Research On The Key Technology And Algorithm Of On-line Fault Diagnosis Of Power Transformer

Since the reform and opening up,with the rapid development of China's economy,the power demand of users surges,and the scale of power grid becomes larger and larger.Transformer is an important transmission and substation equipment in the power network,which is in a pivotal position and extremely expensive in the power network.However,due to the transformer's own structure and working environment and other factors,the failure of the core and winding accounts for more than 90% of the accidents,posing a serious threat to the stable operation of the power grid,and once the failure occurs,it will cause huge irreversible economic losses.It is an important method to ensure the reliability of power supply to conduct on-line monitoring of transformer and find out and eliminate the potential faults before the occurrence of faults.In this paper,the vibration principle of transformer winding and iron core is studied,and the measures to improve the axial stability of winding and control magnetostriction are given to reduce the vibration of winding and iron core.Then the vibration propagation of windings and iron cores is introduced and the feasibility of the vibration method in monitoring transformer windings and iron cores is discussed theoretically.In order to realize the on-line collection of transformer vibration signal,the electric transformer vibration on-line monitoring system is designed,and the vibration on-line monitoring system platform is developed by c # software.The system can collect the vibration signal on the surface of the transformer tank in real time,and the time domain waveform of the signal is displayed in real time and the simple signal analysis is carried out.The vibration signals at different measuring points of transformer are analyzed and studied based on frequency spectrum.The vibration signals collected at different phases of the same position and at different positions of the same phase are transformed into the frequency domain for comparative analysis,and the following three conclusions are obtained:(1)On the side of the same voltage and at the same measuring point,for the vibration signal of transformer with different phases,the fundamental frequency amplitude of the side phase is greater than that of the intermediate phase,and the amplitude of both sides is similar.(2)At the same voltage side,the fundamental frequency amplitude of the vibration signal at different measuring points of the transformer is greater on the upper side than on the lower side,but the frequency characteristics of the upper,middle and lower spectra are similar.(3)The higher harmonic components will be generated,and the higher harmonic components will be continuously attenuated and basically attenuate to 0 after 1000 Hz.Therefore,in the case of low precision requirements,the measurement points can be reduced according to the above conclusions,which can save costs and greatly reduce the workload of data analysis at the same time.In order to extract the characteristics of transformer vibration signal,this paper proposes a signal feature extraction method based on wavelet packet transform.This method can extract the signal features into numerical features,which lays a foundation for intelligent diagnosis of transformer mechanical faults.Finally,the improved particle swarm optimization neural network model is applied to transformer mechanical fault diagnosis.The experimental results show that the proposed method can effectively diagnose the mechanical faults of transformers,and its performance is better than the traditional diagnosis algorithm.