Transformer asset management based on Markov Prediction Model utilizing health index

Transformers are considered among the important assets in a power system network, failures of which could lead to costly consequences. The degradation of transformers is a complex phenomenon that can be affected by multivariable factors. Accurate estimation of condition and life expectancy is quite difficult to be achieved due to complexity of aging mechanisms in transformers. Commonly, the prediction of transformer lifetime for asset management strategies is attained based on the probability of failure or failure hazard rate from the failure recorded data. However, the failure data is very limited and may not be available for the young population of transformers. Nowadays, the majority of utilities have implemented Condition Based Maintenance (CBM) as part of the asset management scheme. Under CBM, a single quantitative assessment known as Health Index (HI) is usually formulated to provide the overall condition of the transformers. Typically, HI is used to determine the present condition state of transformers and there is a potential to utilize HI for future states predictions. Due to the limitation of long-term records and also scattering data of HI, common mathematical approaches such as regression, trend line fitting, and extrapolation techniques are not suitable to determine the condition of transformers due to the overreliance on the data, which may affect the reliability of the predictions. This research presents a study on the application of stochastic probabilistic method known as the Markov Prediction Model (MPM) to estimate the future transformer condition states based on the transformer population HI. The prediction model was designed based on the MPM method that utilized two transition probabilities derivation techniques known as maximum likelihood (frequency of transition) and nonlinear least-squares minimization. The models were developed based on the utility oil condition monitoring data consisted of dissolved gases, oil quality and furanic compounds. Based on the computed HI, 120 transformers were arranged according to the corresponding states and the consecutive year, and the transition probabilities were determined based on the frequency of transition approach. The maintenance costs were estimated based on future-state distribution probabilities according to the developed MPM and the proposed maintenance policy model. In the second MPM model, the HI for 3195 oil samples ranging between 1 to 25 years of age was computed and the transition probabilities were obtained based on a nonlinear least-squares minimization technique. Further, the future performance condition curve of the transformers was determined based on the Markov chain algorithm. A statistical analysis was carried out to test the performance of MPM on the HI data. In addition, the impact of pre-determined maintenance repair rates on the HI data for transformers asset management strategies were performed through a sensitivity study using the updated MPM. Finally, the changes of the HI state distribution of the transformer population and the performance condition curve were discussed. The analysis on the relationship between the predicted and actual computed numbers of transformers using the MPM with the frequency of transition technique reveals that all transformer states are still within the 95% prediction interval. There is a 90% probability that the transformer population will reach ‘very poor’ state after 76 years and 69 years based on the transformer transition states for the year 2013/2014 and 2012/2013 respectively. Furthermore, the total maintenance cost based on the probability-state distribution increases gradually from Ringgit Malaysia (RM) 5.94 million to RM 39.09 million based on the transformer transition states for the year 2013/2014 and RM 37.56 million for the year 2012/2013 within the 20-year prediction interval. On the other hand, MPM with nonlinear least-squares minimization technique indicates that the method can be used to predict the future transformers condition states. The chi-squared goodness-of-fit analysis reveals the predicted HI for the transformer population obtained based on MPM concurs with the average computed HI along the years and the average error is 3.59%. Based on the case study, it is shown that the pre-determined maintenance repair rates can improve the HI state distribution and performance condition curve. The 30% pre-determined maintenance repair rate gives the highest enhancement and it is the most effective for the transformer population at ‘poor’ state.

Text FK 2019 85 - ir.pdf Download (1MB)

Actions (login required)

View Item