Application Of Modified Genetic Algorithm And Back Propagation Artificial Neural Net To Water Resource Projects

APPLICATION OF MODIFIED GENETIC ALGORITHM AND BACK PROPAGATION ARTIFICIAL NEURAL NET TO WATER RESOURCE PROJECTSIn the water resources projects, most problems are nonlinear and are multidimensional for the optimization. Analysis method and enumeration method are adopted to solve these problems traditionally, yet they have different defects in some aspects. The aim of this dissertation is to explore methods to resolve these problems.The genetic algorithm is an advanced kind of global optimization means. First of all, modifications have been made based on research of predecessors in genetic algorithm, then the real coded accelerating genetic algorithm has taken shape. Furthermore, by used in calculating water level curve of natural river course, Calculating depth of water of Spillway dam downsteam section, analyzing the happening frequency curve of hazard, optimizing the partition gives the water system in large irrigated area and in other water conservancy projects, it was proved that the Real coded Accelerating Genetic Algorithm was advantageous in performance and convenient in the global optimization. Multidimension, nonlinearity and lack of samples were most characteristics of those problems in water conservancy projects such as forecast, classification, evaluation, so it is difficult to built exact mathematical models to resolve.The Artificial Neural Net model is a kind of messages treatment system that has performance of self-organized learning, recalled in the connection. Error back-propagation artificial neural net based on the simulated annealing algorithm (abbreviated, SABP) was built by merging the simulated annealing algorithm (abbreviated, SA) in error back-propagation, which made many practical problems in water conservancy projects resolved efficiently.In order to improve the BP-Net's ability further, attempt that using the Genetic algorithm to optimize the weights and the thresholds of BP-ANN was made to form the BP artificial neural net based on RAGA and SA (abbreviated, RAGASABP). Through using in predicting the loss of inundation, in the crop-water model, in the calculation to amount of water requirement of crops, in predicting the groundwater level changed with times, in evaluating water quality by many criterions, the RAGASABP was good at nonlinear mapping, classification and evaluation.Finally, the conclusion of the thesis was that many concrete problems could be settled by the real coding based accelerating genetic algorithm and the error back propagation artificial neural net based on RAGA and SA in water resource projects.Postgraduate: Xing Zhenxiang Supervisor: Fu QiangMajor: Soil and Water of Agoricultrual Engineering...