Research Of Magnetizing Inrush Identification Method In Transformer Based On Wavelet And Neural Network

With the development of the domestic economy, people's increasingly need for electric power raised. Since the scale of the electric power system is enlarged continually, some new requirements of large transformer protection are presented. Differential relay is the master protection for transformer all along, but the magnetizing inrush current which brings out when the transformer suddenly closed with no load caused miss-operation of the differential relay, so to discriminate the magnetizing inrush current correctly is still one of the most key and difficult problems on guarantee for credible operation of the magnetizing inrush current for transformer. The reliability of the protection would be enhanced if the magnetizing inrush current can be discriminated. So, this thesis analyzed and compared some discriminating methods applied in practice or described in related literatures at present. After considering many factors comprehensively, a method to the discriminate magnetizing inrush current based on wavelet analysis and BP neural network is proposed.The magnetizing inrush current is a large impulse current which brings out when the transformer suddenly closed with no load. This thesis analyzed the cause for its produce and the characteristics, and discussed the cause for fault current's produce and its characteristics. In addition to this, it also researched the methods to discriminate the magnetizing inrush current and the fault current.The wavelet transform has the characteristics of multi-scale and local time-frequency, it is particularly well adapted to process the sharply changing signals and to extract characteristics of these signals. When a smoothing function's first derivative is adapted to the wavelet function, the modulus of wavelet transformed signal gets the local maxima at the sudden change point. In this thesis, these characters are applied to measure the magnetizing inrush current angles of the three-phase transformer.The Artificial Neural Network not only has the high adaptability and self-structure and very strong rash-stick and wrong- tolerance ability, but also can close to non-linear relation of complex adequately, so it is adapted to discriminate the magnetizing inrush current and fault current. Firstly, a forward three-layer BP neural network of dynamicstructure is established, then, second harmonic component ratio and dead angles of two phase inrush's dispersion in three-phase transformer are taken as input models. Secondly, using the improved BP network training algorithm, it makes the model of the network's structure gain optimization by model training. The whole results from the experiment shows that the NN model designed in this thesis can discriminate the magnetizing inrush current and fault current successfully and quickly, thus the superiority of this method is revealed.