The Automatic Clustering of Uniformly Distributed Loads for the use in Rural Electrification Planning KM Govender, AS Meyer, B Dwolatzky Information Engineering Research Program School Of Electrical And Information Engineering University Of The Witwatersrand, Johannesburg, PRIVATE BAG 3 WITS 2050 EMAIL: [email protected]

Abstract - Design of new electricity supply systems for communities is an everyday task for engineers in developing countries. South African government policy prioritizes the electrification of the approximately 50% of non-electrified rural communities. Thus an “Electrification Planning Decision Support Tool” was developed by Banks et al. This GIS based tool aids in technology selection (solar, grid, etc) and produces the electrification plans and costs involved. The tool caters for the supply of electricity to any consumer within given load polygons. This paper describes the automated creation of load polygons in randomly distributed rural communities, which is am important stage of the planning process. Intergraph’s GeoMedia™ software is used as the GIS platform and a benefit points system is used to display the consumer distribution density by means of a thematic map. A user interface tool uses inputs like maximum allowable expenditure and cost per connection for the creation of load polygons. Index terms: load polygons, distance method, grid technique, geographical information system

I. INTRODUCTION There are many remote rural communities in South Africa that have no access to a supply of electricity. This is largely due to the fact the planners concentrated more on urban electrification that was considered high priority regions. Thus there is a need and demand for the design of new electrical supply systems for these rural regions by electricity utilities. The manual approach to planning usually leads to an over designed, costly and nonstandardised solution. This results in delays in producing plans and therefore automated engineering tools will aid the planning process with respect to time of delivery and quality of the plan.

The author acknowledges support from Eskom’s Tertiary Education Support Programme and the department of Trade and Industry’s THRIP programme for carrying out this research.

0-7803-7285-9/01/$17.00 (C) 2001 IEEE

The planning process involves identifying the region to be electrified and funds to be allocated to that region. This is called the “rolling out” process and usually takes three (3) years). The next stage of electrification planning involves the clustering of individual consumers together to form a ‘load polygon’. This is a critical stage of the planning process, in that all consumer-related planning information is represented as an aggregate associated with a specific load polygon. In many cases load polygons can be identified ‘by inspection’ since consumers build their homes close together to form natural clusters. This allows planners to use the natural layout to produce plans and determine its validity. However in certain rural areas, settlement patterns are almost uniformly distributed (or scattered) over a wide area and load polygons cannot be easily created or determined. This paper addresses this situation in which there is no obvious clustering of the consumer layouts. Specific attention is given to the Kwa-Zulu/Natal province of South Africa, since most of the remote rural areas are uniformly distributed and have no supply of electricity. However the solution proposed in this paper is flexible and can be used on any type of consumer layout. A. Background “Government has pledged that in ten years time (the year 2010) there will be no community in SA without access to electricity” [1]. This statement was made at a workshop dealing with the development of an electrification plan to help develop South Africa’s rural areas. Currently about 3.2 million rural households have no access to electricity. The Government’s development strategy will see millions of dollars spent on rural power connections in an attempt to uplift the quality of life for its 20-million rural people. Eskom, South Africa’s national power utility, announced recently that it would electrify 600000 new households, with the focus on rural areas, over the next three (2000 - 2003) years at a cost of about $0.2bn [1]. The year 2000 saw Eskom making an estimated 250000 new connections, with 71% new connections made in rural areas.

450 0-7803-7287-5/01/$17.00 (C) 2001 IEEE

B. Previous Work An “Electrification Planning Decision Tool” has been developed to facilitate the electrification planning process [2]. This tool uses Geographical Information System (GIS) data sets and a score sheet to quantify the “assumed benefit” of electrification for target regions. The tool starts its design process on given load polygons. This is done so that certain “average properties” can be defined for the group, namely an average income, an average demand for electricity supply (viz. ADMD), an average weighting factor for the prioritization of projects (based on urgency for the need of electricity as compared to other regions), etc. The tool aids in technology selection (grid or off-grid such as solar) and estimates the costs involved for the electrification plans. It facilitates long-range strategic planning for entire regions and hence may assist in detailed engineering planning. A disadvantage of the tool is that it uses predefined load polygons and assumes that these polygons are correct. C. Current Project The current project addresses the following: Automated creation of Load Polygons using a GIS platform: Clearly, the policy framework for the rural electrification planning process has been laid, both by government and Eskom. Hence there is a need for improvements in current planning tools. This paper proposes the automated creation of load polygons for the rural electrification planning process, using a GIS platform. This process will aid in the planner in producing quality plans at an accelerated rate. Load Polygons: Load Polygons involve the clustering of individual consumers together. This is a critical stage of the planning process, in that all consumer-related planning information is represented as an aggregate associated with a specific load polygon. In many cases load polygons can be identified ‘by inspection’ since consumers build their homes close together to form natural clusters. However, in certain rural areas settlement patterns are almost uniformly distributed over large areas and careful attention has to be given when choosing the members of each polygon. The clustering of consumers into distinct polygons is a critical aspect and, unless it is done correctly, may lead to inaccuracies in the planning of the electrification infrastructure. The clustering process is crucial for the success of an automated placement and will later influence all activities of design [3]. This paper thus presents an engineering tool aimed at automating the process of determining the load polygons. The solution combines the functionality of a geographically referenced database with a sophisticated software interface. The operation on a GIS database will ultimately ensure a standardized and more economical design. The information is

