Research Of Typhoon Track Forecast Optimization Method Based On Mixed Model Ensemble Forecast

Typhoon is one of the most serious meteorological disasters in the world.It directly or indirectly affects and threatens the personal safety and property safety of our people.Therefore,it is of great significance to improve the forecast level of typhoon track and integrate forecasting methods with operational application for improving the mechanism of typhoon prevention and control and enhancing the early warning ability of meteorological disaster in China.At present,the methods of typhoon track prediction include meteorological method,dynamic method and statistical method,among which ensemble prediction is a dynamic stochastic prediction method which can effectively improve the accuracy.Aiming at optimizing the inaccuracy method of typhoon track prediction and the insufficient research of ensemble prediction,this paper explores the multi-element extensible fusion model of neural network based on the research of ensemble prediction idea,constructs an intelligent convergence method for typhoon track prediction,which is highly available,fine-grained and adapts to the timeliness requirements of typhoon prediction.Taking intelligent services of the results of typhoon observation and prediction as a driven demand,a multi-dimensional and multi-scale information service platform for typhoon ensemble forecasting optimization is designed,and an integrated and efficient service system of"model-method-application" is realized.The main contents of this paper are as follows:(1)Aiming at improving the accuracy and timeliness of Typhoon Forecasting and optimizing ensemble forecasting method adequately,a multi-mode collaborative optimization method is designed based on site optimization strategy and factor optimization strategy,and the multi-mode collaborative optimization based on optimization strategy is realized.By updating the parameters and improving the structure of the neural network model,an optimization training model of the neural network is designed.The convergence method of Intelligent Forecasting for typhoon track is formed,and the optimization method system of ensemble forecasting is innovated.(2)The multi-model collaborative optimization model of ensemble forecasting,the partial value optimization model of ensemble forecasting and the tendency optimization model of ensemble forecasting are proposed.Based on the typhoon data of the Northwest Pacific and South China Sea from 2016 to 2018,the three heterogeneous models are tested and explored.By comparing and analyzing the forecasting results under different evaluation scales,the multi-model collaborative optimization model of ensemble forecasting,the partial value optimization model of ensemble forecasting and the tendency optimization model of ensemble forecasting are synthetically compared and analyzed,which verifies the high availability of the convergence method of Intelligent Forecasting of typhoon track.(3)Taking typhoon Mixed model ensemble forecasting data as information support and intelligent service of typhoon observation and forecasting results as demand-driven,an information service prototype system for optimizing typhoon Mixed model ensemble forecasting is constructed,which effectively serves typhoon monitoring and forecasting,and provides omni-directional and highly aggregated decision support for relevant departments in ocean,meteorology and information fields.