of the aggregation of time and space into a single framework. A spatio-temporal system should be able to process, manage and analyze data that is dependable from space and time.

The ability to incorporate temporal change into our data model is fundamental to our ability to respond to the many challenges set forth by the GIS community. Following Peuquet [9], the greatest promise of spatio-temporal systems resides ultimately in their capacity to examine casual relationships and their effects in any of four modes of inquiry: planning, exploration, explanation, and prediction.  Traditional GISs; i.e, spatial systems, are already great at providing means to facilitate planning.  The answers to the other three objectives (exploration, explanation, prediction) lie in our ability to represent dynamic phenomena. Only then, we might be able to better understand change in order to improve our ability to explain past events and make accurate predictions in relation to a given phenomenon.

Over the last decade, a number of spatio-temporal data models have been proposed [1][2][3][7][8][11][12][13][15][16][17][19]. After all, these models express the different views humans hold about change over space and time. In this paper, we make use of the three-domain model [20][21] along with the concepts introduced in [4][5] on identity-based change to describe a simple system capable of capturing a rich class of change semantics as normally perceived by humans.  The rest of the paper is organized as follows: Section 2 gives a brief review of the three-domain model. Section 3 reviews the work of Hornsby and Egenhofer on identity-based change.  Section 4 builds upon the concepts introduced by Hornsby and Egenhofer, considers further this approach to change and shows how to implement a system, within the framework of the three-domain model,  that supports identity-based change operations efficiently. Section 5 describes an example of the type of scenarios that our system can support and section 6 concludes the paper.
 

The Three-Domain Model

In order to incorporate temporal changes into a GIS, Yuan [20] proposed a data model, named the three-domain model, which uses a remarkably flexible structure. Basically, Yuan's model distributes objects across three-domains, stored in three separate tables: the space domain holds spatial objects, the time domain holds temporal objects, and the semantic domain holds semantic objects.  In addition, these three tables are fully interconnected via a fourth table, named the relation table where each record is a triplet of the form (semantic-id, space-id, time-id) providing dynamic links among semantic, temporal, and spatial objects.

Figure 1

The semantic-id is used to uniquely identify an object independent of its attributes and values. This concept provides the model with the level of abstraction needed to model a wide range of real world phenomena. The three-domain model is capable of responding to inquiries involving movement as well as change. In particular, Yuan shows that her model can support all four types of spatio-temporal queries that are identified in [8]: simple temporal query, temporal range query, simple spatio-temporal query, spatio-temporal range query.

We have used the three-domain model to build a spatio-temporal database system capable of tracking down the movement of objects through space.  Without loss of generality, we assume here that we are dealing with objects of type Point in a two-dimensional space.  Suppose we begin our experiment by recording the locations of three objects at time t1.  We refer to these objects as A, B, and C.  In our database, the following data would be recorded.
 

Figure 2

Figure 3

The space table normally contains additional attributes besides the shape (we omit them here to keep things simple). The fields named Previous and Next shown above are used to store relationships of parenthood among spatial objects -- this is similar to Yuan's spatial tree. Also, we use these two fields to track the movement of objects through space. Initially these two fields are unknown and as such their space-id values are set to -1.

Figure 4

Figure 5

Suppose now that we detected the movement of A at time t2 and then at time t3. Suppose also that C moved to a new location at time t4. These changes would be reflected in the database as follows.

Figure 6

In the above table, the most current configuration of the system correponds to those records whose Next field is set to -1.

Figure 7

Figure 8
 
 

 Case Study:

The original application, named GISEL (GIS for Electricity of Lebanon) ,  was originally developed by the GIS group at K&A. GISEL consists of several applications (Facility Siting,Trouble Call, Maintenance, Collection Management, Work Order, ArcFM, Switching,  etc.) for the area of municipal Beirut. Mainly, a GIS system was developed to model the existing network of the EDL (Electricity du Liban). This includes parcels, buidlings,substations, MT/BT (Medium Tension to Low Tension transformer rooms), transformers, medium tension lines, low tension distribution lines,  etc.

Figure 9 below shows a capture of municipal Beirut with its 60 sectors.

Figure 9

Another figure shows the above yellow sector with its  parcels.

Figure 10

In this paper, we consider further this application and develop a system based on the three-domain model to keep track of the locations of the different transformers that are part of the EDL network. This allows us to answer a number of spatio-temporal queries which are of great importance to some of the EDL applications such as the maintenance application, the Trouble Call, The Collection Management, and The Switching application.

System Setup: Our application is initially written for Arcview 3.1 software, and then customized using the avenue code.  The data used is that of the entire area of municipal Beirut, but only one sector is used in the examples for this paper.  This sector includes 30 MT/BT (Medium Tension to Low Tension Transformer Rooms), and 36 transformers located in these MT/BTs.

Figure 11

Some of the possible scenarios that are possible to occur on our network include transactions that can be done on transformers from simple switching operations to maintenance to change location to removal.  All these operations are adopted by our model, from changing related attributes of the transformers to changing their spatial locations.  The transactions are reflected in the model studied.  The model will be able to log and retreive these transactions.  It will also be able to answer many other queries about our database for the period when these transactions occurred.  We will see sample data behind these operations and examples of queries that can be resolved using this model.

Assume the following sequence of events.
1999: Transformer # 18 has been moved to MT/BT  # 2697
A New transformer has replaced the old one.

2000: Transformer # 24 has been removed from MT/BT # 2590

2001: Transformer (previously # 18) has been moved to MT/BT 2698.
A new upgraded transformer has been installed
 

The following figures show captures of the views and the underlying tables before and after the changes (Note that in this implementation we have chosen to store the current configuration in a separate table.)

Figure 12 shows the state of the concerned transformers at time 1998

Figure 12

The corresponding tables are shown in the figures below.

Figure 13 (1998 - the Space Table)

Figure 14 (1998 - The Semantic table)

Figure 15 (1998 - The Time Table)

Figure 16 (1998 - Relation Table)

Figure 17: (1998 - The History table)

Next we will see the 2001 data status

Figure 18: (2001 View)

Figure 19: (2001 - The Space Table)

Figure 20: (2001 - The Semantic Table)

Figure 21: (2001 - The Time Table)

Figure 22: (2001 - The Relation Table)

Figure 23: (2001 - The History Table)
 
 

Sample Queries:

Queries about attributes of entities:
These queries seek information about attributes of features and entities.  Example: What is the status of transformer X?  What was the rating of Transformer Y on 21/5/1999?

Queries about locations, spatial properties, and spatial relationships
They answer questions about where and what.  Where was Transformer X at 1995? Which Transformers were in MT/BT Y at time January 1996?

Queries about time, temporal properties, and temporal relationships
What happened to Transformer X from 1996 to 1997?  In Which Month did we have most number of Switching operations?

Queries about spatiotemporal behaviors and relationships
What happened to the transformers of MTBT Y from Jan. to April 1999?
In which Month and In Which Mt/Bt did we have the largest number of repairs?
 

To answer:
What is the status of transformer X?
Select Transformer ID X from Current Space Table,
Query Value for selected record for status (In Service/Out of Service)

Where Was Transformer X in 1995?
From relation Table select space id = X.
Get Time ID
From Time Table check if time <= 1995 then go to Space table and get its location (Shape).
Otherwise, go to History table and get its previous id value.  If previous-id = -1 then the transformer X did not exist in 1995.
Go to the beginning and check again

What happened to transformer X from 1996 to 1997?
Select Transformer X from Relation Table. Get its Time ID.
From Time table, if time < 1996,  then transformer X attributes have not changed values during the given interval. In this case, go to the space table and get its location (shape); stop.
(a) If 1996 <= time <= 1997,  then
    select Transformer from History table
    Get its previous ID.
    Get its (previous ID) time from the Relation table.
    If time <=1996 then
        select Transformer (corresponding to previous id) from History table
        stop
    otherwise goto (a)
if 1997 < time then
    Get its previous ID.
    Get its (previous ID) time from the Relation table.
    If time <=1997 then goto (a)
 
 
 

Conclusion

This paper has described a system based on the three-domain model to provide a system-level implementation for dynamic changes. The added flexibility provided  by the three-domain model where the space, time, and semantic domains are dealt with separately in three different tables and then linked together via a fourth table gives this model a tremendous maneuver capability in handling real-world scenarios. Mainly, the idea of a semantic identifier used to identity an object independent of its attributes and values gives the three-domain model the level of abstraction needed to model changes as perceived by many people. We believe that the three-domain model is worthy of further investigations. We are currently looking at a number of different scenarios from different disciplines to try to model them within the framework of the three-domain model. Our initial results are very encouraging.
 
