| Summary: | The development of electric power industry is oriented on high reliability, flexibility and efficiency of managing power grids of arbitrary configuration. For such grids an information infrastructure is required. It should consist of various software and hardware, including systems of electrical equipment monitoring and diagnostics for accumulating information about its parameters with controlling and managing its condition. As a rule, data about the electrical equipment is heterogeneous. Thus there is a necessity of certain mechanisms for data processing to provide a possibility of constructing diagnostic models in an automated mode and adapting them to power grid operating conditions. The aim of this work is to develop a mechanism for automated calculation of the electrical equipment diagnostic models parameters. It supposes using historical data analyzing algorithms that ensure high reliability of the diagnosis. Base on this the application of fuzzy clustering for constructing membership functions to assess the features of equipment condition change in fuzzy diagnostic models was considered. Different fuzzy clustering algorithms were analyzed, and a technique for processing data on the equipment operation with constructing membership functions based on fuzzy partition matrix and clusters centres was proposed. The technique allows to approximate the membership functions by typical curves and to choose the most effective variant of clustering in terms of electrical equipment diagnostic reliability. The testing of fuzzy models for assessing the condition of power transformer equipment using clustering results was performed. A high level of compliance of simulation data with the conclusions of specialized organizations performing monitoring of equipment in power supply systems for oil production facilities was obtained. The practical relevance of the results is in using the technique in the synthesis of intelligent expert-diagnostic systems for increasing the electrical equipment diagnostic reliability and reducing the duration of its unplanned downtime.
|
|---|