PDF(1546 KB)
PDF(1546 KB)
PDF(1546 KB)
基于自适应混沌精英变异差分进化算法的中长期水资源优化调度
Medium- and Long-term Optimal Scheduling of Water Resources Based on Self-adaptive Chaotic Elite Mutation Differential Evolution Algorithm
中长期水资源优化调度问题是一类具有非线性、多阶段、高维度和多重约束特性的复杂优化问题。针对经典智能算法在求解此类问题时容易陷入局部最优或者收敛效率较低等问题,应用混沌搜索策略增强算法的探索能力,同时改进传统算法的变异方式,向精英个体学习以提升收敛速度,提出自适应混沌精英变异差分进化(ACEDE)算法。将所提出的算法应用于珠江三角洲水资源配置工程中长期调度进行实例研究,并与经典智能算法进行对比分析。结果表明:①ACEDE算法在全局探索能力、收敛精度与速度上实现了全面提升,并且表现出良好的适应性。相较于传统差分进化(DE)算法,2030年水平年6月份和2040年水平年6月份调度中ACEDE算法所计算的电费成本分别节省了74.23万元和23.55万元,降低了6.68%和1.52%。②在珠江三角洲水资源配置工程中长期调度中,充分利用调蓄水库库容满足高分水量需求,同时放缓月末补水充库过程,能够有效控制泵站的平稳运行,达到降低电费成本的目的。
The medium- and long-term optimal scheduling of water resources is a complex optimization problem characterized by non-linearity, multi-stage high dimensionality, and multi-constraints. To address the local optimization and low convergence efficiency of classical intelligent algorithms, this paper introduce a novel algorithm named the Adaptive Chaotic Elite Mutation Differential Evolution (ACEDE) algorithm. The algorithm leverages a chaotic search strategy to enhance the algorithm’s exploration capabilities while revising the traditional mutation approach to learn from elite individuals, thereby accelerating convergence. The proposed algorithm is applied to the medium- and long-term scheduling of the Pearl River Delta Water Resources Allocation Project (PRD WRAP) as a case study and is compared with classical intelligent algorithms. Results indicate that: 1) The ACEDE algorithm improves significantly in global exploration capabilities and convergence accuracy and speed, demonstrating good adaptability. For the June 2030 and June 2040 level year dispatches, respectively, the ACEDE algorithm, saves ￥742 300 and ￥235 500 in electricity costs compared to the traditional DE algorithm, reducing the cost of electricity by 6.68% and 1.52%. 2) For the medium- and long-term optimal scheduling of PRD WRAP, fully utilizing the reservoir storage capacity to meet high water demand while slowing down the replenishment at the end of month could effectively control the smooth operation of the pumping station and minimize electricity costs.
水资源优化调度 / 差分进化算法 / 混沌映射 / 精英变异 / 珠江三角洲水资源配置工程
optimization of water resources dispatching / differential evolution algorithm / chaotic mapping / elite mutation / Pearl River Delta Water Resources Allocation Project
| [1] |
谢轶. 浅谈我国水资源分布和生态红线[J]. 现代农业研究, 2020(3):143-144.
(
|
| [2] |
任海康, 王晓云. 大型引调水工程向某区域调水的必要性研究[J]. 海河水利, 2023(9):1-4.
(
|
| [3] |
金鹏宇, 王超, 孙嘉辉, 等. 胶东调水工程年度水量调度方案研究[J]. 人民长江, 2023, 54(8): 131-138.
(
|
| [4] |
雷晓辉, 张利娜, 纪毅, 等. 引江济淮工程年水量调度模型研究[J]. 人民长江, 2021, 52(5): 1-7.
(
|
| [5] |
姚柳杉, 赵雪花, 武雯昱, 等. 猎豹优化算法在水资源优化配置中的应用[J]. 中国农村水利水电, 2024(1):135-141,149.