0-7803-7285-9/01/$17.00 (C) 2001 IEEE

displayed on Intergraph’s GeoMedia™ and the software interface manipulates the data and stores the results in the database. Additional benefits such as material management and the reduction of tedious repetitions in the planning phase will ensure a cost effective procedure and a more accurate plan. Note: most of the newly electrified rural regions deal with very low-income consumers and therefore the energy consumption is small. Optimization automated tools are therefore necessary for the capital and running costs to be reduced and the investment to be a made a profitable one II. Geographical Information System A. What Is GIS? GIS [4] integrates database management, computer graphics and spatial modeling into a software environment for the managing of graphic features. Thus Intergraph’s GeoMedia (a GIS mapping software) was chosen as a platform for the creation of the load polygons. It allows all necessary information to be displayed in layers and allows the user the maximum options for the optimization of the polygon creation. This is achieved by means of a dynamic link to the database. A GIS is mapping software that links information about where things are with information about what things are like. Unlike with a paper map, where "what you see is what you get," a GIS map can combine many layers of information, and allows the planner to access various other information like dams and other geographical information that will effect the load polygons, in order to optimize the plan. Figure 1 shows how the layers are arrayed:

Figure 1: layers on the GIS platform- as chosen by the planner to meet his current requirements

451 0-7803-7287-5/01/$17.00 (C) 2001 IEEE

III. Load Polygon Creation A. General Overview The following steps are followed in order to obtain the load polygons. The concepts used here are fully described in the further sections. 1.

All consumer related information is imported from various data sources (Eskom’s database, town planners records, consultants) and stored in a central database that is linked to GeoMedia. 2. All the relevant information (consumer location, type of consumer etc) is accessed and displayed. 3. The benefit weighting factors as determined by Banks et al. [2] are then applied to the consumer information 4. The benefit points are then used to obtain the distribution density map. 5. The planner then specifies manually where the load polygons should start based of the most densely populated area- the centroid of the polygon. This is done on the GeoMedia platform and then exported to the database. 6. The database is then updated with xy coordinates and the planner’s inputs. 7. The software interface then uses a distance method to determine the members of the load polygons, by using various cost factors 8. The members of each polygon is then accessed and displayed via GeoMedia. 9. The planner then displays other features viz. rivers and roads to optimize the design. 10. If changes are made, a re-calculated algorithm is run, to determine the new costs involved. B. Project Parameters - What Data Is Required? The proposed solution has defined certain aspects that have to be followed in order to obtain the load polygons. The information needed for the consumers is as follows: the number and names of all schools, clinics, police stations, households etc., and the location of all potential consumers of electricity, which should include sites that are to be developed in the future. This information is currently available from various sources such as town planning records, Eskom’s HELP and Napoleon databases. Various weighting factors (see below) associated with the region must also be made available. This information will be used to display the consumer information on GeoMedia and to obtain a consumer distribution density map that is unique to every region. Figure 2 shows the attributes associated with each consumer within a load polygon as stored in the database.

0-7803-7285-9/01/$17.00 (C) 2001 IEEE

Figure 2. The attributes from the “CONSUMER” database as seen through GeoMedia.

IV. The components of the Automation Process A. Using The Benefit Points System To Display The Consumer Density This approach is used to assist in the electrification planning prioritizations and technology choices [2]. Potential consumers such as households, schools, health centres, police stations etc. are allocated benefit points. The single dwelling consumer is assigned a point of one (1) and all other consumer are referred accordingly. Each benefit-point has a base value, which is then multiplied by a weighting factor. The weighting factor is derived by considering the average income, the road infrastructure and other factors (social and geographical) about the region. This allows different regions to have different weighting factors and hence allow regions with the highest need for electricity to be considered first (i.e. put in the top half of the electrification priority list). This approach was chosen, as it is easy to apply to automated tools and is attractive to committees and planners. It allows representative committees and experts in the field to establish a uniform scoring and weighting system, which is relatively transparent and familiar to most decision makers. Thus the designer needs to obtain and apply the weighting factor to the base values to the region being considered. B. The Grid If the community is randomly distributed or scattered, it is difficult to see which part of the region is most densely populated. Thus, the “Grid Technique” was developed. This is a simple technique in that a grid is superimposed over the consumer layout and calculations done, using the benefit points system. The calculations produce a condition for a query to be run on GeoMedia. A grid is merely an array of rectangles. The smaller the sides of the rectangle, the more accurate is the display of the concentration of consumers. The most important aspect of the grid technique is to allow all major value-point clients (consumers with higher base values

