Prediction Of Water Level At Urban Waterlogging Point Based On Artificial Neural Network

Urban waterlogging is a serious threat to the property and life of urban residents,it is an urgent problem to be solved in the process of urbanization.In addition to adopting engineering measures to alleviate waterlogging problem,precautionary and forecasting research for urban waterlogging has gradually received attention.Traditional research method for waterlogging prediction is mostly based on the urban rainfall runoff model,but the lack of basic data and the limited knowledge of hydrophysical processes limit the application of the urban rainfall runoff model.With the development and application of the monitoring cloud service system for urban waterlogging,the data of waterlogging is monitored in real time,and the new research method for urban waterlogging warning and prediction based on data-driven model has become a new trend.The monitoring cloud service system for urban waterlogging can operate stably and reliably under the harsh outdoor test environment,which is an important prerequisite for providing reliable data support for the urban waterlogging warning and prediction research.Performance testing of the monitoring cloud service system for urban waterlogging has been finished based on the indoor drainage pipe network model,the equipments have been selected and assembled in this research,such as,the terminal for measurement and control based on Internet of Things,the module for network communication and the power supply device for these equipments,these assembled equipments have been installed in the representative waterlogging point,and the performance in the outdoor environment of the monitoring cloud service system for urban waterlogging has been tested.The test result shows that the monitoring cloud service system for urban waterlogging has good real-time performance,high stability,the selected communication methods is cost-effective,and the battery has strong power supply capability,which can provide reliable data support for the research of urban waterlogging warning and prediction problem.Prediction of water level at urban waterlogging point based on artificial neural network,which is a research method based on the monitoring data of rainfall and water level data at urban waterlogging point,and the the nonlinear approximation ability of the artificial neural network.This method provide a new entry points for the research of urban waterlogging warning and prediction problem.In this paper,the Elman neural network has been introduced into the research,and the particle swarm optimization algorithm has been used to optimize the number of neuron nodes in the input and hidden layers of the neural network.The research example shows that the Elman neural network model is suitable for the research ofurban waterlogging warning and prediction problem.With the increase of prediction duration,the prediction accuracy of the model is reduced,which can basically meet the practical application requirements.At the same time,in order to further improve the prediction accuracy of the model,the echo state network has been introduced into the research of urban waterlogging warning and prediction problem,the echo state network has been used to dynamically approximate the mapping relationship of the rainfall and water level at waterlogging point.Aiming at the problem that the key parameters of the echo state network model and the time series embedding dimension are selected to be subjective,the particle swarm optimization algorithm has been used to jointly optimize the above values to improve the prediction accuracy of the model.The prediction effects of echo state network,Elman neural network and BP neural network have been compared and analyzed.The research example demonstrates the applicability and superiority of echo state network in the research of prediction of water level at urban waterlogging point.