【摘要】 基于以发电量最大为目标的水库优化调度模型，分析了4种具有代表性的优化调度算法，即动态规划（dynamic programming, DP）算法、离散微分动态规划（discrete differential dynamic programming, DDDP）算法、逐步优化算法（progressive optimality algorithm, POA）和浮子算法（float algorithm, FA）的理论基础及求解流程，并以南欧江流域和沅江流域的多个水电站为实例，对比分析了各算法的求解效果及效率。结果表明：DP算法的结果精确，但计算时空复杂度较高，DDDP算法往往能收敛至全局最优解且相对高效，POA需要较优的初始解才能收敛至全局最优解附近；FA计算效率高且解的精度较高，在不发生弃水时几乎可直接得到最优解，在发生弃水时亦能得到较优的初始解。基于水库发电优化调度运行规律建立的浮子算法为水库优化调度模型求解开辟了新思路，具有较好的应用前景。更多还原

【Abstract】 Based on the reservoir optimal operation models with the objective of maximizing power generation, the theoretical basis and solution procedures of four representative optimal operation algorithms, namely dynamic programming(DP) algorithm, discrete differential dynamic programming(DDDP) algorithm, progressive optimality algorithm(POA), and float algorithm(FA), are analyzed. Taking several hydropower stations in the Nan’ou River Basin and the Yuanjiang River Basin as examples, a comparative analysis of the solution performance and efficiency of each algorithm is conducted. The results show that DP algorithm yields accurate results but has high computational time and space complexity; DDDP algorithm can often converge to the global optimal solution and is relatively efficient in computation; POA requires a high-quality initial solution to converge near the global optimal solution. FA exhibits high computational efficiency and high solution accuracy: it can almost directly obtain the optimal solution when no water abandonment occurs, and can also yield a high-quality initial solution when water abandonment happens. The float algorithm, established based on the operational rules of reservoir optimal power generation dispatching, opens up new avenues for the solution of reservoir optimal operation models and has promising application prospects.更多还原