中文摘要：

逐步优化算法（POA）在求解梯级水电站联合优化调度中,其优化结果受初始解的影响较大,不同的初始解对优化迭代过程的收敛性影响不同,导致优化结果可能陷入局部最优。针对这一问题,本文在深入分析POA寻优机制的基础上,探求了影响算法全局收敛能力的关键因素,揭示了POA的两阶段寻优策略和梯级水电站优化调度在求解两阶段问题时传统的＂自上而下逐电站＂寻优模式对算法收敛能力的影响规律,进而提出了基于逐步差分和变阶段优化改进策略的变阶段逐步优化算法,有效消弱了原始算法在求解梯级电站联合调度问题中对初始解的依赖性,在一定程度上保证算法收敛于全局最优解。实例研究表明所提算法优化得到的梯级发电量比POA算法提升0.15%左右,有效克服了原始算法的局部收敛问题,且改进算法效率更高,寻优结果更稳定。

英文摘要：

In optimal dispatch for cascade reservoirs, the progressive optimality algorithm（POA） suffers from premature convergence caused by poor initial solution. To solve the problem, this paper explores the key factors that influence the global convergence behavior of POA and presents a variable period progressive optimality algorithm（VPPOA）. This new algorithm adopts a progressive differential strategy and a variable period strategy to replace respectively the approach used for solution of two-period optimal problems and the two-period optimal strategy in traditional POA, so that it can overcome the premature convergence problem caused by poor initial solution. Simulation results of optimal dispatch for the lower Jinsha River show that the gain in total electricity generation from VPPOA optimization is 0.15% more than that from POA and VPPOA-optimized solutions are nearly the same as those globally-optimized. Thus, this algorithm significantly reduces computational cost and improves the optimal dispatch of cascade reservoirs.