Qinghai-tibet Railway Runs Security System, High Winds Forecast To Optimize The Algorithm And To Promote Applied Research

Strong-wind is one of major meteorological disasters which affect railway transport safety. In recent years, Many railway accidents caused by strong-wind have happened in the world. Make research for evolve -ment rules and real-time short-term forecasting methods of strong-wind can make scientific guidance for railway departments to dispatch and control trains.Qinghai-Tibet Railway is the highest line in the world. Extreme strong-wind weather happens frequently along the line. The year-average strong-wind days are more than 100 days. History maximum wind-speed exceeded 40m/s. Local station wind-speed value is far higher than train over-turned critical wind-speed value. Strong-wind makes a serious threat for trains along this line. So the strong-wind forecasting methods research for Qinghai-Tibet railway has a great project and economic significance.Due to especial air-current and topography of Qinghai-Tibet Plateau, strong-wind along it has uncertainty characteristics. The evolvement of strong-wind is a typical non-stationary random process. Article carries on numerical forecasting for the wind-speed change in local railway region (defined as "typical microphenomenon" by the World Meteorological Organization). Through summarizing the domestic and foreign research results, article has not only proposed an non-stationary modeling method, but also provided an optimized method system based on Time Series Analysis Theory and Kalman Auto-adaptive Filter Theory. The system mainly includes: Kalman-time Series Method (for short as KTSM) and Rolling Time Series Method (for short as RTSM). Experimental results showed: Contrast to the traditional Time Series Method, KTSM can amend wind-speed forecasting-delay phenomena, improve significantly one-step ahead forecasting accuracy; RTSM can attain high-precision of multi-step ahead forecasting. They have good algorithm performances. Based on algorithm results, with Visual Basic.Net2005, SQL Server 2005, Matlab2007 software platform, ADO.Net and ActiveX technology, article has developed Qinghai-Tibet Railway Strong-wind Assisted Modeling System and Modeling & Forecasting Toolbox. Besides, article makes research on how to use KTSM and RTSM for non-stationary random time series, include wind farm wind-speed time series, mechanical failure vibration time series, railway freight traffic volume time series ect. Prediction experimental results have confirmed the stability and validity of KTSM and RTSM, showed that KTSM and RTSM can improve significantly forecasting accuracy for non-stationary random time series, but not increase calculation complexity.