Study Of Condition Monitoring And Comprehensive Diagnosis For Middle-Voltage And Low-Voltage Switchgear

The middle voltage(MV) and low-voltage(LV) switchgear, as one of the important electrical equipments in distribution network, is responsible for the opening, closing, control and protection of power lines. It's safe operation directly affects the quality of people's work, study and life. In order to improve the running reliability, reduce the fault risk, and decrease the economic losses, monitoring equipments can be installed in the switchgear to monitor its running parameters. And by analyzing and processing the parameters, the running status detection and fault diagnosis can be achieved to ensure that the staff can detect problems in time and get the decision basis. Therefore, for the common faults of switchgear, monitoring and analyzing the features to realize the fault detection and classification of switchgear is of great significance. The main contents are as follows:(1) The scheme to monitor features in switchgear is designed, and the fault feature extraction and selection for switchgear based on the cloud sample entropy and improved Laplacian score are proposed. From the features reflecting different types of faults such as insulation, mechanical, arc and so on, the original feature set for switchgear fault can be achieved. The multivariate multiscale cloud sample entropy(MMCSE) fault features of switchgear can be obtained by using the lower half-trapezoid cloud model to quantify the similarity of composite delay vectors of features time series and softening the similar tolerance criterion of multivariate multiscale sample entropy. Two kinds of score with local preserving and global separation is weighted by introducing a weight coefficient and the improved Laplacian method is formed to sort the importance level of fault features, then the optimal fault feature subset of switchgear is obtained.(2) According to the differences of faults, fault comprehensive diagnosis method for switchgear based on Mahalanobis distance and fuzzy support vector machine(FSVM) is proposed. The fault alarm of switchgear is implemented by using Mahalanobis distance to quantify the similarity of fault features and standard samples. This thesis uses the piecewise half-trapezoid cloud model to quantify the uncertainty of the relationship between fault samples. And the regional difference and dispersion of sample space are synthesized to calculate the sample membership, then the classification method based on FSVM is formed and it can identify different fault types of switchgear. Through case analysis on the monitored data, the correctness of the proposed scheme is validated.(3) An online monitoring and comprehensive diagnosis system of switchgear is designed and developed. The overall structure of system consists of three hardware modules and one software module, and the multi-layer data fusion gateway is designed for data communication between modules. The parameters of switchgear such as basic electrical quantities, the temperature of electrical equipment junction and the temperature and humidity of environment are collected by three hardware modules, and the software can use the parameters to realize the data display, query, fault alarm, classification and so on. The system has been running for a year in a switching station, which preliminarily tests and validates its functions and characteristics.