Transformer liquid dielectric is seen as a critical quality characteristic which must be very strong to withstand the voltage level in an abnormal state. In this research, prediction of dielectric strength, fire point and viscosity through response surface methodology (RSM) and adaptive neuro-fuzzy inference system (ANFIS) is developed to improve liquid dielectric life in the automated environment. An experimental design with RSM-based box–behnken method was planned and experiments with three distinct input parameters (Wt.% CNT, Wt.% ZrO₂ and Stirring temperature) were carried out. The set of data has been used to train and analyse using hybrid and back-propagation optimization methods. Model values of ANFIS were compared with experimental and RSM values. The comparison results showed that the developed ANFIS model is a powerful tool for modelling the nano fluids-based mixed insulating fluids.

The objective of this work is to increase the performance of the mixed insulation oil with nanoparticles. To improve the performance, the Mineral oil (MO) is blended with Pongamia Pinnata Oil (PPO) and suspended with Carbon Nano Tube (CNT):Zirconium di Oxide (ZrO ₂ ) nanoparticle. The selection of optimal weight percentage (wt%) of nano particles are quite challenging. In this work, a novel approach (integrated Box Behnken Design (BBD) and Non-Sorted Genetic Algorithm - II (NSGA-II)) has been used to model and study the significant attributes. The Process parameters; CNT (wt%) Stirring Temperature (°C) and ZrO ₂ (wt%) with 17 different combinations based on BBD are experimentally investigated. The quadratic model obtained from BBD are set as multi objective problem to obtain optimal process parameter with maximum Breakdown Voltage (BDV), Fire point and minimum viscosity. The essential characteristics are evaluated in accordance with the established International Elecrotechnical Commission (IEC) and American Society for Testing and Materials (ASTM) standard. Conformation results have proved a good bond with the predicted results. It has been confirmed that this integrated BBD and NSGA-II technique helps to predict the uncovered values of CNT:ZrO ₂ without performing multiple experiments.

This study focuses on the restoration of aged FR3 transformer fluid by optimizing antioxidant formulations to enhance its dielectric and physicochemical properties. A composite antioxidant approach, incorporating tertiary butylhydroquinone (TBHQ), butylated hydroxytoluene (BHT), alpha-tocopherol, and citric acid, is employed to improve fluid performance. The efficacy of various antioxidant combinations is evaluated based on key parameters, including breakdown voltage, acidity, flash point, fire point, and dielectric dissipation factor. Response Surface Methodology (RSM) is utilized to optimize the antioxidant blend, while ANOVA (Analysis of Variance) assesses the statistical significance of the developed models. The optimized antioxidant combination improved breakdown voltage by 45.18% (R² = 0.9529), reduced acidity by 59.09% (R² = 0.9763), and enhanced flash by (37.17%) and fire points by 30.85% (R² = 0.9454 and 0.8022, respectively). Dielectric dissipation factor also showed strong improvement (24.32%) (R² = 0.9534). The high R², adjusted R², and adequate precision values indicate robust model reliability and predictive capability. While breakdown voltage exhibited acceptable lack of fit (p = 0.0255), other responses showed non-significant lack of fit, confirming model adequacy. These findings highlight the method’s effectiveness, synergistic antioxidant effects, and potential for sustainable transformer maintenance.

The paper examines the relationship between scanning electron microscopy (SEM) texture values and service life, in working multi region power transformers (90 No.), which allows conducting a minimally invasive insulation aging and condition evaluation. Five quantitative texture and morphological features are extracted from SEM images of Kraft paper insulation samples: gray level co-occurrence matrix (GLCM) contrast, entropy, porosity percentage and orientation index and fractal dimension. The paper analyses transformers of 11 to 230 kV voltage rating and service life of 0 to 35 years in various parts of Tamil Nadu, India. Correlations are found to be strong between texture features with the years of service with porosity (r = 0.70), fractal dimension (r = 0.72) and orientation index (r = -0.77) showing the strongest relationships. In order to combine the five features into one, a Texture Degradation Score (TDS) is developed by summing the five features and normalizing the score to a 0-100 range. The consistency and predictability of the proposed TDS formulation is validated using a linear regression model with the result of R² of 0.984 justifying the correlation-based strategy. Porosity and fractal dimension also become the most significant predictors and also they give 60 percent of the model predictive power. Multi-class classification has an 88.9% accuracy at classifying transformers as Good, Moderate and Aged condition depending on TDS thresholds. The correlation analysis shows that both rated voltage and power in the correlation analysis have little effects on the evolution of the texture, which implies that they are applicable across all transformer specifications. It is a correlation based methodology that allows quick, inexpensive, prognostic evaluation of transformer health, without destructive tests, to assist in maintenance decision-making and extending the life of assets in power transmission systems based on evidence.