Moisture in the transformer insulation can shorten its life. There are many methods for detecting humidity in transformer paper insulation. One of the methods used in the factory to evaluate the drying process of transformer insulation and determine its humidity is the frequency response analysis method. In this paper, the desired experiments are performed on different transformers, and after obtaining the results of frequency response measurements, the required features are extracted from them. Then, using the k-means method, these features are placed in three clusters (dry, wet, and excessively wet). The cost function of the k-means method is optimized using the particle swarm optimization (PSO) algorithm to get a better result. By applying new data from different transformers, the capability of the proposed method in determining the moisture content of the transformer is evaluated. The results obtained from the evaluation of the insulation condition of another group of transformers indicate the high accuracy of the proposed method.
Type of Study:
Research Paper |
Subject:
Artificial Intelligence Techniques Received: 2020/12/25 | Revised: 2021/10/16 | Accepted: 2021/10/04