中文摘要：

为解决逐步优化算法（POA）求解库群长期优化调度时存在的“维数灾”问题，结合正交试验设计方法提出了正交逐步优化算法（OPOA），分别从阶段维、状态维和空间维进行降维求解。算法采用POA将多阶段决策问题分解为若干两阶段子问题，以目标函数为试验指标，水库为试验因素，离散状态为因素水平，各子问题的优化求解视为分别开展多次正交试验设计，通过逐次加密离散水库状态并构造“均衡分布、整齐可比”的状态集合进行计算，直至获得最优解。乌江梯级水库长期优化调度结果表明， OPOA仅需POA耗时的28.4%即可获得全局最优解，显著优于粒子群算法，是求解库群长期优化调度的有效算法。

英文摘要：

In order to overcome the curse of dimensionality of Progressive Optimality Algorithm （POA） in solving the long-term optimal scheduling of cascade reservoirs, Orthogonal Progressive Optimization Algo-rithm （OPOA） is proposed on the basis of the orthogonal experiment design method, reducing the dimen-sions of stages,the number of discretization of water level and cardinal number of set. Firstly,POA is em-ployed to decompose multistage decision problem into several two stage sub-problems, and then solve each sub-problem by carrying out the orthogonal experimental design multiple times, taking the objective func-tion as experimental index, reservoirs as experimental factors and discrete status as factor levels. All sub-problems are calculated respectively by building water levels in equilibrium distribution, until obtaining the optimal solution of each sub-problem successively. The computer simulation results of 4 reservoirs in the Wujiang River show that OPOA is distinctly superior to particle swarm optimization and takes only the 28.4 percent of the computing time compared to POA in obtaining the global optimal solution, which is an effective algorithm in long-term optimal operation for hydropower system.