Research On Key Technologies And System Implementation Of Intelligent Early Warning Of River Flood Disaster

River-flood and geological disasters happen frequently in mountainous areas,which cause giant loss of life and property and damage of surrounding terrain and vegetation.Therefore,the accurate and efficient mountain torrential flood disaster prediction and warning play important roles in ensuring safety and protecting ecological environment.In recent years,the research on artificial intelligence is increasingly in-depth.In this case,combining it with disaster prediction and warning provides the possibility for mountain torrent disaster prediction and warning,and also provides a new direction and opportunity for the further development of flood prevention.Under this background,this work introduces the thought of intelligence algorithms into the field of river-flood disaster prediction,and proposes methods of scared river parameter prediction model and scared river condition classification method.Besides,the work designs a river-flood monitoring and warning system.The main researches of this work are as follows:1)To solve the problem of multiple features in river-flood parameters,the work takes the similarity calculation between flash floods and hydrological parameters to measure the correlations,and takes strong correlation parameters as influential factors of river-flood.Then,the work uses extreme learning machine to construct the prediction and classification models since extreme learning machine has less training parameters and generalization performance.Besides,the work employs whale optimization algorithm to optimize the model since whale optimization algorithm has less adjust parameters and strong robustness.2)To solve the problem of initial population distribution inequality and weak global search capability,the work applies halton sequence to initialize the population,which makes solutions distribution more balanced in solution space.Additionally,the work combines the nonlinear time-varying factor and inertia weight mechanism to balance and enhance the global search capability.The experiments indicate that the method based on improved whale optimization algorithm outperforms the baseline methods in terms of convergence effect and find excellent accuracy.3)To predict the varying of hydrological parameters accurately,the work propose a river-flood parameter prediction method,called IWOA-ELM.The method uses the improved whale optimization algorithm to optimize the parameters of extreme learning machine.For the small-scale dataset of river-flood disaster,this work designs a sliding window mechanism,a rolling forecast function,to enhance the dataset.Besides,it adopts the intelligence algorithm to predict the next river-flood parameters,and further achieves flood prediction.The experiments indicate that the IWOA-ELM outperforms the baseline methods in terms of root mean square error and fitting coefficient value.4)To solve the problem of river-flood intelligent warning,the work takes the influential factors of river-flood as input features to classify the scared river condition,which is based on the correlation parameters.Then,it classifies the scared river condition by combining improved whale optimization algorithm and extreme learning machine and using designed classified strategy.The experiments demonstrate that the designed classified strategy outperforms the baseline methods in terms of accuracy and precision.5)To achieve the remote monitoring for river-flood disaster,this work designs a remote monitoring and warning system by combining local sensing network and telecommunication network and using Java programing language.The system adopts self-organized network to monitor,collect,and transfer river-flood features,and uses the designed IWOA-ELM model and classified strategy to achieve river-flood disaster prediction and warning.