Design of a fault diagnostic engine for power transformer using data mining

Abstract

The power transformer is one of the main components in a power transmission network. Major faults in these transformers can cause extensive damage which does not only interrupt electricity supply but also results in large revenue losses. Thus, these transformers are needed to be routinely maintained. Due to the large number of transformers of different makes and capacities, routine maintenance and diagnosis of such transformers are rather difficult as different transformers exhibit different characteristics and problems. Moreover, different climatic and operating conditions may not be able to draw correct conclusion to some problems. In Malaysia, the lack of local expertise makes dependency on foreign consultants imminent which are rather expensive. To help in overcoming such problems, a Software for Intelligent Diagnostics of Power Transformers known as ADAPT, using the technique of fuzzy logic is developed in this study. The technique allows the interpretation of the Dissolved Gas Analysis (DGA) to be performed routinely on the transformers. In order to ensure that all the transformers are diagnosed and maintained properly, a new intelligent diagnostic architecture known as Total Intelligent Diagnostic Solution (TIDS) has been developed to improve the diagnosis accuracy of the conventional DGA approaches. The TIDS structure has a main interpretation module which consists of Fuzzy TDCG and Fuzzy Key Gases and a supportive interpretation module which consists of Fuzzy Rogers Ratio and Fuzzy Nomograph. The TIDS structure is incorporated into the ADAPT software which allows for multiple diagnostic methods to reach an ultimate outcome especially when verified by four methods. This new architecture leads to the diagnostic of a wider range of transformer fault types and provides a more detail information about the transformer condition, thus help to reduce maintenance costs, prevent unnecessary force outages and avoid explosion danger.