Building Machine learning datasets for oil-immersed service transformer health assessment using Fuzzy logic method

This paper proposed a low-cost method to build the machine learning training dataset for assessing service transformer health by using fuzzy logic method. The training dataset is tested on a stimulated 50kVA server transformer. The monitoring data is collected from the real-time energy monitoring device which is installed near the transformer to measure ambient temperature, current, and voltage. The condition of transformer is evaluated by using Support Vector Machine algorithm. The data generation proposed in this paper has high feature continuity and good scalability that can be used as a training data for machine learning, deep learning models.