摘要
青藏铁路大风监测预警系统采用时间序列法实现沿线风速的短时预测。利用时间序列法对经过1min平均化处理后的非平稳实测风速序列建立ARIMA(11,1,0)模型,进行超前3步预测计算,获取第3 min平均预测风速,预测平均绝对误差为2.237 2 m.s-1。针对时间序列法第3 min平均预测风速精度低的问题,采用提出的滚动式时间序列法修正时间序列法预测计算步骤,重新获取第3 min平均预测风速,预测平均绝对误差减小为1.1670 m.s-1。使用最小二乘法拟合样本每分钟内最大实测风速和该分钟平均风速的相关系数K,通过K为1.142 8修正滚动式时间序列法第3 min平均预测风速,获取滚动时间序列法第3 min最大预测风速,预测平均绝对误差为2.090 4 m.s-1。滚动式时间序列法第3 min平均风速、最大风速的两者预测均满足精度要求。滚动式时间序列法已经在系统中使用。
Time series method is adopted for building the strong wind monitoring and warning system of Qinghai-Tibet Railway to realize the short-term forecast of the wind speed along the line. ARIMA （11, 1, 0） model for the unsteady in situ tested wind speed series, which have been 1 min averagely treated, is established by time series method. Then three-step ahead forecast calculation is made to get the third minute average forecast wind speed with the average absolute error of 2. 237 2 m·s^-1. Due to the low forecast accuracy of the third minute average forecast wind speed by time series method, an optimized algorithm named rolling time series method is proposed to modify the modeling steps of time series method. The third minute average forecast wind speed is recalculated, and the average absolute error is significantly reduced to 1. 167 0 m·s^-1. The correlation coefficient K between the maximum wind speed per minute and average wind speed per minute is fitted via least square method. By using K as 1. 142 8, the third minute average forecast wind speed is revised by rolling time series method to acquire the third minute maximum forecast wind speed with average absolute error of 2. 090 4 m·s^-1. The third minute average forecast wind speed and maximum forecast wind speed calculated by roiling time series method both meet the accuracy need of monitoring and warning system. tern. The rolling time series method has been used for the sys
出处
《中国铁道科学》
EI
CAS
CSCD
北大核心
2008年第5期129-133,共5页
China Railway Science
基金
"十一五"国家科技支撑计划重点项目(2006BAC07B03)
铁道部科技研究开发计划项目(2006G040-A)
关键词
青藏铁路
大风监测预警系统
风速预测
时间序列法
滚动式时间序列法
Qinghai-Tibet Railway
Strong wind monitoring and warning system
Wind speed forecast
Time series method
Rolling time series method
作者简介
潘迪夫（1957-），男，广东兴宁人，教授，工学硕士。
