Development of A Hybrid Bayesian-Random Forest Framework for Transmission Line Predictive Maintenance in Nigeria

This study introduces a Bayesian-optimization-enhanced Random Forest (BO-RF) framework for predictive maintenance of Nigeria’s aging power transmission lines. Aimed at enhancing reliability and cost-efficiency, the framework automatically tunes the Random Forest hyperparameters via Bayesian Optimization, significantly improving predictive accuracy. Utilizing a dataset inclusive of historical failure records, environmental variables (temperature, humidity, pollution), and sensor-derived health indicators (vibration, partial discharge), t model demonstrates superior performance achieving a fault prediction accuracy of 92%, a falsepositive rate of 6%, and a mean time to failure (MTTF) prediction RMSE of 3.8 days outperforming baseline Random Forest and Gradient Boosting models. The implementation illustrates both practical feasibility and potential impact on strategic maintenance scheduling and grid resilience