The Modeling Research Of Data Mining In Oil Field Measure Program

With the development of oil-fields, it is necessary for the enterprise to make a reasonable measure programming to obtain the better benefit under the stable production. The issue of the oil-field measure programming is to construct a scientific and reasonable model of measure programming prediction. At present, the commonly model is the fixed mathematical model and the precision of prediction is badly. The prediction of measure effect is a complicate nonlinear system. Artificial neutral network is of stronger capability to mapping nonlinear systems and can resolve this kind of problems. The algorithm is presented and improved in the paper to enhance the generalization capability and the stability of the model.The disadvantages of the Standard Genetic Algorithm results in premature convergence, bad stability and low convergence speed. This paper proposed a Genetic Algorithm based on adjust adaptively and nonlinearly. Traditional BP algorithm has low convergence speed and is subject to fall into minimal point. An adaptive and nonlinear Genetic Algorithm is applied to optimized construct and weights. A LMBP algorithm based on gradient jumping is proposed and some rules such as the judgment condition of falling into minimal point and increasing speed of gradient are established. Two hybrid algorithm, which are combination of the modified genetic algorithm (NLSAGA) and the LMBP algorithm are proposed. Finally, existing experiment data samples are used to train the ANN. The simulation results show that model presented in the paper is reasonable and the prediction model based on NLSAGA_LMBP2 is better than others.