Method for predicting typhoon path

Abstract

The invention relates to the technical field of typhoon path prediction, in particular to a typhoon path prediction method. Comprising a preliminary experiment, and the preliminary experiment comprises the following steps: firstly, four models are established, and four sequences with different time steps are used for prediction: model 1: a future (four steps) 24h path is predicted in a complete sequence; according to the second model, the four-step sliding window width is set, and the typhoon position in the next 24 hours is predicted; models III and IV and the model 2 use sliding window width to predict, the window width of the model 3 is changed into 6, and the window width of the model 4 is changed into 8. According to the method for predicting the typhoon path, two-dimensional typhoon path data provided by a CMA tropical cyclone optimal path data set can be used for establishing future typhoon path prediction of an LSTM network, and prediction is carried out in a complete sequence prediction mode.

Description

technical field [0001] The invention relates to the technical field of typhoon track prediction, in particular to a method for predicting a typhoon track. Background technique [0002] my country faces the Pacific Ocean to the east, and is near the Northwest Pacific Ocean, where typhoons occur more frequently in the world. figure 1 The annual number of tropical cyclones in the Pacific Ocean recorded by CMA from 1949 to 2018 is given. Tropical cyclones are generated every year, and the number of generation is almost more than ten times. According to statistics, the tropical cyclones that landed in my country during the 20 years from 1998 to 2017 accounted for 31% of the total number of tropical cyclones generated in the Northwest Pacific Ocean. Typhoons have become one of the main natural disasters affecting our country. The eastern coastal areas are also relatively developed political, economic and cultural areas in our country. They have been affected by typhoons for many ...

AI Technical Summary

Problems solved by technology

[0003] A lot of research has been done on typhoon track forecast at home and abroad, and some typhoon track forecast methods have been established, mainly including subjective experience forecast, objective statistical forecast, and dynamic forecast of numerical model. Subjective forecast has basically withdrawn from the historical stage, and objective statistical forecast is based on a large amount of historical data. , look for physical factors that have a strong correlation with the forecast object, use probability statistics methods, establish the relationship between the physical factor and the forecast object, find out the relevant laws, implement the forecast, and the dynamic forecast of the numerical model is based on atmospheric dynamics, thermodynamics, fluid mechanics As a basis, use relevant principles to describe atmospheric motion, form a closed equation system, set certain initial values ​​and boundary conditions, and use computers to perform numerical solutions to make quantitative forecasts. These methods have good results in practical applications, but they require rich experience. Only people with relevant skills and knowledge can use it. Many application scenarios lack corresponding conditions. After the emergence of artificial intelligence technology, artificial intelligence technology has been applied to many scientific fields, and the field of meteorology is one of them. A lot of research has been done on the application of technology to typhoon forecasting. The research of Li Moumou and Deng Moumou pointed out that if the early factors with specific physical meanings are selected and have a good correlation with the typhoons that appear later, the BP neural network is used to compare the standard It is possible to establish a typhoon forecast model by training the data. Zhou Moumou abstracted and simplified the weather data, and used the BP neural network to predict the direction of typhoon movement. The typhoon movement category and the actual movement category (westward, northward, Northwest shift) generalization rate reaches 97%, but the prediction of difficult typhoon track is not very accurate, which needs to be studied. Huang et al. use BP neural network as the basic model, and extract the features of the predictor group and the variance contribution technology of the predictor Combined methods are used to mine the information of weather forecast factor data. The model research results show that the artificial neural network has strong adaptive learning and nonlinear mapping capabilities, which can well reflect the nonlinear characteristics of the typhoon path, and also point out the abnormality. The path will cause trouble to the forecast. Shao Moumou, Fu Moumou and others conducted path prediction research on the tropical cyclones that appeared in the coast of China based on the BP neural network. Using the typhoon data 24 hours before the forecast, they can use the BP neural network to forecast the next 24 hours. , 48 hours, 72 hours, the location of the typhoon center, and the accuracy meets the requirements. Fang Moumou, Ji Moumou, etc. combined artificial intelligence technology and expert knowledge, established an expert system, and realized the intelligent automatic forecasting of typhoons. A researcher and others will The combination of the support vector machine method and the dimension reduction method enables it to have the ability to identify nonlinear problems and improve the prediction accuracy of typhoon track forecast

Images