 

References

[1] F. Bonfatti and P.D. Monari, "Spatio-Temporal Modeling of Complex Geographical Structures", Computer Support for Environmental Impact Assessment: Proceedings of the IFIP TC5/WG5.11 Working Conference on Computer Support for Environment Assessment, CSEIA 93, Como, Italy, Elsevier, 1994

[2] C. Claramunt and M. Thuriault, "Managing Time in GIS: An Event-Oriented Approach," Recent Advances on Temporal Databases, Clifford J. and Tuzhilin A. (eds.), Springer-Verlag, pp. 23-42, 1995

[3] M. Erwig, R. H. Guetting, M. Schneider, and M. Vazirgiannis, "Spatio-Temporal Data Types: An Approach to Modeling and Querying Moving Objects in Databases," Chorochronos, Technical Report, CH-97-08, 1997

[4] K. Hornsby and M. J. Egenhofer, "Qualitative Representation of Change," in: S. Hirtle and A. Frank (eds.), Spatial Information Theory: A theoretical Basis for GIS, Proceedings of the International Conference COSIT'97, Lecture Notes in Computer Science 1329, Springer-Verlag, pp. 15-33, 1997

[5] K. Hornsby and M. J. Egenhofer, "Identity-Based Change Operations for Composite Objects," Eight International Symposium on Spatial and Data Handling, Vancouver, Canada, T. Poiker and N. Chrisman (eds.), pp. 202-213, July 1998

[6] S. Khoshafian and G. Copeland, "Object Identity," SIGPLAN Notices 21: 406-416, 1986

[7] G. Langran, "A Framework For Temporal Geographic Information Systems," Cartographica, 25 (3), 1988

[8] G. Langran, "Time in Geographical Information Systems," (ed.) Taylor and Francis, London, 1992

[9] D. Peuquet, "It's About Time: A Conceptual Framework for the Representation of Temporal Dynamics in GIS", Annals of the American Association of Geographers, 84 (3):441-461

[10] A. Pavlopoulos and B. Theodoulidis, "Review Of Spatio-Temporal Data Models," Timelab Technical Report, Tr-98-3, October, 1998

[11] A. Renolen, "History Graphs: Conceptual Modeling of Spatio-Temporal Data," in: GIS Frontiers in Business and Science, Vol. 2, International Cartographic Association, Brno, Czech Republic, 1996

[12] A. Renolen, "Conceptual Modeling and Spatio-Temporal Information Systems: How to Model the Real World," in: Hans Hauska, Editor, SCANGIS'97, Proceeding of the 6th Scandinavian Research Conference on Geographical Information Systems, Stockholm, Sweden, June 1997

[13] A. Renolen, "The Spatio-Temporal Object Model: A Conceptual Model Based on Ontology," Article submitted to Geoinformatica, 1997

[14] B. Smith, On the Origin of Objects, Cambridge, MA, The MIT Press, 1996

[15] K. Al-Taha and R. Barrera, "Identities Through Time," in: M. Ehlers and D.R.Steiner (eds.), International Workshop on Requirements for Integrated Geographic Information Systems, New Orleans, LA, pp. 1-12, 1994

[16] N. Tryfona, D. Pfoser And Th. Hadzilacos, "Modeling Behavior of Geographic Objects: An Experience With the Object Modeling Technique," The 9th International Conference on Advanced Information Systems and Engineering (CAISE), Lecture Notes in Computer Science, Barcelona, Catalonia, June 1997

[17] M.F. Worboys, H.M.Hearhshow, D.J.Maguire, "Object-Oriented Modeling for Spatial Databases," Int. Journal of GIS, Vol. 4, No. 4, 1990

[18] M. Worboys, "Object-Oriented Approaches to Geo-Referenced Information," International Journal of Geographical Information Systems 8 (4):385-399, 1994

[19] M. Yuan, "Wildfire Conceptual Modeling for Building GIS Space-Time Models, " Proceedings: GIS/LIS'94, pp. 860-869, 1994

[20] M. Yuan, "Modeling Semantic, Temporal, and Spatial Information in Geographic Information Systems," in: M. Craglia And H. Couclelis (eds.) Geographic Information Research: Bringing The Atlantic, pp. 334-347. London, Taylor and Francis Inc., 1996

[21] M Yuan, "Use of Three-Domain Representation to Enhance GIS Support for Complex Spatiotemporal Queries, Transactions in GIS 3 (2): 137-160