Optimization of Distribution Overhead Powerline Design Using Genetic Algorithm with Memory

Abstract

The increasingly heavy standardization of distribution overhead powerline installations presents an opportunity for the automated design of distribution overhead powerline pole structures, attachments and conductor spans. A successful design automation algorithm must be capable of generating distribution overhead powerline designs that meet all relevant code requirements and utility company standards, be capable of producing a design that is ideally more cost effective than that of a typical human-created design and be able to perform the optimization process within a reasonable amount of computing time. A genetic algorithm optimization tool is developed for use by a distribution facility operator whose service area includes powerline in rural Alberta, Canada. The optimization tool can interpret a survey comma separated value file along with limited user input and then using the supplied data to carry out the economic optimization of a distribution powerline design subject to constraints such as pole structure force loading, conductor span vertical clearance, conductor uplift, grounding and span-tension continuity. The resulting output from the tool contains a completed design in the form of several design documents that comprise a substantial component of a construction design package which is intended for use by the distribution facility operator's design department. Upon testing the optimization tool on a 10 pole, 15 pole and 40 pole three phase distribution overhead powerline new extension, it is found that the tool can produce designs that not only comply with all code requirements and standards but can also result in fewer design omissions compared to the corresponding human-created designs. The overall cost efficiency of the optimized designs either meet or exceed the human created designs by a slight margin. Finally, the total optimization time for the 40-pole structure powerline design using a high-performance desktop computer is found to be almost three hours.