Transformer hot spot temperature estimation through adaptive neuro fuzzy inference system approach

Area of Science:

• Electrical Engineering
• Materials Science
• Artificial Intelligence

Background:

• Power transformer insulation degrades due to thermal stresses from dynamic loading and environmental changes.
• Accurate winding hot spot temperature (HST) estimation is vital for maintaining power system reliability and preventing accelerated insulation aging.
• Conventional thermal models for HST computation are complex and non-linear, necessitating improved estimation methods.

Purpose of the Study:

• To enhance the accuracy of top oil temperature (TOT) and winding hot spot temperature (HST) estimation in power transformers.
• To develop and validate an adaptive neuro-fuzzy inference system (ANFIS) model for real-time thermal performance monitoring.
• To provide a reliable method for asset managers to assess transformer loading and ensure operational longevity.

Main Methods:

• Developed a sub-ANFIS model for top oil temperature (TOT) estimation using loading and ambient temperature as inputs.
• Implemented a hybrid optimization technique to fine-tune ANFIS membership functions for improved accuracy.
• Validated the ANFIS model using field data from a 60/90MVA, 132 kV power transformer under dynamic operating conditions.

Main Results:

• The ANFIS model demonstrated high accuracy in estimating both TOT and HST, achieving coefficients of determination of 0.98 and 0.96, respectively.
• The model exhibited low mean square errors of 7.8 for TOT and 10.3 for HST, outperforming conventional thermal models.
• Validation against field data and analogous models confirmed a precise input-output correlation and the model's reliability.

Conclusions:

• The adaptive neuro-fuzzy inference system (ANFIS) provides a highly accurate and reliable method for estimating power transformer winding hot spot temperatures (HST).
• Accurate HST monitoring using ANFIS facilitates effective asset management, enabling optimized loading recommendations and enhancing overall transformer lifespan.
• This approach addresses the limitations of conventional models, offering a significant advancement in power transformer thermal analysis and operational safety.