This paper deals with the techniques and problems of modelling very large networks which exceed computer hardware resources. Emphasis is placed on the mechanics of modifying the data in preparation for modelling. Two very different databases are addressed - one is a very sample database; the second is a very detailed database. Brooklyn Union Gas (BUG) has been modelling very large networks for many years with varying degrees of success. The successful solution is not easily defined because the techniques used are not conventional and require additional measures of acceptability. I cannot claim that the techniques we employ to reach a solution through modelling are always correct and valid. I do feel that the solutions reached are acceptable when backed up by a good working knowledge of the gas system, modelling experience, and an awareness of the limitations of the design tool in varying applications. Three different size gas piping systems are modelled at BUG - small (less than 300 nodes), moderate (less than 3000 nodes), and very large (20000-40000 nodes). The present company hardware and operating system are capable of handling a 13000 node/17000 NCE problem which requires 6 megabytes of real storage. The purpose of this paper is to focus on the techniques we use to model very large piping systems. The paper is presented in three parts. Part I deals with the problems of handling our very large Brooklyn/Queens low pressure networks which exceed our computer hardware resources; Part II deals with the problems of extracting network data from our Staten Island DFIS corporate data base; Part III discusses the pros and cons of engineering self-sufficiency in this computer application.

A. Background

BUG services approximately 1.1 million customers are supplied from our Brooklyn/Queens (B/O) low pressure system. This system is predominately a 4" and 6" Cast Iron interconnected/looped grid which exists under every street in Brooklyn and Queens. Intersections are normally 4-way interconnections. There are no discontinuities in this system of pipes. Likewise, the 48000 NCE's and 36000 nodes in the data base description are interconnected and continuous. This data base was created and is maintained by our Engineering department. It is not a corporate data base. The data base contains only pipe attributes necessary for modelling diameter and length. It does not contain information on welds, couplings, joints, bends, valves, offsets, and reducers, but instead throws in an efficiency factor to adjust for these elements, as well as any loss in cross sectional area due to deposits left from when gas was manufactured. Pipe roughness is also assigned to each pipe based on diameter. A clerk within my group, working from revised field installation documents issued by the Drafting section of our Distribution department, updates our data base. Loads are assigned to appropriate nodes through a ‘findering’ system commonly used in the industry. In our case, nodes are assigned ranges of customer account numbers (therefore billing history) that fall within 1/2 block of the node.

This content is only available via PDF.
You can access this article if you purchase or spend a download.