| Transformer is the key equipment in the whole power system,its healthy,safe and stable operation affects the whole power system.With the rapid development of China’s power industry,the voltage level and capacity of power grid are further increased,which puts forward higher requirements for the reliability of power transformers.Insulation aging has a direct impact on the healthy and stable operation of power transformers.However,the partial discharge of transformers generally occurs in insulation aging or manufacturing defects.Therefore,the main purpose of this paper is to evaluate the insulation aging status through the measurement and analysis of partial discharge phenomenon.Firstly,this paper studies several types of partial discharge(PD)in insulation aging,analyzes the similarities and differences of suspension discharge,corona discharge and surface discharge,and summarizes the main partial discharge detection methods in power market,which lays a theoretical foundation for the following research.Secondly,this paper studies the method of aliasing data separation based on fast independent component analysis(FastICA).It uses the method of fast independent component analysis to separate the overlapped partial discharge signals,and achieves good results.The practical analysis shows that the fast independent component analysis can effectively separate the partial discharge signals.Then,this paper uses improved particle swarm optimization(IPSO)to optimize the two parameters c and of support vector machine,and verifies through the classic data set.The experiment shows that the method has good separation effect and can be used for insulation aging classification.Finally,this paper uses the simulation data to verify the proposed method.The partial discharge signal is generated by using the partial discharge model built by Matlab/Simulink.After the signal is aliased,the fast independent component analysis is used to separate the signal,and the characteristic component is extracted to calculate the energy entropy,which is used as the input data of the classifier.The processed data are input into the classifier for insulation aging classification experiment,and the experimental accuracy reaches 78.33%,which proves that the pattern recognition classification method studied in this paper is effective. |
