Research On Data Storage And Prediction For Snowmelt Flood Disaster

Ice and snow resources are widely distributed in the northwest region,which bring abundant water resources and ice melting phenomenon.In particular,under the influence of global warming and human activities,the melting process of ice and snow has become more active,and its frequency and intensity have also changed significantly,eventually leading to the occurrence of snow melting flood disaster.Therefore,it is particularly important to actively respond to and adapt to the impact of future climate change and ensure the ecological security of northwest arid areas.Based on the national key research and development plan project,this thesis conducts data storage and prediction research for snowmelt flood disaster.The research areas of data storage are Xinjiang and Gansu.The data prediction part is mainly for Xinjiang,aiming at the collected flood disaster data and meteorological data.The main research work of this thesis is as follows:Firstly,the storage methods of disaster data are compared and studied.According to the complex characteristics of disaster data,the unstructured storage method Mongo DB is selected to construct the database and store the data.Secondly,this thesis compares the characteristics of commonly used neural network methods such as Radial Basis Function Neural Network(RBF),Deep Neural Network(DNN),Convolution Neural Network(CNN),Long Short Term Memory Networks(LSTM),etc.In this thesis,CNN and LSTM are combined to construct the cascade CNN-LSTM neural network model for flood disaster data prediction.In order to avoid the gradient disappearance and gradient explosion caused by the increase of network layer,this thesis constructs the CNN-LSTM neural network model based on residual learning,and divides the collected effective data into three data sets according to the research area and influencing factors,and compares them with the commonly used neural network models such as RBF,DNN,CNN and LSTM.The experimental results show that the two neural network models designed in this thesis are more accurate for disaster data prediction.Finally,in order to study the disaster data more systematically and comprehensively,this thesis designs and implements a flood disaster data platform based on B/S architecture based on Node.js,Vue.js,Mongo DB,Tensor Flow.js and other technologies.The front-end display,data storage,data download,and data prediction are integrated into the same platform,this data platform provides important data support and method reference for disaster prevention and related research in China.