452 0-7803-7287-5/01/$17.00 (C) 2001 IEEE

viz. police stations) to occur within a load polygon. This is easily achieved by moving the grid so as to include these clients within each grid segment. Figure 3 shows the grid superimposed over he consumer layout. The sides of the grid segment represent a 4km x 4km area in GeoMedia.

Note the points of darker concentration represent the more densely populated area. It gives the planner a visual and allows him to determine the most appropriate centroid. V. CREATING THE LOAD POLYGONS A. The Software Interface The application software is dialogue-based. It makes use of the DAO (specifies that the data source is a Data Access Object) database structure and is connected to the central warehouse. It is broken up into three sections: 1. 2. 3.

The planner inputs Equations used The distance method

1. The Planner Input The software interface asks the user to input: Figure 3 “GRID TECHNIQUE”. A grid superimposed over the consumer layout to produce a condition for a thematic map

Hence, all consumers are grouped into small grids and this is the basis for which a thematic map can be drawn. C. THE DISTRIBUTION DENSITY MAP The Thematic Map: A thematic map is merely an indication of a value or range of values on that map by a color. If these values occur on the map, it will be represented by the assigned color. A range of colors can thus be used throughout the map to indicate the density of consumer occurrence. Obtaining a thematic map: The benefit-points that appear in each segment of the grid segment are summed. This gives us a range of values so that GeoMedia can produce the thematic map. Figure 4 shows the results: the points of high consumer concentrations. The planner uses GeoMedia to input his decision as to where he wants the centroid of the transformer zone to be. This is a manual process and give the planner total control of the positions of the zones.

1. The maximum expenditure for this region 2. The cost per point/ cost per connection 3. The cost per km of LV cable These cost inputs are used as the condition to limit the members of each polygon, viz. the number of members in each polygon is decided by the budget allocated to that region. Detailed costing for the region is done by consultants for Eskom and thus a break down of the costs are available. The maximum expenditure for this region should only be the money made available for the conductor costs. The type of service conductors and Low Voltage conductors will be the figure for the cost of the LV conductor input and the cost per point / cost per connection is standardized by Eskom at a cost of R 2800 (varies from region to region). 2. The Equations The benefit-points are used to calculate the consumer cost per connection. This is as follows: cost/connection = cost/point * value-points

(1)

where the cost/point id determined by Banks et al.[2]. The cost of the LV system is calculated by the equation: cost of LV = cost/KM * length

Figure 4. The Thematic map

0-7803-7285-9/01/$17.00 (C) 2001 IEEE

(2)

These equations allow the program to do a consumer-byconsumer cost calculation and hence determine the exact amount spent on the project. These equations are used for every consumer. Note, a ¥2 factor was multiplied to all lengths of the cables. This was done to account for the routing of the cable around corners / objects and the service

453 0-7803-7287-5/01/$17.00 (C) 2001 IEEE

cables to be installed at each consumer. This is a slight over design, but is compensated at consumers where more cable is needed. 3. The Distance Method The “distance method” was used as a condition to decide which consumer is connected to the transformer and which “connected” consumers (to the transformer) are connected to the next consumer. The methodology is simple, but proves effective. Firstly all centroid positions, as inputted by the planner, are identified and put into an array. This is then written to the database and is manipulated by the software interface. The software application then works with one centroid at a time and the distances to each consumer is determined and stored in a second array. The array containing all distances is sorted in order of increasing magnitude to speed up the process. Then the software application searches for the first shortest distance from the transformer point / centroid and connects this point directly to the transformer point- the first connected consumer It then checks to see if the distance from the transformer position to the next consumer is shorter than the distance from the chosen connected consumer to the next consumer. It does not connect all points to the centroid / transformer, but that which forms the shortest distance from each other (a chosen consumer and the next consumer). It implements only the shortest connections rather than just connecting all consumers to the centroid.

attribute in the warehouse / database. This means that this consumer is connected to that specific transformer. The total length of cable, the total cost and the total number of connections is also calculated and stored in the database under each transformer / centroid. This process is repeated for all centroid. VII. RESULTS The load polygons are easily displayed using GeoMedia, which allows all consumers with the “loadpoly” attribute = to A , B, C, etc. to be displayed via a query. Figure 6 shows the output obtained using a R 2800 / point and a R 500 000 maximum expenditure.

