| As important means for describing hydrological characteristics and water resource circular process,conceptual hydrological models are powerful tools for rainfall runoff simulation and flood forecast.In the application of hydrological model,the reasonability of model structure,rationality of relevant model parameters,accuracy of initial and boundary conditions and quality of observed data are main factors affecting the forecasting accuracy of a hydrological model.The Xinanjiang model is a typical conceptual hydrological model that has relatively definite physical meaningful parameters and has been successfully and widely applied,especially in the semi-humid and humid regions in China,since it was proposed.Thus,taking the Xinanjiang model as an example,the paper combines the intelligent algorithms,multi-core parallel technology and weather forecast to conduct the research of model parameter calibration and real-time rolling forecast correction of conceptual hydrological model.The research is based on the three aspects: model parameter calibration,antecedent precipitation correction and rainfall input data to improve the model application,extend the lead-time,therefore,providing decision support for disaster prevention and mitigation and reasonable water resources allocation.The main contents of the paper are as follows:(1)A multi-core parallel genetic algorithm for Xinanjiang model parameter calibration is proposed,since that the Xinanjiang model is involved with numerous parameters to calibrate,which is very time-consuming.In the paper,a multiple criteria model calibration model is built and the Technique for Order Preference by Similarity to Ideal Solution(TOPSIS)is employed to solve the multiple criteria decision-making problem and find the optimum alternative.Regarding the time-consuming characteristic of calculating fitness values,a master-slave parallel genetic algorithm is adopted to improve the computational efficiency.In addition,the tabu strategy is introduced to avoid repetitive computations on the same individuals to alleviate the system workload and improve the computing efficiency.Application results show that the proposed method can not only achieve high quality parameters,ensuring the forecasting accuracy of Xinanjiang model,but also significantly improve the computational efficiency,providing new ideas for further model parameter calibration.(2)On the basis of analyzing of the problems of traditional determination methods of antecedent precipitation(Pa),this paper proposes a gradual correction method of Pa for the Xinanjiang model.The method is based on a particle swarm optimization(PSO)to modify Pa gradually in a flood forecasting process according to the real-time rainfall and streamflows information,thus continually updating the forecast results for the reliability of Xinanjiang model.For further improvement of the PSO performance,chaos initialization is adopted to enhance the quality of the initial solutions,and tabu search strategy is employed to expand searching space and improve the global search performance for avoiding being trapped into local optimum.Moreover,an elitist strategy is introduced between the adjacent correction periods to improve the algorithm efficiency and the stability.Application results indicate that the method can efficiently improve the accuracy of Pa and significantly ameliorate the forecasting precision.(3)A meteorology-hydrology coupling flood forecasting model based on the Climate Forecast System(CFS)and Xinanjiang model is proposed in the paper.A parsing method for CFS product is described,and an unidirectional coupled meteorology-hydrology model is built for flood forecasting driven by precipitation forecast based on the CFS product.The verification results demonstration that the precipitation forecast in the future 24 hours of CFS has high performance on no rain forecast and light rain forecast.Moderate rain forecast can provide a certain reference,even though with a low TS score.However,heavy precipitation and magnitude above heavy precipitation forecast is of low reliability.The application results of meteorology-hydrology coupling flood forecasting model indicate that the Xinanjiang model combined with the future 24-hour precipitation information of CFS is efficient to prolong the flood forecast period,anticipating the flood process ahead of time.Finally,this thesis summarizes the whole thesis and conclusions,and prospects the next work content. |
