Analysis Of Monthly Runoff Characteristics And Prediction Model Research

Runoff prediction is an important direction of hydrology research.To carry out research on runoff evolution law,find out the trend of runoff increase and decrease,and make scientific and reasonable runoff prediction on the basis of this,plays an important guiding role in the comprehensive development,utilization,rational allocation and efficient management of regional water resources.Based on the analysis of the evolution law of inflow runoff of a certain hydrologic station,the SARIMA model,NAR model and SARIMA-NAR combined model are established to predict the inflow of A hydrologic station in the future.The main work contents and research results are as follows:(1)on the basis of time series analysis and random hydrology,the deterministic components in the runoff evolution law of this hydrologic station are analyzed.Kendall rank correlation test and Spearman rank correlation test were used to analyze the evolution trend of inflow runoff series in this hydrological station.Order clustering and rank sum test were used to diagnose the sudden change of runoff series.The results showed that the sudden change of monthly runoff time series was not significant.Wavelet analysis was used to analyze the periodicity of runoff series,and it was found that the runoff evolution process had multi-scale periodic changes,and the first major period of 18 months and the second major period of 9 months were identified.(2)on the basis of time series analysis,the SARIMA seasonal product model is established to fit and forecast the runoff flow sequence,and it is found that this model can achieve better prediction accuracy.(3)based on relevant neural network theories,a runoff prediction model based on NAR dynamic neural network was established to carry out fitting and combining prediction of runoff,and the prediction effect was good.The stability of sarima-nar combined prediction model was improved by testing.(4)develop a matlab-based runoff characteristic analysis and prediction system interface to simplify work steps and improve work efficiency.