论文部分内容阅读
为探索高性能并行计算在水库群优化问题上的应用,该文以经典四水库问题为例,构建多维动态规划模型,采用主从模式策略对动态规划程序进行并行化,利用高性能并行计算机(多达300个核)对该问题求解,得到了不同核数的计算时间、加速比以及并行效率。结果表明:借助分布式计算,动态规划求解水库群优化的计算时间能够有效缩短,加速比将随核数增加进一步提升,并行效率减少趋势缓慢。未来工作需借助分布式内存来克服动态规划的内存过大问题。
In order to explore the application of high-performance parallel computing in reservoir group optimization, this paper builds a multidimensional dynamic programming model by taking the classical four-reservoir problem as an example, adopts the master-slave mode strategy to parallelize the dynamic programming procedure, and uses the high performance parallel computer Up to 300 nuclei) to solve the problem, get the calculation time of different nuclei, speedup and parallel efficiency. The results show that with the help of distributed computing, the dynamic programming can shorten the computation time of reservoir group optimization effectively, and the speedup ratio will further increase with the increase of the number of cores. The parallel efficiency decreases slowly. Future work will need to rely on distributed memory to overcome the dynamic programming of memory problems.