A Hybrid Machine Learning-Genetic Algorithm (ML-GA) to Predict Power Tranformer Life

Abstract

Power transformers are important elements of the electrical system, and it is  important to estimate their remaining age with accuracy in order to plan maintenance efficiently. This research suggests an innovative method to improve power transformer life predictions by combining Machine Learning (ML) With Genetic Algorithms (GA). The method uses parameters like temperature, current load conditions, and operational parameters like transformer oil for training engine learning models on time series data. Genetic algorithms are used to develop and choose the most relevant features for powerful predictions, therefore optimizing the model's performance and feature selection. The purpose of the ML-GA hybrid method is to solve the weaknesses of conventional methods while providing improved accuracy and adaptability to a changing operational environment. The experiments that suggest the efficiency of the suggested method on the real-world dataset are presented in this article. A comparison with current approaches demonstrates the benefits of the ML-GA Hybrid model in power transformer life prediction. According to MSE and RMSE, the resultant error level is actually close to 0, indicating that the ML-GA prediction model is fairly accurate. This innovative frame provides utilities and businesses useful tools to maximize maintenance plans, save downtime, and improve the life of power transformers, all of which improve the general reliability of the electrical power system.