Application Of Well Logging Curves Indentifying Based On Kernel Method

Through the development for several years, Daqing oilfield accumulates a great quantities of logging data which contains much important knowledge, but also incomplete probably. Before, people obtained the important instruction information depend on artificial judgment and experience accumulation, which is waste time and effort, and not easy to identify. With the encrypting mining, completed by artificial to measure logging curve can't satisfy the development progress of oilfield. Adopted computer to measure logging curve automatically is subiected to people more and more. But because of the imperfection of identify model, the difficulty of withdraw the appearance characteristic parameter and the inelasticity of recognition algorithm, many processes can't be completed automatically by computer, therefore it is hard to apply to the spot in the processing of logging data.This paper proposes the application of well logging curves identifying based on kernel method. Kernel methd map the nonlinear problem to high dimension space and translate to linear separable problem.It advances calculating ability of linear learning machine by using point algorithm in high dimension space. Generally speaking, replacing a point algorithm with the kernel function in any calculate which contains point algorithm is called kernel method. At present, most application of kernel function is selecting giving settle of kernel function, researching the classification ability of them. But a determined kernel function just reflect the input vector definition in feature space, there are many nonlinear mapping corresponds to the measure. It is worthy of considering that selecting kernel function according to the data, not selecting in the existing kernel functions. This paper carries on a research on the existing kernel functions, constructs a new kernel function and apply it to well longging curves identify.The main work of the paper:1. From measure well principle and it's explains discuss the situation of logging curve identify, summary the difficulties of logging curve identify at present;2. Set out from the angle of mathematics, study the foundation theories knowledge of kernel function;3. On the base of study on the existing kernel functions, carry on a research to the structure kernel methods, structure kernel function basis the mixture kernel function theories and prove the mixture kernel satisfy the Mercer theorem;4. On the foundation of above theories, carry on identify the data of the Daqing oilfield and analysis the result. The subject applies kernel method to logging curve identify, not only carry on logging curve identify automatically, improve efficiency, but also rise the accuracy for identifying, pushes forward digital oilfield intelligent oilfield process.