Multi-criterion analysis-based artificial intelligence system for condition monitoring of electrical transformers

Abstract

In this present paper, a novel multi-criterion-based fuzzy logic (FL) expert system using different membership functions (MFs) is proposed to determine the overall health index (OHI) of electrical transformers. 30 oil samples from different field transformers installed at various locations in Himachal Pradesh, India, are collected for the analysis and various diagnostic tests are conducted on each of the oil samples. The diagnostic testing data are utilised for the proposed methodology. Initially, the diagnostic data are normalised using the well-known multi-criterion analysis (MCA) method. The normalised input data are grouped into three grades, ie total dissolved combustible gases (TDCGs), oil insulation and paper insulation. Furthermore, a fuzzy logic model is designed based on the three different grades. Output health indices are determined for each of the samples. Comparison and validation of the proposed model is conducted with the expert model, as well as the preknown health status of 150 transformers installed in the Gulf region. The expert model is designed with a trapezoidal membership function, whereas the proposed model considers the popular Gauss-2. From the comparison, it is observed that the accuracy of the proposed model is 98%, while the accuracy of the expert model is 96%, making the proposed model more accurate. Moreover, a plan of action for proper maintenance is also recommended for each transformer, based on the evaluated health index. The proper maintenance of transformers leads to improvements in their service life. The present work is beneficial not only for transformer utilities but also for customers. The model is straightforward to understand, even for inexperienced staff and maintenance managers.