Abstract
The implementation of artificial intelligence to determine the condition of a transformer based on temperature is something new. This study is an initial study to facilitate the process of monitoring or supervising transformers based on temperature. By processing thermal images obtained from the AMG8833 thermal sensor, an analysis of the temperature conditions of the transformer and its surrounding environment can be carried out. The thermal images obtained provide an overview or represent the distribution of temperature around the object observed by the AMG8833 thermal sensor. Furthermore, the thermal image is processed to obtain the percentage of white and non-white colors. The use of the SVM method aims to assist the justification process based on the percentage of white (high temperature) and non-white (low temperature). By using 10 training data that have obtained the percentage values of white and non-white colors, it produces 90% accuracy in recognizing the conditions of training data in high/low temperature conditions which can be interpreted as dangerous or safe. The hyperplane line equation obtained from the SVM method is used to distinguish between dangerous or safe categories. This line is a dividing line on the training data to categorize between high temperature data and low temperature data. Furthermore, the line is used to test 2 test data in determining dangerous or safe conditions. The results obtained 100% accuracy in determining dangerous or safe conditions on the test data. The SVM method successfully classifies the test data well even though the accuracy is not 100% when classifying training data.