Study On Runoff Prediction Of Chaohe River Basin By SSA-SVR Based On CEEMD Decomposition

Hydrological forecasting,especially mid-and long-term runoff forecasting,is a crucial part of water resources management and the operation of irrigation works.Precise and trusy runoff forecasting can furnish precious info for flood control,power generation,water supply and drought control decisions.On the basis of general analysis of the trend and features of hydrological elements,the SVR model was optimized by the Sparrow Search Algorithm,and establishes an annual runoff prediction model based on causality.Combining the Complementary Ensemble Empirical Mode Decomposition method with the SVR model,a time series-based combined monthly runoff prediction model CEEMD-SSA-SVR is established,and use cubic chaotic mapping and bird swarm algorithm(BSA)to further optimize the CEEMD-SSA-SVR model to improve the prediction performance of the model and provide a scientific reference for the optimal allocation and comprehensive utilization of water resources.The main research contents of this paper are as follows:(1)Using MK trend test method,Sen’s slope and MASH moving average method to analyze the trend of changes in hydrological and meteorological elements,which shows that runoff and rainfall have a downward trend over time,and temperature has an upward trend.(2)The C-C method was used to verify the chaotic characteristics of the non-linear dynamic system of the water resources in the basin,and the phase space reconstruction of the runoff sequence was carried out.(3)Using the SSA algorithm to optimize the SVR and BP neural network models,and established SSA-BP and SSA-SVR annual diameter survey models,compared with the multiple linear regression model,which shows that the SSA algorithm can effectively optimize the model,and the non-linear prediction model has more excellent prediction performance.(4)For the monthly runoff sequence decomposed by CEEMD,a time-series-based SVR monthly runoff prediction model(CEEMD-SVR)was established,and the SVR model was optimized using Differential Evolution algorithm,Artificial Bee Colony algorithm(ABC)and SSA algorithm.CEEMDABC-SVR,CEEMD-DE-SVR,CEEMD-SSA-SVR monthly runoff prediction models were established respectively.The results show that CEEMD can effectively remove sequence noise,the SSA algorithm has stronger global search capabilities,and the CEEMD-SSA-SVR model has higher prediction performance.(5)The cubic chaotic map and the BSA algorithm are introduced to further optimize the SSA algorithm,and the bird flock fusion sparrow search algorithm(C-BSSA)based on the chaotic map is proposed,and the C-BSSA-SVR model is constructed to predict the 6 modal components and establish A CEEMD-based C-BSSA support vector machine monthly runoff prediction model is used.The results show that the optimization performance of the C-BSSA algorithm is improved,the running time is effectively reduced,and the defects that are easy to fall into the local optimum are improved.The prediction effect of CEEMD-C-BSSA-SVR model is better.