Study Of UHF PD Monitoring And Knowledge-Based Analysis Method For Transformers

The power system is growing into ultra high voltage and super large capacity, so the voltage level and capacity of a power transformer is improving as well. If the accident of a power transformer occurs, it will result in large area blackout in power systems. The stable operation of the power system may even be destructed, thus it will bring huge economic losses. Statistical analysis shows that insulation fault is the main reason to the transformer failures, and partial discharge (PD) is an important sign of insulation deterioration. The insulation condition assessment can be performed by online PD monitoring, so PD monitoring becomes a research focus in recent years. One difficulty of PD monitoring is interference suppression. Owing to the high frequency band of ultra high frequency (UHF) detection method, the external interference in the field can be suppressed very well. Therefore, it is very meaningful to carry out online UHF PD monitoring for transformers.However, as the pulse sources of PD monitoring are complex, most PD monitoring systems have deficiencies in the following aspects: pulse knowledge acquisition, online pulse recognition, pulse raw data storage and pulse source analysis. Since the operating conditions of the transformer are not considered in most PD systems, the reliability of PD analysis results is reduced. In addition, high speed oscilloscope was utilized to perform data acquisition in most PD measurements. The limitations can be summarized as: (a) the online memory is small; (b) the sampling time slot is long; (c) low responsiveness to transient pulses caused by over-voltages.To address these issues, the integrated monitoring and comprehensive analysis of transformer PDs is performed under the integrated condition monitoring and diagnostics framework of HOMS (Hydropower plant Optimal Maintenance System). The PD activities of 4 power transformers in Gezhouba Hydro Power Plant are studied using UHF technique.According to the problem of current PD monitoring systems, the integrated monitoring method of transformer PDs is proposed. An online UHF PD monitoring system for transformer is developed under the integrated framework of HOMS. The system is composed of internal UHF sensor installed in the oil tank, envelope detector, high speed data acquisition and real-time data analysis etc. Due to the integrated monitoring, the operating environments and operating conditions of the transformers have been acquired synchronously. Thus the association analysis of PDs with respect to operating conditions can be performed.Pulses recognition and comprehensive analysis must be established on the basis of prior knowledge of the onsite pulses. A large quantity of pulses from the 4 transformers have been continuously collected and organized for years. A sliding window method is employed to determine the boundaries of individual pulses. Thus the sampling data is converted into pulse-time series. The feature identification method of these onsite pulses was studied with the aid of the platform of HOMS. Different types of pulses are mapped to different feature spaces. The feature knowledge of onsite pulses was acquired; furthermore, the pulses knowledge base (PKB) was built. The PKB can be used for online pulses recognition and comprehensive analysis of the pulses.According to the problem of pulse raw data redundancy and typical pulse knowledge missing, a knowledge-based pulses recognition method is proposed. Based on the PKB a recognition tool using hierarchical sub-classifiers is developed. Linear and nonlinear discriminant functions are used to divide the feature spaces into regions. The multi-class problem is converted into multiple small-class problems. Therefore curse of dimensionality, the well-known difficulty in pattern recognition, has been overcome using this method. Under the integrated framework of HOMS, the method of comprehensive analysis of PDs is proposed based on the integrated PD monitoring. The comprehensive analysis of transformer PDs is performed as follows: (a) periodical PDs caused by power frequency voltage under stable operating conditions; (b) the PDs caused by transient over-voltages under unstable operating conditions; (c) special analysis of transient pulses induced by over-voltages. The characteristics of PD activities with respect to relevant states are studied with the aid of the integrated framework of HOMS.At present, the system has been applied to 4 power transformers, and the reliability of PD analyzing results is verified by the analysis of large mount of onsite data.