Tricks of the Trade

Tricks of the Trade

Jan Scheurwater
Tensing”SKS
Wielkamp 3
NL 5301 DB Zaltbommel
Netherlands

Introduction
By the end of the 19th century the availability of electricity, gas and water from public utilities was a privilege for few in the urban areas in the prosperous parts of the world. Now we regard these utilities as ordinary facilities that should be available to everyone. During the last decades, the use of computers in the operation and control of utility networks has become standard. Many major utilities around the world have the information that is required to operate their networks available in digital form and depend on it for their services. Where utilities in the past often enjoyed the luxury of a de facto monopoly, they now are confronted with competition. This requires focus on customers and efficient operation. The availability of digital information about the utilities’ key assets, their networks, facilitates the integration of new techniques that will give them the competitive edge.

To that end, a group of new techniques is available that originates from within Operations Research (0.R.). This paper will discuss the general background of O.R. and indicate applications for the utility industry. It will expand on the use of O.R. techniques for the design of networks and indicate the major requirements on the GIS systems and applications that are used in our industry.

What is ‘tO.R.”?
Operations Research (or: Operational Research or: Mathematical Programming) is a branch of applied mathematics that arose since World War II in conjunction with the development of computers as an answer to the growing need for tools to solve complex production and scheduling problems. O.R. is concerned with the solving of mathematical optimization problems. Many of these problems are combinatorial in nature, i.e. a very large number of possible “solutions” exist, and the problem is to find the best (or, in many real world cases: a very good) solution. A large group of problems relevant to GITA consists of so called network optimization problems. Some well-known examples (with their translation to our business world) are:

shortest route from A to B over a network. How do I best travel from A to B, but also: Given available space in conduits, what is the best way to add a cable that connects A with B.
traveling salesman problem. The salesman/field engineer has to visit a number of clients and wants to minimize total distance traveled.
minimum spanning tree: given a particular network, what is the shortest sub-network that can be designed to serve all customers.
Steiner Tree: define the network that connects given points using the minimal amount of wire.
Chinese Postman: what is the shortest route that traverses every link in a network at least once (e.g. what is the best route for checking equipment such as street lights, poles, cabinets, manholes).
Network flow: what is the maximum flow through a network. This is implicit in any gas or water network calculation package.
and many more.

In the above examples “short” can be replaced by “cheap”, as distance is used as an arbitrary way of measuring cost. E.g. the shortest route from A to B is associated with the route that requires the lowest costs in terms of material and construction.

Real world problems often are too large and too complex to find the one and only “best” solution. Even with the best computer hardware available today time and cost required for identifying the “best” solution may be prohibitive. In recent years techniques have been designed to identify “good” solutions in a limited amount of time. The quality of the results generated by so called “genetic algorithms” is usually a major improvement over solutions designed without computer assistance as well as quite sufficient for practical purposes.

It will be evident that there are many application areas within the scope of the utility industry for Operations Research. The main usage can be identified in network design and in routing of servicemen. The techniques have existed for a number of years but it is only now that sufficient digital data is becoming available for application in the AM/FM industry. On the other hand, the application of these techniques reveals all too often that the data collected and digitized at great cost is inadequate for O.R. purposes. Paraphrasing an old carpenter wisdom, “Measure once, saw twice; measure twice, saw once”, we could say “Think once, digitize twice; think twice, digitize once”. The use of these techniques can lead to significant savings in cost and in response time and the incorporation of these tools in daily practice will improve the utilities’ competitiveness significantly.

Network construction & what O. R. can do for you
The design of a network in a newly built area starts with an inventory of relevant data. Supposedly, the number and the location of the future customers are known, either from a layout plan or from field surveys of customers-to-be. Generally, the network will be positioned alongside the roads and in urban areas underneath the curbs. Only a limited number of locations are suitable for the location of equipment such as high voltage transformers and gas stations. These stations have to be connected to the existing mains network. Each location has an associated cost of linking it to the mains network. Also, the location itself has costs associated to it. The costs per unit length of network construction depend on the type of cable but also on the method of construction. Overhead networks have a different price structure from underground networks where the costs of digging depend on issues like soil type, type of pavement or the availability of conduits with free space. Road crossings are usually costly, so limiting the number of road crossings can be a design issue.

With respect to other issues, no precise knowledge is available and planning and design will be based on assumptions. For example: what percentage of households will subscribe to the services at a later point in time? Does that require additional network capacity? In the past many utilities had a local monopoly and could build over-dimensioned networks. In today’s competitive world this is no longer an alternative. In fact, the opposite is the case: what use can be made of the over-dimensioned networks of the past? Is a network upgrade really necessary or does the existing network have sufficient capacity?

There are major differences in the design details (and consequently, in the definition of the Operations Research problem that has to be solved) between countries. In large parts of North America the electrical and telephone networks are above the ground. This gives a great deal of freedom of choice for the network location. In urban areas overhead networks may not be an option and a system of conduits may have to be designed. In areas that do not have a rocky soil (such as all of the Netherlands) the preferred way of building the network can be to dig trenches and put the network in the ground without protection.

