Parameters Estimation of an Extended Debye Model for Power Transformer insulation using Voltage Recovery Method (RVM) and Genetic Algorithm (GA)

The Recovery Voltage Method (RVM) serves as a valuable tool in assessing the state of power transformers' insulation systems by estimating their polarization spectrum, crucial for evaluating aging and moisture-induced degradation. Typically, the insulation behavior is modeled using an extended Debye model (ED) comprising parallel branches, each representing various dipole groups within the dielectric. However, obtaining analytical solutions for the Debye equations with multiple branches poses a challenge, often leading to simplified models that inadequately correlate with measured RVM data. To address this, this paper proposes a multi-branch ED model incorporating two or more polarization dipoles (RPI-CPI) and introduces a parameter estimation method utilizing a time-domain mathematical model. This approach integrates key RVM parameters such as recovery voltage peak, peak time, and initial slope of the dielectric response. Parameter estimation employs a Genetic Algorithm (GA), minimizing an objective function considering both experimental and calculated RVM data. Evaluation of the proposed method involves examining ED circuits with two and three branches to understand how the number of dipoles influences accuracy. Results demonstrate the effectiveness of the technique, showing good consistency between calculated and measured RVM curves. Notably, employing three branches enhances curve coincidence compared to two branches, suggesting that including an additional branch improves the equivalent circuit's performance and accuracy.