A Hybrid KNN-FWA-ELM Flood Forecasting Model And Its Applicability For Typital Medium And Small Watershed In The Loess Region

China is the largest loess distribution region in the world.The loess accounts for nearly 7% of the national land area.The loess hilly is the main landform of the loess region,which is characterized by the strong topographic cutting,the vertical and horizontal gully,and the steep slope.In the loess region,the rainfall period is mainly concentrated from July to September and the high intensity of rainfall causes the flood in the medium and small watershed to be of rapid flow production,high flood peaks,short duration,rapid flow rate and destructive power.The underlying surface of the loess region is susceptible to natural and human activities.The forecasting of the flood in medium and small watershed caused by sudden rainstrom and local heavy rainfall has always been an important research topic in related fields such as natural disasters.The paper uses a hybrid intelligent computing method,applying Fireworks Algorithm-FWA and K Nearest Neighbor-KNN method to improve the Extreme Learning Machine-ELM model.A data-driven flood forecasting model for medium and small watershed in the loess region under the condition of underlying surface changes is established and its applicability research is conducted.The research not only has important theoretical significance for flood forecasting methods,but also has high application value for flood control and disaster reduction in medium and small watershed in the loess region.This paper takes the Gedong basin in the loess region of western Shanxi as the research area,by analyzing the characteristics of the underlying surface of the basin,the underlying surface change is divided into three periods,and the hydrological characteristics of the basin and the rainfall losses during the flood process are analyzed and calculated.The paper proposes a KNN-FWA-ELM hybrid flood forecasting model,which is further applied to flood forecasting in the Gedong basin.Then the applicability of KNN-FWA-ELM model in flood forecasting of typical medium and small watershed in loess region is evaluated,through comparing the simulation results of KNN-FWA-ELM and HEC-HMS models under different underlying conditions and different flood characteristics,and analyzing the application effect of KNN-FWA-ELM model in the Wangjiagou experimental plot.The main findings are summarized in the following aspects:(1)Analysis of hydrological characteristics of the basin.The annual precipitation in the basin from 1964 to 2012 does not decrease significantly,but the annual runoff depth shows a significant downtrend.The multiple linear relationship between annual runoff coefficient and land use change shows that increasing the forestland area in the loess region can effectively intercept runoff and reduce mountain flood disasters.The comparative analysis of flood evolution characteristics in different periods indicates that water conservation measures such as building terraces and afforestation can effectively intercept precipitation,attenuate flood peak discharge and flood volume,and delay the flood peak lag time.Under the same level of rainfall and rainfall intensity,the attenuation amplitudes of flood volume and flood peak discharge are 9.32%-92.90% and 4.61%-80.22%,respectively,and the flood peak lag time extends 0.58-1.38 h.The rainfall losses in the basin are affected by the forestland area percentage and slope.The rainfall losses increase with the increase of the forestland area percentage and the decrease of slope,and the impact of the forestland area percentage on the rainfall losses is greater than that of the slope.Under the same magnitude of rainfall,from 1964 to 2012,as the percentage of forestland area increased,the correlation coefficient between the percentage of rainfall losses and the percentage of forestland area increases from 0.6397 to 0.8413.(2)Establishment of KNN-FWA-ELM flood forecasting model.For the special topography,micro-geomorphology,vegetation and other underlying surface conditions of the loess region,the advantages and disadvantages of the ELM model in the flood forecasting of typical medium and small watershed in the loess region are analyzed.In order to achieve model parameter optimization and continuous flood forecasting,MATLABR2016 b is used as the platform to self-program,and the ELM model is improved by using fireworks algorithm and K nearest neighbor method.A hybrid data-driven model(KNN-FWA-ELM)for flood forecasting in medium and small watershed in the loess region is proposed and established,and the changes of the underlying surface characteristics of the basin in different periods are considered in the input and output of the model.(3)Application and evaluation of KNN-FWA-ELM flood forecasting model.The KNN-FWA-ELM and ELM models are applied to the Gedong basin,and the floods in different periods and different grades are simulated.The flood peak discharge error,peak-appearence time difference,Nash coefficient,correlation coefficient and root mean square error are used to analyze and evaluate the flood simulation results of the two models.The results demonstrate that KNN-FWA-ELM model has better simulation effect and higher simulation accuracy than ELM model.(4)Applicability analysis of the KNN-FWA-ELM flood forecasting model.In order to further explore the applicability of the KNN-FWA-ELM model in the flood simulation of the medium and small watershed in the loess region,the simulation results of the KNN-FWA-ELM model and the HEC-HMS model for all floods and floods in different periods and different flood grades are comparatively analyzed according to the flood simulation qualified rate,flood peak discharge error,peak-appearance time difference,Nash coefficient,correlation coefficient and root mean square error.Based on the root mean square error and mean absolute percentage error,the paper compares the simulation results of flood peak discharge,flood volume and flood peak lag time of the KNN-FWA-ELM model and the HEC-HMS model,and applies the KNN-FWA-ELM model to the flood simulation of the Wangjiagou experimental plot.The results show that the KNN-FWA-ELM model has a good overall effect on the simulation of floods in Gedong basin and Wangjiagou experimental plot,and the simulation results of the KNN-FWA-ELM model in the Gedong basin is better than that of the Wangjiagou experimental plot.From the aspect of the floods in different periods,the KNN-FWA-ELM model is appropriate for floods with rapid peaking and relatively long flood duration,relatively long interval between multiple peaks or flood with the first peak as the main peak,irregular process and a long duration;from the aspect of different grades of floods,the KNN-FWA-ELM model has a better simulation effect on the great flood;from the aspect of flood peak discharge,flood volume and flood peak lag time,the KNN-FWA-ELM model has a better simulation performance on the flood peak lag time.