What can current applications do?
Several consultants and software vendors do have applications available that assist network designers in the fast generation of cost effective network alternatives. In their simple form, these tools require the following steps by the designer

specify the area that has to be designed,
specify the customer locations,
specify locations for transformer stations,
provide digital base maps, land use data, etc.

The software will than automatically generate the least cost network layout based on context specific rules. Such rules may be:

the network location should be parallel and at a predefined distance to the roadsides,
in some areas digging is not allowed,
soil types, land use and street surfaces have specific associated construction costs,
the crossing of a road may not be allowed,
above ground networks have to consider tree crowns, underground networks similarly have to consider the tree roots. Cutting of branches and roots may be prohibited by environmental regulations,
.......

Of course this type of software can also generate a bill of materials and will produce cost reports.

In some countries/areas, one utility company provides several services, e.g. gas, electricity and CATV. Asthecosts ofdigging area major padofthe construction costs, the integrated design and simultaneous construction of all networks may result in major savings. It also limits the inconvenience for your customers.

The methods used for integrated design usually start with the design of the most restrictive network (often gas). The other utilities will than use the same trenches but their own locations for cabinets and manholes.

What are the benefits?
In order to assess the benefits of the use of O.R. methods, one has to do the design twice: once in the standard (manual) way and once using the combination of GIS and O. R.. As this is costly and time consuming (and a bit of a threat to the manual network designer) there are not very many cases where comparable information is available. The PNEM, one of the major Dutch utilities has gone through this exercise for several residential areas (200 to 1200 houses). The network designs that were generated automatically were 10-1 5% cheaper in construction than the manual designs. The time to do the design with these tools went down to a third of the manual time. The time saved by automated design allows the designer to create and evaluate different alternatives. For example, you can create designs based on different locations for your transformer stations or on a different number of transformer stations.

It should be noted that the savings that can be achieved in this way quite often exceed not just the cost of implementing the O.R. software on top of a GIS, but also the cost of the data acquisition and conversion for the GIS. In other words: these approaches may pay for your GIS in one project.

Requirements to GIS applications
The effective use of Operations Research techniques in a GIS environment raises a number of issues. In fact it imposes requirements on the GIS environment, the O.R. software, the network data model, the spatial model for the geographic data used as well as on data quality.

To allow the integration of O.R. software in a GIS environment, both software components have to be open. It should be possible to build integrated solutions using standard interfaces. Furthermore the O.R. software should allow human intervention as a good design is usually generated by cooperation between man and machine. From the point of view of the utility company it is desirable that the GIS software and the design software are independent software components. Only then the choice of the GIS software does not influence the choice of the O.R. software and a component can be replaced when required.

Network design requires an adequate network data model in the GIS system. Part of the design process is network calculation. Network calculation requires integrated data about the location, the components, and the characteristics of the components and the connectivity of the network. Obviously scanned raster images will not do. The data structure of standard drafting software generally does not handle location and technical characteristics of components in an integrated way. Also, the most advanced GIS software is not always used to its full capacity since the requirements for network design were not considered at the time of data model design or regarded as too costly during data conversion.

The network design uses a range of “background” geographic information and consequently makes demands on these data. If you want to generate a network parallel to the street edges, the base map should contain all relevant street edges and these should be gee-coded as such. Also, the street edges should be modeled adequately as “polylines” (strings of lines and arcs). Short polylines (in particular if they consist of one line or arc) will have a dramatic effect on computer processing time. Also, the endpoints of these polylines should coincide exactly. In summary: The ideal base map for network design has a topological structure. Where land use data or road surface data is used, again this surface related data has to have a topological structure in order to calculate the length of network in a particular area.

Even an excellent GIS system does not guarantee the implementation of a good data model and even an excellent data model does not guarantee good quality data. Visual inspection of data on the screen or on a plotted map does not tell everything about data quality. The more advanced analysis and design applications are the real test for your GIS data. Network connectivity and topological structure can only be controlled by automated procedures. This must be available when the data is converted and when the data is updated. If these quality control procedures are not in place data quality will deteriorate in the long run. On the other hand ruthless data quality control is the key to improved company performance and customer satisfaction.

Conclusion
The use of Operations Research techniques in combination with GIS is still in its infancy. This paper has focussed on the integration of O.R. techniques with network design. An equally important area in the competitive utility world is the planning and routing of service and repair engineers. Clearly, customer satisfaction is directly affected by efficient correction of failures and outages. In addition to the requirements that have been discussed in the previous chapter, excellent means of communication between a planner and the field workers are a necessity. Now these are coming available, tools to assist field operations that integrate GIS, ERP-systems and O.R. techniques will emerge. In fact, integration of information systems will be the trend for the next decades as it will be the way to improve the competitiveness through cost savings. In the unfortunate case that current GIS systems do not meet the requirements for operations as discussed here, additional work will have to be done and additional costs will have to be made. But the way to recover the costs of GIS systems is not in doing the old manual process of record keeping the digital way. The only justification of these systems lies in their full exploitation in the daily operations of the utility industry, even though that may require business process re-engeneering.

The 20th century gave rise to an enormous leap in technology. In the utility industry, it is just the beginning.