机组组合的扩散并行遗传算法

研究了在细粒度并行机上的扩散并行遗传算法．遗传算法中个体为矩阵个体，选种采用竞争法．并行处理机拓扑结构为三维网格．对一个十机系统的机组组合问题进行了串行模拟，结果表明，当最大遗传代数或并行处理机个数增大时，均可找到更好的解，同时加速比也得以提高，且异步法优于同步法．

A Penetrating Parallel Genetic Algorithm for Unit Commitment

A penetrating parallel genetic algorithm on a fine grain parallel computer is investigated. In the genetic algorithm, matrix individuals and a tournament selection are used. Three dimensional grid is used as a parallel processor topology. A sequential simulating test is conducted upon a 10 unit commitment problem. It is found that the greater the maximal generation number or the number of parallel processors, the better the solutions. Meanwhile, the speedup can also be improved. In addition, asynchronous method is prior to its synchronous partner.