文摘
Distribution system planning is a process through which an electric utility assures that the delivery of electric energy to its customers is reliable, efficient, and economical. Good plans will save the utility significant portions of their initial investments and reduce the annual operating costs without sacrificing the service reliability of the system. Many researchers have addressed different aspects of such a planning process. No matter what kind of methodologies these researchers employed, they all agreed that three issues--uncertainties of loads, geographical deployment of the system, and large number of components--complicate the planning process. The spatial distribution and connection of customer loads are hard to model in an accurate mathematical form, thus making most optimization techniques inadequate for planning studies. The combinatorial explosion of the search space and vagueness in performance evaluations have made distribution system planning too hard to optimize. Instead of searching for absolute optimality, this work attempts to develop a framework that evolves a better solution based on available spatial heuristics and human knowledge in a fully automated planning environment. In so doing, this dissertation firstly devises a generic data model that utilizes geographic information system (GIS) capabilities in spatial modeling and management. Secondly, it develops a simple planner that generalizes the planning tasks from the topological perspective. Thirdly, it exploits a set of spatial heuristics that helps filter less desirable candidates while searching for the optimal design in the planning process. Fourthly, it proposes a heuristic based search algorithm for optimal cable routing. Finally, it presents an intelligent, adaptive search algorithm with implicit parallelism that employs the learning capabilities of genetic algorithms (GA's) as an underlying mechanism to help approach optimality in the planning domain. The novel features of this work are (1) its innovation, according to the geographical relationship among different objects, to streamlining the planning process of the different analogous levels in a distribution network, (2) taking advantage of spatial relationships to resolve the NP-hard cable routing and site selection problems, and (3) evolving high quality designs that are near-optimal and unable to explore in a traditional problem solving environment. Practical examples of secondary system planning, primary system planning, and street lighting planning demonstrate the feasibility of the proposed approach.