• P-ISSN 0974-6846 E-ISSN 0974-5645

Indian Journal of Science and Technology


Indian Journal of Science and Technology

Year: 2021, Volume: 14, Issue: 39, Pages: 2971-2981

Original Article

Analysis of 11kV/ 430 V 500 kVA Transformer Dissolved Gas using Pre-processing Techniques through Duval’s Triangle

Received Date:12 October 2021, Accepted Date:06 November 2021, Published Date:25 November 2021


Objectives: Analysis of dissolved gas in transformer oil is very important because the fluctuations in the oil will change the process of content of dissolved gases which will show effect on many factors such as oil temperature and external environment. Hence, it is important to propose a preprocessing technique through Duval’s triangle which is important step in the machine learning process for predictive maintenance of transformer. Methods: The dissolved gas content for 500 kVA transformer oil based on the training and testing data for fault free condition taken for the period of 5 years (2013-17) and faulty condition taken for 3 years (2018-2021), has been evaluated using Duval’s triangle method to express gas concentrations of CH4, C2H2 and C2 H4 in ppm as percentages of their total. High pass filter, scaling and windowing techniques are used for preprocessing the giant DGA data. Findings: The results show that unusual and undesirable intensity in the CH4, C2H2 and C2 H4 gasses is eliminated by using proposed high pass filter technique. More visible output signal by the way of reducing its dimensionality is obtained by applying proposed scaling and windowing techniques for the filtered DGA data. Novelty: Most of the researchers have used limited samples of DGA data for condition monitoring due to which diagnosis of transformer faults is not accurate. To accurately diagnose transformer faults, large set of training and testing data is required. In this paper, the pre-processing of giant DGA data consisting of 960 samples of fault free training with data size of 480000 x 55 and 500 samples of fault free testing with data size of 250000 x 55, faulty training and faulty testing with data size of 960000 x 55 is presented. The preprocessing method adopted will be used in real time application to provide suitable data to train the advanced deep neural network like LSTM through MATLAB interfacingwith graphical processor unit. This is required for condition monitoring and accurately predicting the faults in above mentioned transformer.

Keywords: Dissolved Gas Analysis; Duval’s Triangle method; Predictive maintenance; Data Ensemble; Pre-processing; Scaling; Windowing


  1. Mharakurwa ET, Nyakoe GN, Akumu AO. Power Transformer Fault Severity Estimation Based on Dissolved Gas Analysis and Energy of Fault Formation Technique. Journal of Electrical and Computer Engineering. 2019;2019:1–10. Available from: https://dx.doi.org/10.1155/2019/9674054
  2. Alqudsi A, El-Hag A. Application of Machine Learning in Transformer Health Index Prediction. Energies. 2019;12(14):2694. Available from: https://dx.doi.org/10.3390/en12142694
  3. Kabir F, Foggo B, Yu N. Data Driven Predictive Maintenance of Distribution Transformers. 2018 China International Conference on Electricity Distribution (CICED). 2018. doi: 10.1109/CICED.2018.8592417
  4. Witten IH, Frank E. Data Mining: Practical Machine Learning Tools and Techniques (4). Morgan Kaufmann..
  5. Rogers R. IEEE and IEC Codes to Interpret Incipient Faults in Transformers, Using Gas in Oil Analysis. IEEE Transactions on Electrical Insulation. 1978;p. 349–354.
  6. Data Analytics cases for Asset Awareness. Electric Power Research Institute . 2019.
  7. Kumar R, Markam K. Performance Analysis of Kaiser-Hanning Window For Digital Filter Design. Journal of Emerging Technologies and Innovative Research. 2018;5(10). Available from: https://www.jetir.org/papers/JETIR1810880.pdf
  8. Ghosh S, Dutta S. Ensemble Machine Learning Methods for better Dynamic Assessment of Transformer Status. Journal of The Institution of Engineers (India): Series B. 2021;102(5):1113–1122. Available from: https://dx.doi.org/10.1007/s40031-021-00599-1
  9. Zhang X, Wang S, Jiang Y, Wu F, Sun C. Prediction of dissolved gas in power transformer oil based on LSTM-GA. IOP Conference Series: Earth and Environmental Science. 2021;675(1):012099. Available from: https://dx.doi.org/10.1088/1755-1315/675/1/012099


© 2021 Srirama Sarma et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Published By Indian Society for Education and Environment (iSee)


Subscribe now for latest articles and news.