Dealing With Data Uncertainty for Transformer Insulation System Health Index

Abstract

Health index has been widely accepted as a powerful tool for monitoring the condition of power transformer insulation system based on various diagnostic parameters. While this approach has been extensively discussed in the literature, not much attention was given to provide effective solutions to the uncertainty in the used data. According to CIGRE 761, data quality issues may arise due to measurement accuracy as well as incompleteness and unavailability of the required data. Therefore, this article presents the implementation and evaluation of a certainty level model for transformer insulation system health index to deal with data uncertainty. The impact of data unavailability on the health index results is also investigated. Certainty level of the health index is determined by the criticality level of available data, and is reported along with the health index result. A method to handle unavailable data by predicting the oil interfacial tension (IFT) using Random Forest approach is also presented. The proposed certainty level model is designed to accommodate the predicted value of missed data into the health index model while considering its prediction accuracy. The robustness of the developed model is validated through its application in assessing the health condition of six in-service power transformers. The results indicate that by including the proposed certainty level and the prediction approach to eliminate the issue of uncertain and missed diagnostic data, an asset management decision can be taken on operating power transformer fleets with high level of confidence.

Review of Transformer Health Index from the Perspective of Survivability and Condition Assessment

“…In cases when condition data are not complete, or even when some are unavailable, neither weighted-score-based methods nor those based on ANN, fuzzy logic or regression methods are capable of adequately calculating the health index of the transformer (fleet) of interest. Under such circumstances, probabilistic-based techniques like Monte Carlo simulation and Bayesian belief networks are superior for handling the missing information and uncertainties in transformer health index calculation [49,[95][96][97][98][99][100].…”

Connective Action and Digital Activity of The Cipayung Group in Protesting The Increase of Fuel Price

“…It will be easier for the government to get closer to the people as a result of the development of social media. The public will be encouraged to participate in democratic practices with actors, including decision-making processes regarding the public interest on certain issues (Prasojo & Abu-Siada, 2021).…”

An intelligence expert system for quality health monitoring of power transformers using testing attributes

“…This system includes the implementation of continuous condition monitoring and regular maintenance of power transformers. This results in efficiency improvement, service life extension, and optimal performance of transformers [3,4]. In general, a noticeable part of failures in power transformers is primarily caused by their insulation failure [4][5][6][7][8][9].…”