| Disconnector is the most essential piece of primary equipment in a substation,and its operation with a low defect failure rate and high health directly affects the stability and dependability of the entire power grid.Due to the unfavorable natural environment,power grid load fluctuation and impact,manufacturers’ manufacturing process,and other objective factors,disconnectors are susceptible to equipment defects.Presently,the commonly employed technical means of fault diagnosis in production are limited,and the characteristic data extracted on-site cannot reflect the equipment fault condition in a timely and accurate manner.Therefore,substation inspection personnel can only detect defects of a serious nature.Disconnectors account for 21.3% of all equipment in Yunnan Northeast Power Grid,and their health has a direct impact on the grid’s ability to operate safely.In this paper,the motor current of the disconnector serves as the research object,and a deep learning is used to diagnose the common fault types with high frequency.The primary work comprises the following three components:1.Design the technical scheme for detecting the characteristic quantity of an electric disconnector.This paper examines the selection and analysis of research characteristics,the selection of test objects,the analysis of equipment failure types,and the selection of testing equipment individually.Finally,a number of test schemes are compared,and an initial test scheme is determined.The results indicate that the characteristic amount of motor current is closely related to disconnector failure.By matching data acquisition equipment with design parameters,it is possible to collect the characteristic quantity data of various typical states of the equipment,which can serve as the test data set for subsequent work.2.This paper adopts envelope analysis to preprocess the data,extracts the envelope curve data that reflects the distribution law of motor current variation under different working conditions,and performs descriptive analysis on the envelope curve data in order to address the issues that the motor current data set collected on site contains interference factors and it is difficult to extract characteristic quantities.Under the same category of equipment,the preliminary descriptive analysis reveals that the change states of characteristic data are highly convergent.The DTW algorithm is then used to analyze all samples further.The results indicate that the characteristics of current in the same working condition of the equipment are highly convergent,whereas different fault states will reflect glaring differences in the characteristics of current.The disconnector motor’s current signal can be utilized as an effective characteristic quantity for dependable equipment fault diagnosis.3.This paper designs an outdoor electric disconnector fault diagnosis algorithm based on a DBN network,and conducts a comparative experiment using SVM and BP algorithms.The aim is to address the problem of the disconnector fault diagnosis network model’s low accuracy.To further evaluate the algorithm’s viability and enhance its performance,we propose an enhanced FS-DBN network and introduce the Fast shapelets(FS)time series pruning method.The comparison demonstrates that the fault diagnosis performance of the FS-DBN algorithm proposed in this paper is superior,and that it can effectively mine the characteristics of fault current data sets and make accurate diagnoses. |
