Study On The Application Of Distributed Hydrological Simulation And Optimization Algorithm Based On Multi-source Information

As a technology-effective approach in recognizing the law of hydrologic cycle evolution, physically-distributed basin hydrological simulation is one of the priority areas in hydrology nationally and internationally right now. However, two problems existing in the application of physically-distributed basin hydrological simulation to the northern basin of our country are as follows:(1) non-availability of measurements at a scale required for distributed models due to a low density of hydrometric station networks, and (2) complication of simulation due to interaction between complicated human activities and physical mechanism of runoff yield and conflux. Two methods for solving the two problems are as follows:(1) employing all available multi-sources to reduce data uncertainty, and (2) modifying model structure to be appropriate for dealing with complicate human activities. Then fast and stable optimization algorithms are required for modelling hydrological processes due to the complicate simulations and increasing parameters in distributed models. With the series of problems as objects, utilizing the observation information based on3S technique, the engineering information based on statistics and investigation, and the heuristic information based on intelligent data analysis in Biliu river basin, Fengman reservoir basin and Tang-Wang river basin, this dissertation does research on the application of distributed hydrological simulation and optimization algorithm based on multi-source information. The study contents and results are as follows:(1) Localizing construction of distributed basin hydrological model SWAT. The model structure, principal and computing methods of main modules are introduced. The standardizing methods for constructing data required for localizing model application, such as meteorologic data, soil data, and land use data, are elaborated. Then the researches on interpolation and extending approach based on image recovery technique, distributed hydrological simulation considering small-and medium-sized water storages, and optimizing search algorithm considering heuristic information are supported greatly.(2) Combining image recovery technique with normal multi-correlation analysis, precipitation interpolation and extending approach based on image recovery technique is proposed. Firstly, the stations established recently and the stations with short-term data only are set as unknown stations with incomplete information, and the multi-correlation of short-term precipitation data between unknown stations and their adjacent stations could be analyzed and excavated; secondly, the precipitation data of unknown stations could be interpolated and extended with known data of adjacent stations. Through the application of the approach proposed in this paper for precipitation interpolation and extending of stations in Biliu river basin, such as Kuangdonggou, the approach is examined. Compared with normal IDW and multi-correlation analysis, annual precipitation precision is improved by8.30%and4.61%respectively, and precipitation precision of flood season is improved by12.50%and5.54%respectively in Kuangdonggou. The application of the interpolated and extended precipitation is examined with SWAT on the basis of the areal rainfall simulation method. Compared with several scenario simulations of multiple normal approaches, monthly simulated runoff precision is improved by5.67%,10.19%and6.32%respectively, i.e. basin simulated runoff precision is improved more with the precipitation interpolation and extending approach based on image recovery technique.(3) Hydrological simulation approach for water storages in SWAT is improved, and distributed hydrological simulation approach for small-and medium-sized water storages is proposed. Firstly, with the Landsat TM data for flood seasons in wet years and design information for water storages, relationships between water surface area and storage volume for water storages are constructed; secondly, water balance of the series-parallel scheme based on the spatial topological relationships among water storages is constructed; finally, the physical basin model parameter and human interference parameter calibrations are processed with minimal human activities and stable human activities. Compared with several parameter scenario simulations of SWAT, the improved model is examined. Compared with original SWAT, the model performance is improved by5.81%(R2) and22.39%(NSE) over the validation periods, and the model performance is improved by9.52%(R2) and29.69%(NSE) for the flood seasons over the validation periods at Panshi and Dongfeng hydrologic stations. The performance improvements over the validation periods are mostly due to the performance improvements in the flood seasons over the validation periods, and the performance improvements in the flood seasons over the validation periods are mainly due to the improvements of the spatial topological relationships among water storages in the improved SWAT.(4) Dynamically dimensioned search (DDS) algorithm is improved, with reference to the changing sensitivity information of multi-dimension decision variables, the improved DDS algorithm based on sensitivity analysis is proposed. The key feature of DDS was motivated by our experience with manual calibration of watershed models where early in the calibration exercise relatively poor solutions suggested the simultaneous modification of a number of model parameters but as the calibration results improved, it became necessary to only modify one or perhaps a few parameters simultaneously so that the current gain in calibration results were not lost. In short, the DDS algorithm searches globally at the start of the search and becomes a more local search as the number of iterations approaches the maximum allowable number of function evaluations. In the improved DDS algorithm, firstly, changing sensitivity matrix of multi-dimension parameters and cumulative sensitivity of each parameter are updated dynamically during the process of optimization according to the relationship between dimension selection and optimizing efficiency; secondly, the probability of choosing each parameter is calculated during the process of optimization according to its cumulative sensitivity; finally, the optimization strategy with unequal probabilities of choosing parameters during the process of parameter optimization is realized to improve convergence rate and optimizing efficiency. The improved DDS performance is compared to the DDS algorithm for Tangwang river basin SWAT model parameter calibration. Results show the improved DDS to be more efficient and effective than DDS. Compared with DDS, monthly simulated runoff precision with the improved DDS at Yixin hydrologic station is improved by5.83%in100total model evaluations, and the time required by the improved DDS is reduced by9.37%to find equally good values of the objective function (SSE=4.7×106). Additionally, the improved DDS performance is compared to the DDS algorithm for joint operation of water resources closely related to hydrologic cycle simulation and evaluation to examine the improved DDS universality. Results of the application to operation of a large multi-reservoir (multi-reservoir MR) in the northeast show water supply and its assurance rate, belonging to reservoir A and aggregative reservoir contained in MR and their supplying regions downstream, are improved with fixed water consumption. Compared with DDS, water supply for agriculture of reservoir A and water assurance rate for agriculture in its supplying regions downstream with the improved DDS are increased7million cubic kilometer and6.77%respectively, total water supply and water supply for agriculture of aggregative reservoir and water assurance rate for agriculture in its supplying regions downstream with the improved DDS are increased18.87million cubic kilometer,16.3million cubic kilometer and10.62%, water assurance rate for agriculture in the common supplying region of reservoir A and aggregative reservoir with the improved DDS is increased by12.54%in1000total function evaluations. The availability of the improved DDS in hydrologic cycle simulation and operation of water resources is justified in the two cases.Finally, a summary is given and some problems to be further studied are discussed.