Research On Meteorological Data Processing And Visualization Platform Based On Microservice Architectur

With the rapid development of the digital economy,meteorological data has already been used in many fields such as weather forecasting,agricultural production,aerospace and emergency disaster reduction.At the same time,the type,scale and update frequency of meteorological data also show very typical characteristics of big data.However,most of the meteorological data service platforms at home and abroad adopt the traditional single application architecture or SOA architecture,which is more and more difficult to meet users’ needs for large data volume,high concurrency and high availability.On the other hand,how to give full play to the stock advantage of meteorological data in the platform and dig out the laws contained in the massive meteorological data is also a huge challenge.In view of the above problems,this thesis designs and implements a meteorological data service platform based on micro-service architecture.The main work is as follows:(1)The platform is divided into multiple service modules such as data sharing,data visualization,and data forecasting,and a series of core components of Spring Cloud Alibaba are used to complete the service governance.Completely decoupled between different services,and can adapt to the rapid growth of business functions through dynamic expansion of services,truly realizing agile development.(2)The overall design goals and development criteria of the platform are analyzed,and the overall framework and specific functional modules of the platform are designed in detail according to user needs.In order to reduce the occurrence of single point of failure,the architecture of cache and database adopts sentinel mode and master-slave mode respectively.(3)The analysis,storage,archiving and sharing of multi-source meteorological data are completed by using Java language,POI function library and other technologies.Using Python script,D3.js and other technologies to visualize multi-source meteorological data.(4)A weather data forecasting model DACNN-RNN(Dual-stage Attention and Convolutional Neural Network based Recurrent Neural Network)based on an improved two-stage attention mechanism is proposed.The model is based on an encoder-decoder architecture as a whole,and a target attention mechanism and a convolutional neural network are introduced in the input stage of the encoder to model the correlation between all input features,a temporal attention mechanism is introduced in the input stage of the decoder to better grasp the dependencies of long time series.The model was finally integrated into the platform,providing support for meteorological data forecasting services.This thesis conducts a comparative experiment based on real air quality related data in Nanjing.The experimental results show that the DACNN-RNN model outperforms other comparative models in terms of mean absolute error(MAE),root mean square error(RMSE)and coefficient of determination(R-Square).Finally,the functional test and performance test of the platform are carried out.The test results show that the platform has achieved the expected design goals.It has been deployed and put into trial operation on the intranet of the Institute of Atmospheric Physics,Chinese Academy of Sciences.