Based on dimensionality measurement, the Thesis puts forward a multi-kernel SVM fault diagnosis method for power transformer, defines the transformer dimensionality measurement, designs the dimensionality measurement system and workflow and then utilizes the features of multi-kernel SVM to map the input fault features from the original data space to high-dimensional data space and utilizes Lagrange method to achieve the dual solution of primal problem and achieve the multi-kernel SUVM fault diagnosis of power transformer. Meanwhile, during the fault diagnosis of transformer, it is difficult to obtain the fault sample, so how to achieve effective promotion of recognition precision under small sample is a very meaningful research. From the experiment result, it can be seen that MSVM fault diagnosis process applied can achieve the high-precision recognition under small sample and synchronously achieve the promotion of recognition efficiency, which shows the performance advantage of the method applied. The algorithm above adopts matlab modeling method in the realization process. How to design the algorithm with specific controller, to set up the real operating platform and verify the algorithm under the real environment is the research focus in future.
展开▼