Figure 6. The Load Polygons

1

3 2 Figure 5. The Distance Technique

Clearly, the route that will be followed: 1 and then 2 (if the star marks the transformer / centroid) as compared to going from the transformer via route 3. This proves cost effective in the use of materials in that the shortest distances are considered for the use of the cabling. VI. THE AUTOMATED CREATION OF POLYGONS The software uses the “distance method” working with one transformer at a time. The distances are compared and the shortest distance is found. Equations 1 & 2 are then added and if the total cost is less than the maximum expenditure allowed, the programs writes an A, B, C, etc., depending on the number of transformer / centroid zones, to the “loadpoly”

0-7803-7285-9/01/$17.00 (C) 2001 IEEE

The load polygons can be clearly seen, and additional consumers can be added to the transformer, if the cost can be accommodated. Thus the designer is assured of the maximum number of consumers that the transformer can supply. The planner can then display other layers such as roads, rivers and mountains to optimise the design of the load polygons. If members are changed either added or removed, a “re-calculation” program is run to update with the new expense for the region and number of members in that zone. VIII. LIMITATIONS AND FURTHER WORK REQUIRED The solution described in this paper produces the load polygons for remote rural communities to be electrified. There are a few areas that could be improved to improve on the solutions obtained. These include: 1.

Data availability and quality- the data viz. consumer layout, maximum expenditure, cost per point/cost per consumer and cost per LV cable should be correct and verified.

454 0-7803-7287-5/01/$17.00 (C) 2001 IEEE

2.

The routing of the shortest distance for the cablesapply an optimisation tool, as this approach merely considers a √2 factor.

Electrification design Software”, IEEE Computer Applications in Power, vol 11, no.1, January 1998. (ISSN 0895-0156) XI BIOGRAPHIES

IX. CONCLUSION The infrastructure or policy framework for rural electrification has been laid. All tools that will make the planning process easier are being considered and utilized. This paper presented an engineering tool to automatically determine the members of a load polygon. Load polygons form an important part of the planning process the basis of many other proposals [3]. They are critical for further design decisions and the quality of the design. The planner can be sure that the maximum number of consumers is connected to the transformer. This allows plan to be efficient from the beginning of the planning stage and does away with the long tedious over designing and repetition of the planning process. The designer will be confident about accurate results of his design. The automated load polygon creation method does work. This will prove to be a vital tool for planners in the future.

X. REFERENCES [1] This article appeared in the Sep 27 2000 Robyn Chalmers Business Day 1st Edition http://www.bday.co.za/bday/content/direct/0,3523,7083616099-0,00.html [2] D I Banks, F Moche, E C Jonck, E Labuschagne, R Eberhard, “ELECTRIFICATION PLANNING DECISION TOOL”, Rural Area Power Solutions, Mockes Bussiness Consultants, Megasub Software Development, Cape Assoc., paper prepared for DOMESTIC USE OF ENERGY CONFERENCE, Cape Town, April 2000, www.raps.co.za

Govender Krishnan received his BSc (Eng) in electrical engineering from the University of the Witwatersrand, Johannesburg, South Africa, in 1999. HE is currently working towards an MSc at the same University.

Alan Meyer is an honorary lecturer in the Department of Electrical Engineering at the University of the Witwatersrand, Johannesburg, South Africa. He obtained a B.Sc.)Eng) degree in 1952 and an MSc (eng) in 1968. He returned to the University after a long career in industry and consulting engineering practice. During this time, he worked initially as a designer of electric machines, then as a general manager of GEC Large Machines Company in South Africa and as director of the GEC Power Distribution with responsibility for the technical development. This latter included equipment for power distribution. After leaving this organization, he became the electrical partner in the multidisciplinary consulting engineering firm. Barry Dwolatzky received his B.Sc (Eng) in electrical engineering and PhD from the University of the Witwatersrand. He is the acting head of the Department of Electrical Engineering at the same University. During the 1980s, he spent several years doing postdoctoral research at the UMIST, Manchester and at the Imperial College, London. His major current area of interest is in the design of software tools for the use in the design of low-cost electrical distribution networks in developing countries.

[3]. Z. Sumic, S.S Venkata, T. Pistorest, Automated Underground Residential Distribution Design. Part 1: Conceptual Design IEEE Transactions on Power Delivery, Vol. 8, No. 2, April 1993. Paper presented at the IEEE/PES 1992 Winter Meeting, New York, January 26 – 30. [4], R. Kasturi et al, “Map Data Processing in Geographic Information Systems”, IEEE Computer, pp. 10-21, December 1989. [5] T. Rajakanthan, A.S. Meyer, B. Dwolatzky Computer Generated Transformer Zones as part of Township

0-7803-7285-9/01/$17.00 (C) 2001 IEEE

455 0-7803-7287-5/01/$17.00 (C) 2001 IEEE