Improved Multi Grained Cascade Forest Model for Transformer Fault Diagnosis

This project focuses on the diagnosis of transformer faults using a multi-grained cascade forest model, applied to the power transformer fault detection and Remaining Useful Life (RUL) dataset. The dataset contains various fault categories, including "Normal Mode," "Partial Discharge," "Low-Energy Discharge," and "Low-Temperature Overheating," each representing different fault scenarios in power transformers. We applied machine learning models like XGBoost, Decision Trees (DT), Random Forest (RF), LightGBM, and Artificial Neural Networks (ANN) for classification tasks. To enhance model interpretability, SHAP and LIME were implemented for feature importance and explanation. A user-friendly web application was developed using Flask, HTML, CSS, and JavaScript. The platform allows users to log in, register, and input transformer data to predict fault categories based on the trained models. This solution enables efficient fault diagnosis in transformers, improving predictive maintenance and reducing unexpected failures.

Keywords:
Transformer Fault Diagnosis, Multi-Grained Cascade Forest, XGBoost, Decision Trees, Random Forest, LightGBM, ANN, SHAP, LIME, Flask, Predictive Maintenance, Power Transformers, Fault Classification.