Design And Implementation Of Rainstorm Forecasting Analysis System Based On Bayesian Network

The rainstorm process may cause floods,floods and geological disasters in the city,which will affect the lives and property of the people.To this end,the forecast and service of the rainstorm weather process is an important part of the disaster prevention and mitigation work.According to the rainstorm forecast,the forecasting expert constructs a local storm forecast concept model based on historical experience methods and threshold indicators.After combining the real-time and numerical model data analysis,a more reliable forecast conclusion can be obtained.Therefore,on the basis of inheriting the valuable experience of forecasting experts,it is necessary to use modern artificial intelligence technology to reconstruct and realize more intelligent storm forecasting analysis tools.The main work of this paper is to research and implement the "Bayesian network-based rainstorm forecast analysis platform",analyze the existing storm forecast concept model in detail,and design a set of storm forecast analysis tools based on the existing weather forecast business requirements.Support reasoning and analysis.The system adopts the C/S architecture,and the construction and application of the predictive analysis model are deployed on the server side and the client side respectively.The server side adopts the Python technology system to construct the data collection of the Zigong local rainstorm forecast analysis.The data set is from May to September in 2008-2017.The data set includes the 24-hour rainfall in Zigong City,the live elements of the station and the European Center.Analyze the data set ERA-Interim.The intuitive Bayesian network model was constructed using the mature Ge NIe platform to discretize continuous random variables.The prior probability of random variables depends mainly on the prediction expert experience and historical data statistics.A good prior probability can make up for the shortcomings of the smaller data set size.The client adopts the.NET environment and completes the storm forecast analysis tool based on the secondary development of the MICAPS4 framework.When the client runs,the CIMISS data interface is used to obtain data such as ground maps,high-altitude maps,and numerical forecast model products.The SMILE engine is used to load the storm forecast analysis model built on the server side to provide forecasters with rainstorm prediction analysis conclusions.The live forecasting conclusions help.After the local rainstorm test on August 21,2018,20:00 to 20:00,the Bayesian network model based on causality,EC fine grid numerical prediction model data and local live observation Data,the probability of "storm" in the forecast object is the highest,and it is consistent with the actual situation.Through the rainstorm forecast analysis tool,it has good practical significance to help forecasters better carry out rainstorm forecast service work.