Research On Methods For Single-objective Optimization Of Hydrological Model And Multi-objective Decision Making Of Multi-reservoir Operation

Hydrological forecast and watershed water resources utilization and management are two important issues in water science, the former is mainly performed by various hydrological models and the latter depends on 1) water conservancy facilities such as reservoirs, water diversion and transfer projects, and 2) non-engineering measures, for example, reservoir operation. The joint optimization operation of reservoirs is an effective tool to maximize the comprehensive benefit of the multi-reservoir system in flood control, power generation, water supply, navigation and ecology, and at the same time to realize the sustainable development of society and economy. Moreover, the operations of multi-reservoir systems highly depend on reliable hydrological forecast, and also have influence on the hydrological system which may change the evolution law of hydrological phenomena under natural conditions and bring significant challenges to the water resources development and utilization. Therefore, seeking for excellent parameter calibration method for hydrological models to improve their forcast accuracy and exploring reasonable multi-reservoir operation rules are two key technical problems to be solved in water science. This thesis focuses on solving two parametric model optimization problems:1) single-objective optimization of parameters of the hydrological forecast models and 2) multi-objective optimization and decision-making method for the multi-reservoir operation rules, the main work and innovative achievements are listed as follows:(1) In order to solve the global optimization problems of parameters of hydrological models, a hybrid GA/CS algorithm (GACS) is developed and its performance is examined by four complex constrained optimization test problems. The GACS employs the real-coded pattern to improve its computational efficiency, the adaptive probabilities of crossover and mutation to ensure its optimization ability, the elitist strategy to accelerate the convergence, and a tournament selection operator based on comparisons of both fitness and constraint violation between different individuals to handle constrained conditions which avoids the difficult problem to choose the appropriate penalty parameters. In addition, the search mechanism of Levy flight in the cuckoo search (CS) algorithm is also introduced to expand the search scope and increase the population diversity, which further improve the global search ability of the new algorithm.(2) First, a novel nonlinear Muskingum flood routing model with a variable exponent parameter and simultaneously considering the lateral flow along the river reach (named VEP-NLMM-L) is developed. Then, the GACS is successively applied for precise parameter estimation of the VEP-NLMM-L. Finally, the validity and applicability of the VEP-NLMM-L are tested on three watersheds with different characteristics (Case 1 to 3). To overcome drawbacks of incomplete considerations of elements, low forcast precision, poor applicability that exist in traditional linear and nonlinear Muskingum models, several improvements are presented in the VEP-NLMM-L:1) introducing an exponent parameter ? to represent the nonlinear relationship between the channel storage and the weighted discharge,2) adding a coefficient parameter ? to account for the effect of the lateral flow,3) dividing the inflow range into L levels (i= 1,2,-, L) and each level uses a different value ?i to describe the characteristics of unsteady flow in river flood routing,4) the new model includes (2L+2) parameters and (L-2) constraints for the model structure.(3) A newly developed multi-objective decision making (MODM) methodology of combined use of the NSGA-II algorithm and the SEABODE approach is proposed. The NSGA-II algorithm aims at solving the multi-objective optimization problems (MOPs) to generate enough number of design alternatives, which form a decision space. The SEABODE which is a multi-attribute decision making (MADM) approach is responsible for further evaluating, sorting and sieving these alternatives. The SEABODE is based on the theorem of efficiency of order k with degree p or [k,p]-Pareto-optimal points to successively eliminate inferior alternatives, which can avoid the weakness in traditional aggregation approaches such as the AHP, TOPSIS, Vague set, grey correlation analysis, etc. The demonstration of a numerical example also illustrates the effectiveness of the SEABODE approach.(4) The NSGA-?—SEABODE method has applied on a regional water supply system with three reservoirs located in the Jialing River for extraction and preference ordering of the multi-reservoir water supply rules in dry years, where the results examine the practicability and merits of the proposed methodology. First, the well-designed multi-reservoir operation rules consists of a two-point type time-varying hedging policy for a single reservoir and a simple proportional allocation policy of common planned water demand between parallel reservoirs. Then, to comprehensively evaluate the reservoir water supply rules, the reservoir performance criteria and the water shortage indices compose a complete evaluation index system. Next, we establish the mult-objective optimization model for the selected multi-reservoir water supply system with the objectives of minimizing 1) the total deficit ratio (TDR) of all demands of the entire system in operation horizon and 2) maximum deficit ratio (MDR) of water supply in a single period. Finally, we obtain a sizeable number of noninferior alternatives of the multi-reservoir operation rules by using the NSGA-? and find out the most preferred alternative for each reservoir with the help of the SEABODE approach and according to several specified evaluation criteria.