| Summary: | 碩士 === 國立臺灣大學 === 建築與城鄉研究所 === 92 === In urban study, there are many redistricting problems (RP), which subdivide spaces based on various activity intensities. In the past, such problems usually set their objective functions in accordance with the Euclidian distance from certain points. However, distance is not the only consideration, in stead, market shares is probably the other factor to determine the service areas of some industries, like electronic power, telecommunications or cable TV. Therefore, a new algorithm for market based RP (MRP) is needed.
In this study, to comply with the MRP which has multi-minimal solutions, the simulated annealing algorithm (SA), capable of escaping from a local minimal to seek a global one, is adopted by taking power service MRP as an illustrative case. The study area is represented as a grid system. After estimating the volumes of power consumption for every grid, the study area is partitioned into a given number of contiguously grid regions subject to the total difference of power consumptions of all the regions is minimized. This research focuses on finding appropriate initial values of some parameters of SA to improve computational efficiency while satisfied solutions can still be obtained.
In the illustrative case, it concludes that: (1) SA is confirmed to be a valid algorithm for MRP; (2) A grid size representing a 2×2 KM2 area is recommended; (3) Given a tolerable error, the number of iteration times can be sufficiently set to to have credible solutions with 95% confidence level.
|
|---|
