| Gas Insulated Switchgear(GIS)has high reliability.With the acceleration of urbanization and electrification,GIS has been widely used in power grid.However,due to the limited production process,improper operation,and uncontrollable factors in long runs,PD phenomenon of GIS occurs from time to time.In severe cases,insulation breakdown may occur,leading to power grid failure.Therefore,it is necessary to timely diagnose Partial Discharge fault of GIS equipment.The main content of this paper is pattern recognition of PD signal of GIS equipment.Analyzing the existing methods,this paper adopts the PD pattern recognition method based on the combination of AVMD and RF.Using the Artificial Fish Swarm Algorithm(AFSA)to optimize the VMD algorithm to decompose the PD signal and obtain all IMF components.Feature sets are constructed by extracting features from time domain,frequency domain,IMF energy entropy and sample entropy,etc.,and the RF classifier is input for training,prediction and classification,and the PD type is judged.The details are as follows:(1)PD signal feature extraction.Analyzing the existing methods,this paper proposes a signal processing method based on AFSA-VMD algorithm.The simulation model and measured data are used to carry out experiments.By comparing with EMD algorithm,the superiority of VMD algorithm in overcoming mode aliasing is proved.The effects of decomposition layer K and penalty factor α on VMD algorithm are analyzed.Aiming at the current VMD parameter selection problem,the minimum envelope entropy is used as the fitness function,and the AFSA algorithm is used to achieve adaptive VMD decomposition.The simulation signals are decomposed by PSO-VMD and AFSA-VMD.The results show that the PSO-VMD is easy to fall into local optimal,and the running time is longer,while AFSA-VMD has a shorter running time.The IMF component decomposed by AFSA-VMD is consistent with the signal component contained in the original signal.The time-domain and frequency-domain features are extracted,and the sample entropy and energy entropy of IMF are calculated,which are input into RF classifier as feature parameters to realize PD signal pattern recognition.(2)Constructing the classifier model.Aiming at the problem of low accuracy of single classifier in fault diagnosis,RF classifier is used for PD pattern recognition.In order to verify the effect of RF algorithm on PD pattern recognition,BPNN,SVM,KELM and other algorithms are used for experimental comparison.The results show that AVMD-RF has the highest recognition rate and the time is short. |
