Design and Simulation of a Predictive Maintenance Model for a Power Transformer: A Case Study of a Transmission Substation in Nigeria

The most crucial component of a high-voltage transmission system are power transformers, whose unplanned failure can cause major operational interruptions, monetary losses, and widespread power outages. A predictive maintenance framework and simulation for a 300-MVA power transformer at a 330/132 kV transmission substation in Lagos State, Nigeria, is presented in this study. Gradient Boosting algorithm (GBA) was utilized to analyze the transformer's historical data in order to predict potential transformer faults. Important data were collected and pre-processed, including Dissolved Gas Analysis (DGA), load profile, insulation resistance, top-oil and ambient temperature. The GB model was then trained and tested using the dataset. The obtained results indicates that the most significant indicators of transformer malfunctions were the presence of acetylene and hydrogen compounds. The model achieved 93% accuracy, with recall, Area Under the Curve (AUC), and average precision of 88%, 91% and 88.4% respectively. This suggests that the model has an effective fault prediction capability. Because it helps with problem identification, the study has demonstrated that GB can be utilized to increase the reliability of power transformer operations. Utilities seeking to minimize transformer downtime will find the model useful.