中文摘要：

在无功优化协同进化计算中，将控制变量合理地分区分组是算法正常运行的前提，也是获得良好并行性能的关键。参考无功优化控制变量分区问题与分级电压控制中电网分区问题之间的关系，提出将控制变量分区问题转换为降阶电网分区问题，并构造降阶电网分区优化模型。在此基础上，引入种子分区编码方法，提出一种能自动确定分区数目的方法。该方法使用向上归并法进行初步分区，降低了分区规模，并采用种子分区编码法将分区数目等信息编入染色体，解决了分区数目难以确定的问题。系统计算表明，新分区方法能自动确定分区数目，快速地对系统控制变量进行合理地划分。将该方法应用到协同进化计算中，能提高协同进化计算的并行性，保证算法寻优效率。

英文摘要：

In coevohitionary computation for reactive power optimization, partitioning control variables correctly and properly is the basis on which the algorithm can operate normally, and may also be the key point to obtain high performance of parallel computation. According to its relations to the grid partition problem in the secondary voltage control, the problem of control variables partition is converted into that of grid partition, of which the mathematical model is also formulated in this paper. And then the coding scheme of seed cluster is introduced and a novel partition method which can determine the number of the partitions is proposed. This method partitions the grid roughly in the initial stage to reduce the dimension of the problem using agglomerative hierarchical approach. By coding the information of the partition number into the chromosome, the new coding scheme overcomes the difficulties about determining the partition number. In the numerical simulation on IEEE l 1 S-bus system, the new partition can determine the partition number automatically, and can group the control variables rapidly and correctly. The efficiency of coevolutionary computation can be promoted by this novel partition method greatly.