水库发电优化调度模型的快速求解算法及应用

针对水库“以水定电”优化调度数学模型求解方法中，传统动态规划方法易陷入“维数灾”，而现代智能算法搜索效率低且易陷入局部最优解等不足。本文基于水库在高水头运行时耗水率小的特点，提出了水电站水库“以水定电”调度模型的快速求解方法——浮子算法。该方法在水电站满足保证出力和最小下泄流量等要求的前提下，尽可能提高水库调度期的运行水位以达到直接寻找优化调度解的目的。通过案例分析并与离散微分动态规划（DDDP）方法求解结果比较，本算法结果与DDDP算法十分接近，计算时间约为后者的50%。与其他算法联合使用，可在提高优化效果的同时缩短模型的求解时间，为水电站水库“以水定电”调度计划的制定和快速决策提供技术支持。

Fast algorithm for solving optimal power generation model of reservoir operation and its application

In solving models for optimal power generation of reservoir operation in a plan of regulating power generation by water supply, traditional dynamic programming are easy to fall into a dimension disaster; while modern intelligent algorithms using random search are inefficient and easy to fall into a local optimum. To overcome these shortages, this paper develops a floater algorithm, a fast solution method for reservoir dispatching model, based on the characteristic that the water consumption rate of power generation is relatively low at a high operating reservoir stage. This method directly finds the optimal solution by keeping the reservoir stage as high as possible during operation, while satisfying the requirements of guaranteed output and minimum generation discharge. A case study and comparison with the results of discrete differential dynamic programming (DDDP) shows this algorithm has the advantages of achieving high power generation and fast computation (two times as fast as DDDP) for general situations. Thus, our floater algorithm improves the optimization results and solving efficiency for reservoir operation, helping formulate a regulating-power-generation-by-water-supply operation plan and rapid decision-making.