PA-DDS算法在水库多目标优化调度中的应用

Pareto存档动态维度搜索（Pareto-Archived Dynamically Dimensioned Search，PA-DDS）算法是一种求解多目标问题的随机搜索启发式算法。本文将PA-DDS算法引入考虑供水和发电的多目标优化模型优化水库调度图，与非支配排序遗传算法（NSGA-II）和多目标粒子群算法（MOPSO）对比了收敛性，并在求得非劣解分布的均匀性和与理论Pareto前端的相似性方面与NSGA-II进行比较，分析该算法求解水库多目标调度问题的优化性能，对比分析不同目标下的优化调度图。结果表明：PA-DDS算法能够得到更高质量的非劣解集，优化调度图与原设计调度图相比，能更有效协调供水和发电的矛盾，在小幅降低（0.96%）发电量的前提下显著提高（8.07%）水库供水量，平均每年增加经济效益0.55亿元。

PA-DDS algorithm for multi-objective reservoir operation

PA-DDS (Pareto-Archived Dynamically Dimensioned Search) is one of the multi-objective evolutionary algorithms designed to obtain multiple solutions, which offer different trade-off of the problem objectives. The algorithm was introduced into reservoir operating rule curves optimization for the targets of water supply and power generation, and compared with NSGA-II (Non-dominated Sorting Genetic Algorithm II) and MOPSO (Multi-Objective Particle Swarm Optimization) in convergence. Furthermore, the performance of PA-DDS and NSGA-II are also analyzed in distribution and approximation between whole Pareto fronts of non-dominated solutions. The capacity of PA-DDS solving multi-objective reservoir operating problems and optimal rule curves were analyzed. It is shown that PA-DDS algorithm performs better than NSGA-II algorithm in getting the non-dominated solutions. The optimized rule curves can effectively mitigate the conflict between water supply and power generation by dramatically increasing (8.07%) the water supply yield with slight decrease (0.96%) in power generation, and can improve the economic benefit by 55 million yuan annually.