水资源优化配置能够解决区域水资源短缺及不合理利用问题，近年来优化算法成为获得水资源优化配置结果的重要工具。为有效缓解现有优化配置算法中存在的收敛缓慢、局部优化和早熟等问题，介绍了一种新型的元启发式算法：猎豹优化算法（the Cheetah Optimizer，CO），利用该算法收敛速度快、鲁棒性强、寻优能力强等优点，将其运用至水资源优化配置问题中。以太原市为研究区域，建立了以经济效益、社会效益和生态效益为目标的水资源优化配置模型，采用CO求解该模型，并与遗传算法（Genetic Algorithm，GA）结果进行对比分析。结果表明：CO的污染物排放总量在规划年（2030）控制目标内，区域用水满意度达到0.98，能够满足基本用水和发展用水需求；与现状年（2020）相比，生活用水占比下降22.58%，三次产业用水占比分别上升10.53%、25.81%、12.5%，地表水和地下水供水比例分别降低21.70%、11.14%，回用水和引黄水供水比例分别提高了5.43%、27.40%，符合太原市水资源配置思路，验证了CO的有效性；与GA相比，CO的区域满意度、综合用水价值和配水公平性分别提高了7.10%、3.57%和56.52%，污染物排放量和缺水率分别降低了3.11%和3.85%，同时有更好的协调性，验证了CO的优越性。研究成果表明用CO求解水资源优化配置模型合理可行，为水资源优化配置模型求解提供了新方法。
(
|
| [6] |
|
| [7] |
刘孟桦, 何中政, 盛卫荣, 等. 基于改进差分进化算法的万安水库防洪优化调度[J]. 水电能源科学, 2023, 41(6): 54-58.
(
|
| [8] |
陈立刚, 陈芳. 基于遗传算法的水库调度策略研究[J]. 水利科技与经济, 2022, 28(10): 36-39.
(
|
| [9] |
许银山, 梅亚东, 魏婧, 等. 梯级水库群长期优化调度的逐次优化-遗传算法[J]. 中国农村水利水电, 2008(9):25-27.
(
|
| [10] |
孙桂凯, 石锐, 刘思怡, 等. 基于长期与中长期嵌套的水库优化调度[J]. 长江科学院院报, 2022, 39(8): 23-28.
为解决水库优化调度过程中时间尺度不同而导致的水库综合效益不高的问题,以防洪与兴利为目标,将水库长期优化调度与中长期优化调度进行嵌套,建立水库多目标优化调度嵌套模型。其中,多目标问题通过约束法将防洪目标转换为硬性约束后,再转化为求解发电量最大值的单目标问题。算法方面,综合对比后确定长期优化调度采用动态规划算法求解、中长期优化调度采用遗传算法求解。为验证嵌套模型的优化效果,以澄碧河水库为例,进行水库优化模拟调度。结果表明:在满足防洪目标的前提下,嵌套模型优化调度方案相较于长期优化调度方案和实际调度方案效益均更高,验证了嵌套模型在水库优化调度问题中的优越性。
(
The comprehensive benefit of reservoir is not high due to different time scales in the process of optimal scheduling.In view of this,a nested model of multi-objective optimal scheduling of reservoir is built with both flood control and operational benefit as objectives by nesting the long-term and medium-to-long-term optimal scheduling of reservoir.The multi-objective problem is transformed into a single objective problem of solving the maximum power generation after the flood control objective is transformed into a hard constraint by the constraint method.In terms of algorithm,the long-term optimal scheduling is solved by dynamic programming algorithm,and the medium-to-long-term optimal scheduling is solved by genetic algorithm.The optimization effect of the nested model is verified with Chengbihe reservoir as a case study.The results demonstrate that under the premise of meeting the flood control objectives,the nested model scheme is more effective than the long-term scheme and the actual operation scheme,which verifies the superiority of the nested model.
|
| [11] |
王攀, 方国华, 郭玉雪, 等. 水资源优化调度的改进量子遗传算法研究[J]. 三峡大学学报(自然科学版), 2016, 38(5): 7-13.
(
|
| [12] |
王奇. 基于混沌折射差分进化算法的梯级水库群蓄水期优化调度研究[D]. 武汉: 华中科技大学, 2022.
(
|
| [13] |
原博, 王文倬, 李武璟, 等. 基于混沌进化算法的梯级水电群实时优化调度方法[J]. 微型电脑应用, 2022, 38(7): 182-184, 190.
(
|
| [14] |
|
| [15] |
汪涛, 徐杨, 刘亚新, 等. 基于多种群引力粒子群算法的金沙江下游—三峡梯级水库群优化调度[J]. 长江科学院院报, 2023, 40(12): 30-36, 58.
