基于协同进化粒子群算法的水库优化调度与应用

Study on the Reservoir Operation Based on Cooperative Coevolutionary Particle Swarm Optimization Algorithm

水库供水优化调度中存在多个供水目标、多个决策变量和复杂多约束条件,从而呈现出高维度、非线性、强约束特性。针对传统粒子群算法求解此类问题容易出现的收敛速度慢、计算效率低及早熟问题,将合作型协同进化思想与粒子群算法相结合,提出一种基于种群停滞搜索技术的协同进化粒子群算法,一方面通过种群内部个体间的竞争进化模式来提高种群竞争力,另一方面通过种群之间的相互合作模式提升算法全域搜索能力,各种群依次实行进化过程和协同过程,以保持种群进化过程中的多样性,并从“种群进化过程监视”的角度出发,提出了防止算法早熟的种群停滞探测技术,提高算法收敛速度。将提出的算法应用于徐家河水库供水优化调度模型求解中,结果表明,相对于传统粒子群算法,本算法计算的生活、工业和灌溉累积缺水总量分别降低了47.2%、33.3%和14.4%,供水保证率分别提高了1.7%、1.9%和4.4%,缺水指数分别降低了0.064、0.071和0.076,年均弃水量减少了1.9%,水资源利用效率有所增加。算法性能方面,协同进化粒子群算法在迭代早期(约65次迭代)就开始收敛,并在一定程度上避免了粒子陷入局部最优,降低了算法的不确定。综上表明,本文提出的算法易于实现,求解效率高,为水库优化调度模型求解提供了新的思路。

There exist multiple objectives,multiple decision variables and complex multi-constraint conditions in the reservoir water supply dispatching,which presents the characteristics of high dimension,nonlinear and strong constraint.Aiming at the problems of slow convergence rate,low computational efficiency and precocity,a cooperative coevolutionary particle swarm optimization algorithm(CCPSO)based on the population stagnation detection technology was proposed,whitch combines the idea of cooperative coevolution with particle swarm optimization.CCPSO improves the competitiveness of the subpopulation through the competitive evolution among the individuals,improves the global search ability through the cooperation among the subpopulations,and improves the convergence speed through the population stagnation detection technology,and improves population diversity through the evolutionary synergy of various populations.CCPSO is applied to solve Xujiahe reservoir water supply dispatching model,compared with the original particle swarm optimization algorithm,the cumulative total water shortage of domestic,industrial and irrigation calculated by CCPSO is decreased by 47.2%,33.3%and 14.4%,the probability of water supply is increased by 1.7%,1.9%and 4.4%,and the water shortage index is decreased by 0.064,0.071 and 0.076.The results show the application of CCPSO can improve the efficiency of water usage.In terms of algorithm performance,CCPSO can start convergence earlier(about the 65 iterations),can avoid falling into local optimality,and reduce uncertainty.Therefore,the algorithm is easy to be implemented and the solution efficiency is high,which provides a new idea for the solution of reservoir optimal dispatching model.