This paper presents a study to estimate future Health Index (HI) of transformer population based on Hidden Markov Model (HMM). In this paper, HI was represented as hidden state and the condition parameter factors in the HI algorithm namely Dissolved Gas Analysis Factor (DGAF), Oil Quality Analysis Factor (OQAF) and Furfural Analysis Factor (FAF) were represented as the observable states. A case study of 1130 oil samples from 373 oiltyped distribution transformers (33/11 kV and 30 MVA) were examined. First, the mean for HI in each year was computed and the transition probabilities for the condition data were obtained based on non-linear optimization technique. Next, the emission probabilities for each of the condition parameter factors were derived based on frequency of occurrence method. Subsequently, the future states probability distribution was computed based on the HMM prediction model and viterbi algorithm was applied to find the best optimal path sequence of HI for the respective observable condition. Finally, the predicted and computed HI were compared to the hypothesized distribution. Majority of the predicted HI agrees with computed HI. Predicted HI based on OQAF records the most accurate estimation throughout the sampling years. Inconsistencies are observed in year 2 and between year 7 and 10 for the predicted HI based on FAF. The predicted HI based on DGAF is in line with the computed HI during the first 2 years and deviates at the later stage of the sampling period.

Tenaga Nasional Berhad (TNB), an electric utility company in Malaysia has embarked on condition assessment and life management program for 707 units of in-service oil immersed free breathing distribution power transformers with on-load tap changer. The program consists of threetiered approach. The first tier involved transformer population screening by means of on-line diagnostic techniques through assessment of the insulating oil, physical, thermal and operating performance. The second tier focused on off-line diagnostic testing on few identified transformers from the fleet using basic electrical testing techniques. The third tier involved more advanced diagnostic techniques namely frequency response analysis and partial discharge test. Results of the condition assessment on individual transformer were analyzed and transformer health index was formulated to provide engineering solutions and asset management direction.

This paper presents a study to estimate individual condition parameters of the transformer population based on Markov Model (MM). The condition parameters under study were hydrogen (H2), methane (CH4), acetylene (C2H2), ethylene (C2H4), ethane (C2H6), carbon monoxide (CO), carbon dioxide (CO2), dielectric breakdown voltage, interfacial tension, colour, acidity, water content, and 2-furfuraldehyde (2-FAL). First, the individual condition parameter of the transformer population was ranked and sorted based on recommended limits as per IEEE Std. C57. 104-2008 and IEEE Std. C57.106-2015. Next, the mean for each of the condition parameters was computed and the transition probabilities for each condition parameters were obtained based on non-linear optimization technique. Next, the future states probability distribution was computed based on the MM prediction model. Chi-square test and percentage of absolute error analysis were carried out to find the goodness-of-fit between predicted and computed condition parameters. It is found that estimation for majority of the individual condition parameter of the transformer population can be carried out by MM. The Chi-square test reveals that apart from CH4 and C2H4, the condition parameters are outside the rejection region that indicates agreement between predicted and computed values. It is also observed that the lowest and highest percentages of differences between predicted and computed values of all the condition parameters are 81.46% and 98.52%, respectively.

This paper presents an investigation of the condition state distribution and performance condition curve of the transformer population under different pre-determined maintenance repair rates based on the Markov Prediction Model (MPM). In total, 3195 oil samples from 373 transformers with an age between one and 25 years were tested. The previously computed Health Index (HI) prediction model of the transformer population based on MPM utilizing the nonlinear minimization technique was employed in this study. The transition probabilities for each of the states were updated based on 10%, 20% and 30% pre-determined maintenance repair rates for the sensitivity study. Next, the HI state distribution and performance condition curve were analyzed based on the Markov chain algorithm. Based on the case study, it is found that the pre-determined maintenance repair rates can affect the HI state distribution and improve the performance condition curve. The 30% pre-determined maintenance repair rate gives the highest impact, especially for the transformer population at state 4 (poor). Overall, the average percentage of change for all HI state distributions is 16.48%. A clear improvement of HI state distribution is found at state 4 (poor) where the highest percentage can be up to 63.25%.

No abstract