Study On Application Of Genetic Algorithm To Municipal Engineering

For the municipal engineering is a complicated and huge system, it is difficult to optimize, model, predict, imitate, evaluate, make policy and analyze in the municipal engineering by the traditional methods. After analyze the existing methods on modeling, evaluation and prediction and gain the corresponding shortcomings, this thesis aims at trying to perform the optimization, evaluation and prediction in municipal engineering by Projection pursuit models, Matrix games models and BP neural networks models based on Accelerating Genetic Algorithm.First, on the foundation of theoretical study of empirical value of window radius, this thesis modifies one-dimensional clustering modeling. Then, put forward an evaluation model, which aims at improving the generalization of neural networks, based on multi-dimensional projection pursuit and BP neural networks. Next, a new method that solves matrix games and a selecting best scheme model based on genetic algorithm are presented. Finally, a nonlinear combination forecasting model based on neural networks is applied to forecasting water command in order to take advantage of various information from other technologies. All the models that stated above are applied to the practical cases of municipal engineering, and reasonable and effective results, which show the feasibility and reliability of the models, are obtained. It is concluded that the models stated above can be applied to municipal engineering generally.