摘要
针对验潮站潮位预报的需求,提出一种基于分群策略的粒子群优化神经网络(SSPSO-BP)的预报方法。该方法通过建立多个不同功能且具有交流能力的智能粒子群,经SSPSO和BP的两次优化,构建潮高预报模型。实验研究表明,SSPSO-BP模型在Oga站的潮位资料上高、低潮位间的时刻基本保持一致,高潮时刻最大潮高差为7.37 cm,低潮时刻最大潮高差为4.21 cm,该模型比标准BP神经网络及PSO优化神经网络在准确度和精度上有了很大的提高,其平均绝对误差、均方误差相较于BP神经网络分别提升了16.2%、79.2%,相较于PSO-BP神经网络提升了13.9%、79.6%。
Aiming at the demand of tide level prediction of tide gauge stations, a prediction method based on swarm strategy particle swarm optimization-back propagation(SSPSO-BP) based on clustering strategy was proposed. In this method, several intelligent particle swarm optimization with different functions and communication ability were established, and the tide height prediction model was constructed through the twice optimization of SSPSO and BP. The experimental research showed that the SSPSO-BP model basically kept the same time between high and low tide levels in the tide level data of OGA station. The maximum tidal height difference at high tide was 7.37 cm, and the maximum tidal height difference at low tide was 4.21 cm. The accuracy and precision of this model were greatly improved compared with standard BP neural network and PSO optimization neural network, and its average absolute error and mean square error were increased by 16.2% and 79.2% respectively compared with BP neural network, Compared with PSO-BP neural network, it increased by 13.9% and 79.6%.
作者
张宇
贺小星
孙喜文
ZHANG Yu;HE Xiaoxing;SUN Xiwen(School of Civil and Mapping Engineering,Jiangxi Institute of Technology,Ganzhou Jiangxi 341000,China;East China University of Technology,Nanchang Jiangxi 330013,China)
出处
《北京测绘》
2023年第1期131-136,共6页
Beijing Surveying and Mapping
基金
国家自然科学基金(42104023,41904002)
江西理工大学高层次人才科研启动项目(205200100564,205200100588)
江西理工大学大学生创新创业训练资助项目(202210407032)。
关键词
BP神经网络
分群策略
变异算子
SSPSO-BP模型
潮高预测
back propagation(BP)neural network
Clustering strategy
mutation operator
swarm strategy particle swarm optimization-back propagation(SSPSO-BP)model
tide height prediction
