改进WNN模型在GNSS高程拟合中的应用
Application of Improved WNN Model in GNSS Elevation Fitting
郑明丹 1于长龙 1王飞文 1孙五斌1
作者信息
- 1. 浙江省测绘科学技术研究院,浙江 杭州 311100
- 折叠
摘要
针对小波神经网络模型在GNSS高程拟合中存在的缺陷,本文提出了一种遗传模拟退火(GSA)算法优化的小波神经网络(WNN)组合模型,将模拟退火(SA)算法加入遗传算法(GA)种群更新中,保证了种群的多样性.该组合模型充分利用GSA的全局搜索能力,对WNN模型参数进行自动寻优,实现WNN模型全局最优解的获取与GNN高程拟合精度的提升.以某地区D级GNSS水准网实测数据进行实验,结果表明,相较于传统的BP神经网络模型与WNN模型,本文提出的GSA-WNN模型的GNSS高程拟合精度与稳定性更高,更适用于实际工程实践场景.
Abstract
Aiming at the defects of wavelet neural network model in GNSS(Global Navigation Satellite System)elevation fitting,this paper proposes a wavelet neural network(WNN)combination model optimized by genetic simulated annealing(GSA)algorithm and adds simulated annealing(SA)algorithm to genetic algorithm(GA)population updating,then the diversity of the population is guar-anteed.The combined model makes full use of GSA's global search ability to automatically optimize WNN model parameters,so as to achieve the acquisition of global optimal solution of WNN model and the improvement of GNN elevation fitting accuracy.The experi-ment is carried out with the measured data of D-order GNSS leveling network in a certain area.The results show that the GSA-WNN model proposed in this paper has higher accuracy and stability of GNSS elevation fitting than the traditional BP neural network model and WNN model,and is more suitable for practical engineering scenarios.
关键词
高程异常/GNSS高程拟合/遗传算法/模拟退火算法/参数寻优Key words
elevation anomaly/GNSS elevation fitting/genetic algorithm/simulated annealing algorithm/parameter optimization引用本文复制引用
出版年
2024
NETL
NSTL
万方数据