Model-aided Diagnosis for High Voltage Circuit Breakers

ENGLISH) Motivated by economic pressures, power utilities are increasingly looking for new ways of reducing their costs while maintaining the quality of supply. A promising concept is predictive or condition based maintenance of high voltage equipment, based on continuos knowledge of each device’s condition. Power transformers and high voltage circuit breakers (HVCBs) have been the focus of attention in recent years. However, as a survey of available literature indicated, most of the presently available condition monitoring systems for HVCBs offer only limited functionality in providing users with the information they need. Therefore, research was done into online condition monitoring and diagnosis for HVCBs, with the goal of creating an intelligent system that would perform true diagnosis, i.e. yield a clear statement of the location, nature, and criticality of a detected problem. Such a system starts with various sensors on the breaker and ends at the human-machine interface, where the user is presented with the computation results, with various data processing and transmission stages in between. Literature shows that much progress has been made in each individual discipline, but combining the different state-of-the-art technologies into a comprehensive HVCB condition assessment system remains yet to be done. As a basis for further work, statistics of problems with HVCBs were compiled from failure records of a manufacturing company. The results do essentially agree with other published statistics, in particular that by CIGRE Working Group 13.06. Again it was confirmed that the greatest number of failures originate in the operating mechanism, in particular due to hydraulic or pneumatic leaks. The operating mechanism is therefore the place to start with condition monitoring.—Other findings indicate that most failures could have been detected by a suitable monitoring system, using mostly sensors that are readily available today. Many failures could have been predicted, too, thereby allowing the operators to perform maintenance before the problem became serious. With these results in mind, a computer model of a HVCB was created as a basis for a diagnostic system. It was assembled from models of the individual circuit breaker components, with particular attention to the operating mechanism. This modular construction facilitates adaptation to different breaker types and technologies. Patterned after a specific exemplar of HVCB, the model was tuned to give an optimum representation of its behavior under various circumstances and is therefore suitable for use in a model-based system. This model was used for implementation of a novel diagnosis strategy, named “Model-Aided Diagnosis” (MAiD). It was developed with the aim of providing diagnostic ability with good flexibility at low computational cost, in order to make it suitable for upgrading existing condition monitoring devices.— MAiD is a combination of the model-based with the case based approach (both known from literature), featuring certain strengths of both while avoiding certain shortcomings. Its main advantage is that the diagnosis itself requires only little computing power and can therefore be used even on a cost-sensitive device such as a HVCB. Of course, it can also be applied to various other kinds of technical devices. A simple test implementation of MAiD was created, using the circuit breaker model described above. The diagnostic module was tested with simulated data as well as actually measured data on a real circuit breaker in the laboratory, where certain artificial “faults” were introduced into the circuit breaker under observation. In every case, the fault was diagnosed correctly, with no false alarms. This demonstrates the ability of MAiD to reach correct conclusions under a variety of circumstances and is promising for future field installations. Finally, suggestions for integration in an overall monitoring and control system and for possible improvements on the basic MAiD strategy are given. With that, a MAiDbased HVCB condition assessment system can be realized at reasonable effort.