Application of wavelet cumulative energy and artificial neural network for classification of ferroresonance signal during symmetrical and unsymmetrical switching of three-phases distribution transformer

In the case of the presence of ferroresonance in distribution transformer due to a faulty switching operation, ferroresonance signals should be discriminated among its initiations due to opened single-phase, opened two-phases, and opened three-phases, so that ferroresonance mitigation can be conducted appropriately. However, the performance of mitigation system itself is highly determined by its accuracy in classification of such ferroresonance signals. This paper dealt with the application of wavelet cumulative energy as input of artificial neural network (ANN), that was feed-forward backpropagation network. Ferroresonance was initiated by varying grading capacitance of circuit breaker and switching operations. The fifth order of daubechies wavelet transform up to nine levels was applied to the secondary voltage of transformer. The detail signal at ninth level decomposition was then calculated its cumulative energy for the input of ANN. The ninth level detail signal and its cumulative energy showed that the ferroresonance signals were clearly distinguished between opened single-phase, opened two-phases, and opened three-phases. The ANN output also performed the satisfactory classification result.