An Improved Particle Swarm Optimization Algorithm And Its Application In Hydrological Model

Improving the accuracy and efficiency of runoff simulation and flood forecasting is of great signifcance for carrying out the spirit of the No.1 Central Committee of the CPC in 2011,completing the project of water resources monitoring and capacity building,and futher enriching the information of water resources supervision in China.Runoff forcasting and flood forecasting rely mainly on hydrological model.Xinanjiang model is the representative of the lumped hydrological model,and the EasyDHM model is the representative of the emerging distributed hydrological model,as the object of this study.Reasonable model parameters are very important for im-proving runoff simulation performance.Particle swarm optimization(PSO)algorithm is simple in structure and fast in conergence,so it is suitable to solve the problem of parameter calibration of hydrologcal model.This paper analyzes the PSO algorithm mechanism and improves the peformance of PSO algorithm to improve the accuracy and efficiency of hydrlogical model runoff forecasting.And the improved PSO algo-rithm is extended in different hydrological models.The commen parameter setting of PSO algorithm is not universal for specific op-timization problems,such as hydrologic model parameter calibration.In order to anlyze the influence of PSO parameter setting on the simulation results of Xinanjiang model,the orthogonal experiment was designed.The results show the optimal PSO parameters(pop=80 w=1.3~0.4,c1=1.85,c2=2.5,linear decreasing,m=0.05),and the population size and inertia weight has significant influence on the results of simu-lation.This paper proves that the reasonable PSO algorithm parameter setting can efectively improve the simulation accuracy of Xin'An River model.In this paper,an improved particle swarm optimization(IPSO)is proposed by changing the inertia weight strategy.All particles are divided into four parts,using constant inertia weight,time-varying inertia weight,random inertia weight and adap-tive inertia weight strategy update their speed,and four strategies are parallelly com-puted.IPSO and the other 4 kinds of classic PSO algorithms are appliced in nurmical experiments and parameter calibration of Xinanjiang model.Considering the aspects of simulation results,quality of convergence value,convergence speed and stability of the algorithm comprehensively,IPSO is better than other 4 PSO algorithms.To verify the applicability of the IPSO algorithm,it is also used to solve the pa-rameter calibration problem of EasyDHM model which is a distributed hydrological model with 28 parameters.Runoff simulation of 50 representative floods in the upper reaches of Hanjiang was carried out.84%of the flood water balance error is less than0.2,71%of the flood Nash efficiency coefficient is higher than 0.7,84%of the flood peak flow simulation is qualified,and 90%of the flood no peak time error.In general,the IPSO algorithm is suitable for solving the parameter calibration problem of EasyDHM model.Comparing the IPSO algorithm with the PSO algorithm,70%probability IPSO is more suitable for the EasyDHM model than the PSO algorithm.When the flood runoff is less than 10000(m~3/s)or greater than or equal to 30000(m~3/s),it is necessary to replace the PSO algorithm with IPSO to determine the pa-rameters of the EasyDHM model.