跨流域调水条件下水库群联合调度图的多核并行计算研究  

Multi-core parallel computation for deriving joint operating rule curves in multi-reservoir system under the condition of inter-basin water transfer

作　　者：彭安帮[1] 彭勇[1] 周惠成[1] 

机构地区：[1]大连理工大学水利工程学院,辽宁大连116024

出　　处：《水利学报》2014年第11期1284-1292,共9页Journal of Hydraulic Engineering

基　　金：重大国际(地区)合作研究项目(51320105010);国家自然科学基金资助项目(51379027)

摘　　要：为提高跨流域调水条件下大规模复杂水库群优化调度的计算效率和求解精度,采用并行PSO算法进行联合调度图模型的多核并行求解。该算法充分利用PSO搜索速度快、天然并行性等特点,引入多种群思想保证种群的多样性,提高算法的全局收敛能力;采用基于分治策略的Fork/Join框架实现将子种群分配到不同CPU内核进行独立求解;利用Java并发过程中的同步和通信机制实现子种群间的信息交流,避免陷入局部最优。最后通过实例分析表明,多核并行PSO算法能够充分利用多核资源,有利于提高联合调度图模型的求解速度和精度,是解决大规模复杂水库群优化调度的一种高效实用的方法。In order to improve the computation efficiency and solution accuracy of the optimal operation in large-scale multi-reservoirs under the condition of inter-basin water transfer, a multi-core parallel PSO al？gorithm is proposed to solve the joint optimal operation model. This method with the merits of rapid search？ing and easy-to-parallel in basic PSO algorithm, also adopts a multi-swarm strategy to guarantee the swarm diversity and enhance the global searching ability. In addition, a Fork/Join framework based on the divide-and-conquer strategy is employed to assign each sub-swarm to different CPU cores for evolving sepa？rately. At the same time, the synchronization and communication mechanisms of Java are used to imple？ment the information exchange between sub-swarms to avoid trapping in local optimum. The results show that the proposed method can take full advantage of multi-core resources so as to effectively improve the convergence speed as well as the quality of solution. Therefore,it can be a high-efficiency method for solv？ing the optimal operation of large-scale multi-reservoir system.

关 键 词：跨流域调水 水库群 并行计算 多核 PSO算法 Fork/Join框架 

分 类 号：TV697[水利工程—水利水电工程]