The transformer winding deformation based on finite element method and regular limit learning machine

The frequency response method is characterized by its stability, efficiency, lossless, economy and simplicity so it is widely used in transformer testing project. The frequency response method mainly uses the comparison of the amplitude spectrum of transfer function to determine the winding state. Subsequently, correlation coefficient and mean variance value concept were introduced successively. Frequency response analysis method with simple light test equipment, test sensitivity, visual experimental spectrum analysis, the data quantity analysis has the advantages such as comparability, so this article use the frequency response method to test the winding deformation. After the detection of transformer winding deformation by frequency response method, it is necessary to identify the deformation type, in this paper, the frequency response characteristic curve is quantized by using the limit learning machine, and the transformer winding deformation category is judged based on the unified standard. Thus minimizing the human error of the staff. Firstly, the capacitance and inductance parameters of the equivalent circuit are calculated by finite element method. Then the transformer winding deformation is detected by frequency response method. Finally, the identification of deformation type is completed by the ultimate learning machine.