金沙江下游—三峡已形成六库联合调度格局,调度维数增多,约束庞杂交织,优化目标多样,给调度方案制定带来极大困难。针对传统粒子群算法求解调度模型寻优能力不足的难题,提出多种群引力粒子群算法,建立优化调度模型并应用改进算法求解。算法测试和应用结果表明,多种群引力粒子群算法寻优性能更加先进,更适用于求解梯级水库优化调度问题。实例表明,上游龙头电站通过减少自身发电量可以使下游电站和梯级发电量增加。
(
The construction and operation of the Three Gorges Project (TGP) have significantly altered the hydrological regime of the middle and lower reaches of the Yangtze River, thereby impacting irrigation water supply. This study focuses on analyzing the effects of TGP operation on the safety of irrigation water supply in the Poyang Lake areas, specifically considering water quantity and fulfillment of water supply. To achieve this, the evolution of water resources and the situation of water supply before and after TGP operation were investigated. Following the operation of TGP, the Poyang Lake witnessed a decreasing trend in available water supply, with the most significant reductions occurring in September and October during impoundment periods. Moreover, the overall water level in the lake area has dropped by 1.1-2.5 meters. Although the total water quantity can still meet irrigation and water supply demands, there is a noticeable decline in the fulfillment of both irrigation and water supply requirements in urban and rural regions, leading to a significant increase in total water shortage. The fulfillment of secured rural and urban water supply and irrigation water supply requirements has dropped from 99.1% and 83.6% before TGP operation to 87.5% and 42.7% after operation, respectively. Simultaneously, water shortage for rural and urban water supply and irrigation water supply has increased from 0.78 and 70.32 to 93.85 and 186.11 million cubic meters, respectively. However, it is worth mentioning that early or delayed impoundment of TGP can slightly mitigate the project’s influence on water supply and irrigation during impoundment, thereby increasing the water supply capacity by 0.2% and 0.3% respectively.
|
| [16] |
林秀丽, 李均利, 聂君凤. 参数自适应的精英变异差分进化算法[J]. 小型微型计算机系统, 2023, 44(8): 1693-1706.
针对经典差分进化算法（DE）的优化性能容易受到变异策略和控制参数影响的问题，提出了一种参数自适应的精英变异差分进化算法（A parameter Adaptive Elite Mutation Eifferential Evolution algorithm，AMEDE）.首先，提出一种精英变异策略的方法，其目的是为了方便获取优秀个体信息；其次，引入新的控制参数，使得算法可以在更大的搜索空间进行搜索；最后，利用自适应参数学习方法，为种群中的每个个体赋予不同的控制参数值，并根据种群多样性和精英个体的信息动态更新个体的参数，使算法避免过早的收敛并提高算法的收敛精度.对本文提出的AMEDE算法与其他6种改进差分进化算法（DE，CoDE，JaDE，JDE，SaDE，GPDE）在16个基准测试函数上进行了三组对比实验.实验结果表明，AMEDE算法在高维函数和低维函数上都具有搜索精度高、收敛速度快和鲁棒性强等优点.
(
Aiming at the problem that the optimization performance of classical differential evolution algorithm（DE）is easily affected by mutation strategy and control parameters，a parameter adaptive elite mutation differential evolution algorithm（AMEDE）is proposed.Firstly，a method of elite mutation strategy is proposed，which aims to facilitate the acquisition of excellent individual information；Secondly，new control parameters are introduced to make the algorithm search in a larger search space；Finally，using the adaptive parameter learning method，each individual in the population is given different control parameter values，and the individual parameters are dynamically updated according to the population diversity and the information of elite individuals，so as to avoid premature convergence and improve the convergence accuracy of the algorithm.The AMEDE algorithm proposed in this paper is compared with six other improved differential evolution algorithms（DE，CoDE，JaDE，JDE，SaDE，GPDE）on 16 benchmark functions.Experimental results show that AMEDE algorithm has the advantages of high search accuracy，fast convergence speed and strong robustness in both highdimensional and lowdimensional functions.
|
| [17] |
邹强, 鲁军, 周超, 等. 基于并行混合差分进化算法的梯级水库群优化调度研究[J]. 水力发电学报, 2017, 36(6):57-68.
(
|
| [18] |
|
| [19] |
|
| [20] |
|
| [21] |
|
| [22] |
|
| [23] |
|
| [24] |
|
/
| 〈 |
|
〉 |
