Optimization of Mineral Oil Reclamation Using Response Surface Methodology and Multiobjective Genetic Algorithms

Abstract

Aged transformer mineral oil undergoes chemical and physical degradation due to continuous thermal, oxidative, and electrical stresses, resulting in the formation of acidic compounds, moisture, sludge, and gases. These changes compromise key oil properties, including breakdown voltage (BDV), total acid number (TAN), moisture content, interfacial tension (IFT), dielectric dissipation factor (tan $\delta $ ), and color index, ultimately affecting transformer reliability and service life. Reclamation using low-cost adsorbents such as activated bentonite and activated carbon provides an environmentally friendly solution. However, optimizing the treatment process is essential to maximize performance. In this study, response surface methodology (RSM) is employed to model and evaluate the effects of key process parameters: adsorbent dose, contact time, and temperature on six critical oil quality indicators. A multiobjective genetic algorithm (MoGA) is then applied to identify optimal operating conditions that simultaneously balance and enhance these parameters. The optimized treatment resulted in substantial improvements: TAN decreased by 78%, moisture content by 66%, BDV increased by 55%, IFT improved by 38%, tan $\delta $ decreased by 41%, and visual clarity was restored close to the original color index. RSM models exhibited strong predictive accuracy ( ${R}^{{2}} \gt 0.94$ for all parameters), while MoGA successfully mapped the Pareto-optimal tradeoffs. This integrated approach demonstrates a reliable and scalable strategy for sustainable transformer oil regeneration, leveraging data-driven optimization and intelligent decision-making.