| The optimization of large scale cascade reservoirs is a nonlinear andhigh-dimensional problem. In addition, the optimization of large reservoir suffers lowefficiency resulting from searching in a wide range of its large capacity and high waterhead. Analysis on the reservoir optimization results under different flow conditions isalso not stressed yet.This study presents a methodology which establishes a rational optimization spaceto improve the optimization efficiency for large scale reservoir to solve the problem ofcontinuous optimization in wide range. The method couples Incremental DynamicProgramming (IDP) to Heuristic Algorithm (HA). Taking advantage of the convergenceproperties of IDP and powerful ability of global optimization of HA, the IDP-HAmethod uses the optimized result from IDP as a reference and builds a small scale spacewhich covers the global optimum for HA to handle high-dimensional optimization. TheIDP-HA method improves the computational efficiency for large reservoir optimizationproblem.For the optimization of cascade reservoirs, this study decouples the problem into (i)dispatching water among different reservoirs in the cascade system and (ii) dispatchingwater among different types of hydropower units in each hydro power station. Based onthe analysis of the characteristics of hydropower units, two different linearization waysare adapted for two kinds of reservoirs. Heuristic algorithm is used to solve the problem(i), and Simplex Method (SM) is to solve the linear programming problems for (ii). Twoproblems are coupled by the hydraulic and electric connections in the cascade system.Thus a HASM method is proposed for the optimization of large-scale cascade reservoirs.Dynamically Dimensioned Search algorithm (DDS) is firstly used in reservoiroptimization.To analyze the optimization results under different flow conditions, this studyfirstly takes advantage of the nonparametric response surface method based on radialkernel function. By the establishment of response surface, which describes therelationship between the inflow and the corresponding optimal release, the optimalreservoir operation plan can be found quickly, and this method can provide a referencefor the actual reservoir operation. Finally, the HASM method is applied to the Three Gorges-Gezhouba cascadereservoirs for the long-term optimization. The results prove that the simulative accuracyof the model is high, and there is still improving room for the operation of the ThreeGorges–Gezhouba cascade system. |
