| 陈娟,张璐,孙飞飞,邓如霞,钟平安.基于云计算的水库群实时防洪多目标风险调度模型[J].水利学报,2025,56(7):885-897 |
| 基于云计算的水库群实时防洪多目标风险调度模型 |
| A multi-objective risk operation model for real-time flood control of reservoir groups based on cloud computing |
| 投稿时间：2024-11-14 修订日期：2025-06-04 |
| DOI：10.13243/j.cnki.slxb.20240738 |
| 中文关键词: 水库群 实时防洪调度 不确定性 云计算 分布式集群 |
| 英文关键词: multi-reservoir system real-time flood control operation uncertainty cloud computing distributed cluster |
| 基金项目:国家自然科学基金项目(52479011,51909062); 国家重点研发计划项目(2022YFC3202801) |
|
| 摘要点击次数: 67 |
| 全文下载次数: 39 |
| 中文摘要: |
| 水库群实时防洪调度是受诸多因素影响的不确定性调度，开展水库群实时防洪多目标风险调度研究对保障流域防洪安全具有重要意义。本文考虑水文预报误差的不确定性，以各水库最高水位最低、下游防洪断面最大过水流量最小为目标函数，建立水库群实时防洪多目标风险调度模型；针对复杂水库群“维数灾”问题，基于云计算采用多目标金鹰优化算法（MOGEO）求解水库群实时防洪多目标风险调度模型，从智能优化算法、风险因子模拟、并行计算、云计算四个角度优化模型求解时间，满足实时防洪调度高时效性的需求；再根据非劣解集的空间分布，提出改进的点云体素下采样法提取最优调度方案。以淮河中上游史灌河流域为背景开展实例研究，结果表明：MOGEO在防洪调度方面的适应能力较强，模型求解时间由第三代非支配排序遗传算法（NSGA-Ⅲ）的1542 s缩短至830 s，快速收敛至Pareto前沿；基于改进拉丁超立方的风险因子模拟在确保抽样精度可靠的情况下，缩减了2/3的计算时间；采用云分布式集群的模型求解时间为113 s，为云服务器单机12核并行计算时间的1/6、串行计算时间的1/30，大幅度提高了模型求解效率。 |
| 英文摘要: |
| Real-time flood control operation of a multi-reservoir system is a risky operation influenced by many uncertain factors.This paper considers the uncertainty of hydrological forecast errors and establishes a multiobjective risk operation model of a multi-reservoir system with the objectives of minimizing the highest reservoir water level and minimizing the maximum discharge in the downstream section.Using the multi-objective golden eagle optimization algorithm （MOGEO）,the "curse of dimensionality" problem is addressed from four perspectives:intelligent optimization algorithms,risk factor simulation,parallel computing,and cloud computing.And then,an improved point cloud voxel down sampling method is proposed to extract the optimal operation scheme according to the spatial distribution of the set of non-inferior solutions.The Shiguanhe river basin is selected for case study.The results show that MOGEO reduces the calculation time of the model from 1,542 s of Non-Dominated Sorting Genetic Algorithm Ⅲ to 830 s.The improved Latin hypercube sampling method can ensure the sampling accuracy while reducing the calculation time by 2/3.The calculation time using the cloud distributed cluster is 113 s,which is 1/6 of that on a single cloud server with 12-core parallel processing and 1/30 of the time for serial computation. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |