云理论在配电网络变电站选址定容中的应用

Application of Cloud Theory to Optimal Planning of Substation Locating and Sizing

变电站选址与定容优化规划属于大规模组合优化问题,基于云在定性概念描述与定量数值表示之间转换过程中的优良特性,借鉴遗传算法"优胜劣汰,适者生存"的进化思想,提出一种基于云理论的优化算法(cloud theory optimization algorithm,CTOA),并将其应用于电力系统配电网络变电站站址、站容的优化规划研究中。在该算法中,用云的期望代表父代个体的优良特征,用云的熵和超熵控制遗传和变异的程度,用正态云算子完成概念空间到数值空间的转换,产生种群,实现遗传操作。CTOA在定性知识的指导下能实现空间范围的自适应控制搜索,可以有效改善智能优化算法易陷入局部最优解和早熟收敛等问题。最后结合某装备制造基地变电站选址定容实例,分别采用改进自适应遗传算法(improved adaptive genetic algorithm,IAGA)、改进多组织粒子群优化算法(refined multi-team particle swarm optimization algorithm,RMPSO)和CTOA算法对其进行了优化规划研究。结果表明,CTOA在收敛时间,搜索精度性能指标方面优于IAGA、RMPSO算法,且该算法无需编码,操作流程简单,易于实现,能更好的满足配电网络中大规模变电站规划的需求。

The optimizational planning of locating and sizing for a distribution substation is a large combinatorial optimization problem. To solve this problem, a cloud theory optimization algorithm （CTOA） is proposed based on the outstanding characteristics of cloud in transforming a qualitative concept to a set of quantitative numerical values, and in combination with the principle of “survival of the fittest” of the genetic algorithm. The core thought of the algorithm is that the hereditary characteristic of the parent individual first is represented by the expected value of cloud, then the degree of the heredity and mutation is controlled by entropy and hyper-entropy of the cloud, and transformation between the qualitative concepts and their quantitative expression is realized by normal cloud, so as to realize the reproduction. With the instructions of the qualitative knowledge,the extent of the searching space is self-adjusted, and the possibility of prematurity and the probability of trapping in local best optimization are greatly reduced. The proposed CTOA is tested by a realistic planning project to verify the effectiveness and feasibility. The calculation speed and search accuracy of CTOA are obviously superior to those of the improved adaptive genetic algorithm （IAGA） and refined multi-team particle swarm optimization algorithm （RMPSO）, without the process of coding and crossover and easy implementation. The method proposed has a promising application in large-scale practical problems.