Research Of Multi-model Integrated Weather Forecast Based On Cloud Platform

With the development of numerical forecasting technology,weather forecasting quality has been significantly improved,but the current numerical forecasting technology has gradually matured,accordingly,the numerical forecasting technology itself can not make the forecasting to be improved greatly.Multi-model integrated weather forecast is a kind of statistical forecasting technology thus generated,it can integrate the forecasting advantages of each forecasting mode and further improve the forecasting quality.Applying genetic algorithm to multi-model integrated weather forecast can improve the forecasting quality and the stability of the forecasting results,but it has a high computational cost.The purpose of this study is to combine the cloud computing technology and genetic algorithms,establish a multi-model integrated weather forecasting model based on cloud platform,and develop the corresponding multi-model integrated weather forecasting system which can improve forecasting quality at low computational cost and provide accurate and uniform reference results for forecasters.Multi-model integrated weather forecasting system based on cloud platform selects the members of integrated forecasting model firstly according to the characteristics of the numerical forecasting products in Tianjin Meteorological Station,and normalizes and revises forecasting data of each model and training set,then gains the weight coefficient of each numerical forecasting model by the Spark which is a distributed computing platform based on the memory and CPGA(Coarse-grained Parallel Genetic Algorithm),finally integrates the forecasting results using the weight distribution method in order to achieve more accurate and fast weather forecast combining the advantages of each model.This paper analyzes the research status of multi-model integrated weather forecast firstly,describes the advantages and prospects of multi-mode integrated weather forecast using cloud platform and genetic algorithm,then builds a new multi-model integrated forecasting model based on SCPGA(CPGA based on Spark)and weight distribution method,and then develops a multi-model integrated forecasting system using the model.Integrated forecasting system developed has the functions of automatic or manual integration,classified query and cluster monitoring,it can achieve 24-hour temperature forecasts and 48 hours precipitation forecast of 232 sites in Tianjin.The results show that the integrated weather forecast based on SCPGA is not only superior to each model and the traditional genetic algorithm,but also can obtain higher efficiency.It can provide accurate and reliable reference for the forecasters,and has a good prospect.