Study On River Water Level Prediction Based On Data Cleaning Method

Frequent flood disasters is extremely serious impact on the economic development of our country. Flood forecasting, as a very important non-structural flood control measures for flood control is of great importance and significance.Combining with the project "Tongren City, Guizhou Province hydrology remote monitoring and integrated nformation management system development",flood forecasting model is used to predict certain river water level changes.Due to the impact of external factors, natural river hydrological data collection will produce an error, we have to remove these error data to improve the accuracy of prediction of river water.BP neural network is highly adaptive, nonlinear mapping and fault tolerance,which is widely used to solve issues such as river water level forecasting.But BP neural network has some shortcomings, such as slow convergence,being easy to fall into local minimum problems etc.Thus improved BP neural network algorithm is often applied.In this paper,the PSO algorithm is analyzed and used to optimize weights and threshold parameters in BP neural network algorithm. affect particle swarm optimization analysis to determine the PSO optimized BP neural network processes and PSO-BP neural network model.Additionally, the influence on the parameters of particle swarm optimization algorithm is analysed and the PSO BP neural network optimization process and the PSO-BP neural network model is determined.The paper explains the sample data sources, analyzes the causes of error in the sample data,uses the outlier detection,which is one of the data cleaning method, and determines the Excluding outliers process.The process is as follows:first of all preliminary confirm the sample data on flow measurement software;secondly location the sample data outliers and remove;finally constructe the sample data.According to analysis of the study area as well as the complexity of the relationship between river hydrology object, using the river water level forecast flood forecasting,building BP neural network forecasting model and PSO-BP neural network prediction model.By studying and trainingthe history sample data,the output data of the two models is obtained. Then assessing the predictive accuracy of the two models based on "intelligence hydrological forecasting practices" (SL250-2000). From the view of engineering application,The results show that the accuracy of PSO-BP neural network model is higher than BP neural network model. The study can provide a reference for the river’s water level change research.