【作者】 赵强；

【导师】 敬东；

【作者基本信息】 四川大学 ， 电力系统及其自动化， 2003， 硕士

【摘要】 配电网络规划是一个复杂的组合优化问题，其具有多约束、非线性和离散性的特点。由于该问题的NP困难性，我们找不到可在多项式时间内求得该问题最优解的算法。为了满足求解此问题的时间、空间要求，寻找某个次优的、可有效计算的可行解成为自然而合理的办法。启发式方法就是基于这一思想所设计出的。 用传统的优化方法求解配电网络规划问题时，往往只是求得局部最优解。近年来，现代启发式智能方法被用来求解此类问题取得了较好的效果，这些算法具有全局寻优以及通用性强的特点。本文使用一种新型的模拟进化算法求解配电网络规划问题，研究表明该算法在求解组合优化问题是有效的。算法中，蚂蚁之间通过信息素进行交流，相互协作，使得蚁群表现出智能行为。在文中，蚁群算法通过不断修改各待建支路上的信息素，使得问题的解不断得到进化。 现代启发式智能算法在求解配电网网络规划问题时，其解的表示不能很好地处理网络的辐射性约束条件，导致算法搜索大量不可行解。现有的解的表示方法大部分都不能彻底的解决此问题，本文针对配电网络的辐射型特点，提出一种新的解的表示方法，文中使用图论中的支撑树来表示问题的解，该方法避免了配电网的辐射型检查，使得算法的搜索域限定在可行解域，从而改善了蚁群算法的搜索效率。算例表明该算法的实用性和有效件。更多还原

【Abstract】 Distribution system planning is a complicated combinatorial optimization problem, which is characterized by multi-constrains, non-linearity and discreteness. Because of NP-hard, we cannot find the polynomial algorithm which can achieve the best solution to the planning problem. For satisfying the requirements of the time and space, it is reasonable and natural to compute certain sub-optimal feasible solution, and heuristic methods are designed based on such an idea.The traditional optimal methods usually only compute the local optimal best solutions to the distribution planning problem(DPP). In recent years, the modern heuristic and intelligent methods can efficiently solve such complicated discrete problem, and these algorithms are characterized by global optimization and generalization. This paper employs a new type of simulated evolutionary algorithm, Ant Colony Algorithm (ACA), to solve the DPP. Many combinatorial optimization problems have been efficiently solved by ACA. In the algorithm, in order to make the ant colony show an intelligent behavior, the ants cooperate by exchanging the pheromone. Through changing the pheromone of the candidate feeders, the algorithm gradually achieves the better solutions.But most of these algorithms have not find a good method to represent thesolution of DPP so that many unfeasible solutions are searched duringcomputation process. According to the radiallity of distribution system, a new representation of the solution is suggested. This paper uses the spanning tree of Graph Theory to represent the solution so that radiallity checking procedure is avoided. In addition, because this kind of solution representation makes the searching process confined to only feasible solutions, the computational efficiency of this algorithm is greatly improved. The example systems demonstrate the effectivity and the practicality of the proposed method.更多还原