Studies On On-line Monitoring And Fault Diagnostics For Mechanical Condition Of On-load Tap Changers

On-load tap changers (OLTC) operational within power transformers are extensively used in power systems to achieve voltage stabilization in load centres, regulation of the reactive power flow and flexibility in power dispatch, whose principal malfunctions or failures are of mechanical type. Hence, on-monitoring of the mechanical characteristics of OLTCs is of utmost significance for a safe, stable and reliable operation of the power systems.With reference to the state-of-the-art development in mechanical condition monitoring and fault diagnosis of OLTCs and the pending key technological issues, systematic research work has been done in this dissertation which mainly concentrates in the following several aspects:Based on the non-invasion sensing technology, monitoring methods for multiple signals are studied without affecting the normal operation of transformers, which leads directly to the formation of an on-line monitoring hardware system for OLTCs. From the chaotic dynamics point of view, feature extraction method is investigated which can reflect the feature variance of different vibration modes with OLTC operations, and the corresponding fault diagnosis strategy is also established. The Hidden Markov Model (HMM), normally utilized in the speech recognition domain, is introduced into the mechanical condition monitoring and fault diagnostics of OLTCs, and a unique fault diagnosis methodology is thereby developed from a novel view angle. With analysis and elucidation in details to account for the prevailing faults as well as their dominant mechanisms for a SYJZZ-type OLTC, a synthesized expert system is constructed for mechanical fault diagnosis of the OLTC. Finally, the on-line monitoring software system is developed based on graphical programming and the virtual instrument technology.The main conclusions based on the above mentioned investigations are drawn as follows:The non-invasion scheme and the multi-sensor measurement technique are applied to design an on-line monitoring system for OLTCs' mechanical characteristics. Without affecting the normal operation of transformers, the on-line monitoring system will readily acquire the multiple signals to reflect the operational conditions of the equipment, which facilitates the mechanical feature extraction diverse and much effective.On-line monitoring signals from both a rotating angular sensor and a current transformer attached to the drive motor operational in an OLTC, which indirectly reflect the dynamic changes of the rotating angles and moments, can effectively indicate the underlying deficiency or even potential faults of the mechanical drive system. With combination of several applied algorithms including first-order derivative, instantaneous frequency tracking and wavelet analysis, a feature extraction methodology is presented to evaluate the mechanical condition of the OLTC's drive mechanism, which provides an effective alternative reference to realize fault diagnosis.With chaotic dynamics analysis to the mechanical vibration modes of OLTCs' operations, a geometric feature extraction scheme is developed for phase portrait structure based on a multi-dimensional coordinate space. The fuzzy set theory is further introduced into OLTCs' mechanical fault diagnosis, with a view to establishing the normal distribution membership function of every condition's fuzzy set. Regarding two sample categories, i.e. single analyte sample and analyte sample sets, both principles of the largest membership and the optional nearly are used for fuzzy reasoning to finally realize mechanical fault diagnosis of OLTC.By partition, normalization and vector quantization of the power density spectrum of the vibration signals, a feature vector extraction methodology is presented for the discrete spectrums, which can to the farthest extent retain the unique features and difference of all the mechanical modes, and also well meet the requirement for HMM exemplar training. With the sampled data from both experimental simulations and on-site measurements, a trained HMM norm modes database for fault diagnosis is established based on the different mechanical conditions of the OLTC. Large quantities of verifications demonstrate that, the new proposed HMM-based mechanical fault diagnosis scheme for OLTC is feasible and applicable, with outstanding behavior for fault classification.Based on analysis of prevailing faults and their dominant mechanisms specifically for the SYJZZ-type OLTCs, the fault feature changes of the on-line monitoring signals are summarized. Further, an expert system for synthesized mechanical fault diagnosis of OLTCs is established based on a decision tree and integrated analysis of the monitoring signals. The decision tree is preferable for dealing with non-numerical data and rapid reasoning, in addition, multi-angle analysis and synthesized fault diagnosis can be achieved through combination of every diagnosis subprograms and rpocedures.A synthesized and internet-based "on-line monitoring and fault diagnosis system for OLTCs' mechanical characteristics" has been successful developed, and has found practical applications in Shandong power grids. The monitoring and diagnosis system consists of both the hardware and the software subsystems, which has also realized remote diagnosis and "virtual hospital" based on the local area network. In the hardware subsystem design, a practicable installation method is developed to ensure operational stability of the rotating angular sensor, and effectively improve the measurement accuracy as well as extend the sensor's life expectancy. The on-line monitoring software subsystem is developed in a modular architecture with excellent transportability and expansibility, which also possesses manifold functions such as data acquisition, analysis, storage, print, network transmission and so on.