粗粒度并行自适应混合粒子群算法及其在梯级水库群优化调度中的应用

为了充分利用现今普及的多核配置计算机,提高大规模梯级水库群优化调度问题的求解效率,提出了梯级水库群优化调度的粗粒度并行自适应混合粒子群算法。该方法以自适应混合粒子群算法为求解基础,采用粗粒度并行设计模式,利用Fork/Join多核并行框架的分治策略,将其初始种群递归划分为多个子种群,平均分配到不同的内核逻辑线程中实现并行计算,并在各子种群优化结束后,合并优化结果集从而输出全局最优解。以澜沧江下游梯级水库群发电优化调度为例,利用该方法进行计算。结果表明,该方法能充分发挥多核配置的计算性能,在4核环境下最大加速比达到3.97,缩短计算耗时1 787.2 s,计算效率显著提高,为我国不断扩张的大规模梯级水库群优化调度提供了一种切实可行的高效求解途径。

Coarse-grained Parallel Adaptive Hybrid Particle Swarm OptimizationAlgorithm and Its Application to Optimal Operation ofCascaded Reservoirs

To improve the computing efficiency of optimal operation of large-scale cascaded reservoirs, a coarse-grained parallel adaptive hybrid particle swarm optimization (PAHPSO) algorithm is proposed in full use of the popular multi-core computers.The method is based on adaptive hybrid particle swarm optimization (AHPSO) algorithm, and adopts the coarse-grain model and divide-and-conquer strategy of Fork/Join multi-core parallel framework to divide the initial population into multiple small-scale subpopulations, which are assigned to different logical threads averagely for parallel computing.After the optimization computation for all subpopulations, the optimization result sets are merged to obtain the globally optimal solution.The proposed algorithm is applied to the generation and operation of cascaded reservoirs located on the lower stream of Lancang River.Results show that the method gives full play to multi-core computer performance, and the maximum speedup in 4-core parallel environment reaches 3.97 with the time-consuming cutting down by 1 787.2 s.The computing efficiency has improved significantly and it provides a feasible and efficient solution for the optimal operation of increasingly expanding large-scale cascaded reservoirs in China.