并行多维动态规划算法在梯级水库优化调度中的应用  

Application of Parallel Multi-dimensional Dynamic Programming Algorithm in Cascaded Reservoirs Optimal Operation

作　　者：王丽萍[1] 孙平[1] 蒋志强[1] 纪昌明[1] 张验科[1] 

机构地区：[1]华北电力大学可再生能源学院,北京102206

出　　处：《水电能源科学》2015年第4期43-47,80,共6页Water Resources and Power

基　　金：国家自然科学基金项目(51279062);中央高校基本科研业务专项资金项目(13XS23;13QN22;13XS24)

摘　　要：为有效缓解多维动态规划算法运行时间长并增强其对梯级水库群联合优化调度问题的求解能力,基于.NET4的并行拓展库,将并行处理技术与该方法相结合,利用梯级水库状态点间的计算独立性构建了并行多维动态规划算法,并以李仙江流域三库梯级为例,根据不同状态离散数和不同CPU核数构建了3个串行计算方案和9个并行计算方案,从运行时间、并行加速比及并行效率等方面对该并行算法进行性能分析。结果表明,在现有计算条件下该并行算法能有效缩短程序运行时间,从而缓解多维动态规划的维数灾问题,并继承了多维动态规划的全局收敛性。In order to effectively alleviate the long-running defect of multi-dimensional dynamic programming（DP）algorithm and enhance DP＇s ability to solve the joint operation optimization of cascaded reservoirs,the parallel multi-dimensional DP algorithm was proposed based on.NET4 parallel extension library by using the independence of calculations among the state points and the parallel processing technology.Lixianjiang cascaded reservoirs was used as the study instance in this paper,and 3serial computing schemes and 9parallel computing schemes were built according to the different number of discrete points and CPU cores.A detailed performance analysis of this parallel algorithm on run-time,parallel speedup and parallel efficiency had been carried out based on the calculation results of each scheme.The results show that this parallel algorithm can effectively shorten the program run-time and alleviate the dimension disaster of multi-dimensional DP algorithm under the conditions of existing computer hardware.

关 键 词：梯级水库 水库调度 多维动态规划 维数灾 并行算法 李仙江流域 

分 类 号：TV697.1[水利工程—水利水电工程]