基于BIRCH聚类和递归神经网络的高铁强风预警算法  

An algorithm for strong wind speed warning of high-speed train based on BIRCH clustering and recurrent neural network

作　　者：樊仲欣[1] FAN Zhong-xin(Experimental Teaching Center of Atmospheric Science and Environmental Meteorology,Nanjing University of Information Science&Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学大气科学与环境气象实验教学中心,南京210044

出　　处：《信息技术》2024年第10期162-167,174,共7页Information Technology

摘　　要：针对高铁临近风速预测需要克服数据的无周期性规律以及随机性较强的问题,构建了一种基于BIRCH聚类和LSTM递归神经网络算法的临近风速预测预警系统。该系统先做历史数据的交叉验证,然后用BIRCH进行在线增量聚类,最后根据聚类结果选取最接近当前预测时间序列的数据做LSTM的滚动训练并进行预测后得出预报预警结果,因此具有无需依赖数值预报产品以及随机数据适应性强的特点。实验证明,该系统的两种算法同时并行化在线运转,运行效率较高,预测效果较好,是解决强风预警问题的一种新方法。To solve the problem that non periodicity and randomness of data in predicting the near wind speed of high-speed trains,a near wind speed prediction and warning system based on BIRCH clustering and LSTM algorithm is constructed.The system firstly performs cross validation of historical data,then uses BIRCH for online incremental clustering.Finally,based on the clustering results,the data closest to the current predicted time series is selected for LSTM rolling training and predicted to obtain the prediction and warning results.Therefore,the method has characteristics of not relying on numerical prediction products and adapting to random data.Experiments show that two algorithms of system are run in parallel,with high efficiency and good prediction results.Thus,it is a new way to solve the problem of strong wind speed prediction.

关 键 词：高速铁路 风速 预测预警 聚类 递归神经网络 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]