摘要
针对传统模型在河流流量估计方面精度低、自适应能力差、易陷入局部最小值的问题,提出一种改进麻雀搜索算法(improved sparrow search algorithm,ISSA)优化反向传播(back propagation,BP)神经网络的算法,利用雷达采集的河流水位、断面积、流速对河流流量进行估计。针对麻雀搜索算法(sparrow search algorithm,SSA)易陷入局部极值问题进行改进,通过莱维飞行对SSA初始个体优化,从而提高种群多样性;引入捕获繁衍机制至麻雀发现者中,以便麻雀个体能够最大程度搜索且提高其跳出局部最优的能力;使用基于适应度的差分变异算法对麻雀加入者进行优胜劣汰的选择,以提高流量估计的准确性和稳定性。使用云南河边水文站2022年6月和7月两个月的数据,搭建了ISSA-BP、SSA-BP和BP神经网络进行对比。实验结果表明,ISSA-BP较SSA-BP和BP平均绝对误差(mean absolute error,MAE)分别降低了25.5%和40.2%,证实了该算法在多特征值河流流量估计方面具有较好的可行性和性能。
To address the problem of low accuracy,poor adaptive ability,and tendency to fall into local minima in traditional models for river flow estimation,an improved sparrow search algorithm(ISSA)optimized back propagation(BP)neural network was proposed to estimate river flow based on radar collected river water level,cross-sectional area,and flow velocity data.To improve the sparrow search algorithm(SSA),which is prone to local extremum,the initial individuals of SSA were optimized by Levy flight to improve the population diversity.The capture and reproduction mechanism was introduced into the sparrow discoverers so that the sparrow individuals can search to the maximum extent and improve their ability to jump out of the local optimum.The differential mutation algorithm based on fitness was used to select the sparrow participants to improve the accuracy and stability of river flow estimation.Using the data of Yunnan Riverside Hydrological Station in June and July 2022,ISSA-BP,SSA-BP and traditional BP neural networks were built for comparison.The experimental results show that the mean absolute error(MAE)of ISSA-BP is 25.5%and 40.2%lower than that of SSA-BP and BP,respectively,which confirms that the proposed model has good feasibility and performance in multi eigenvalue river flow estimation.
作者
许荆
张文鑫
杨鸿波
李健
于然
XU Jing;ZHANG Wenxin;YANG Hongbo;LI Jian;YU Ran(School of Automation,Beijing Information Science&Technology University,Beijing 100192,China)
出处
《北京信息科技大学学报（自然科学版）》
2023年第3期59-65,共7页
Journal of Beijing Information Science and Technology University
关键词
神经网络
麻雀搜索算法
莱维飞行
差分变异
捕获繁衍机制
河流流量
估计
neural network
sparrow search algorithm
Levy flight
differential mutation
capture and reproduction mechanism
river flow
estimation
