Application of the Random Forest Method in Predicting the Risk of Power Network Disruption Based on PLN UP3 Lhokseumawe Maintenance Data

Abstract

Maintaining the reliability of electricity distribution systems remains a major challenge for PT PLN (Persero), Indonesia's state-owned electricity utility, particularly in minimizing unplanned outages caused by equipment degradation and operational disturbances. Conventional maintenance strategies based on corrective and preventive approaches are often insufficient to prioritize assets with higher disruption risks. Therefore, this study proposes the implementation of the Random Forest algorithm to predict the risk of power network disruption using historical maintenance data from PLN UP3 Lhokseumawe. The dataset consisted of maintenance-related variables, including inspection activities, maintenance findings, measurement activities, Fuse Cut Out (FCO) replacement, load balancing, transformer grounding repair, and climbing barrier installation. Data preprocessing included data cleaning, standardization, and Risk Priority Number (RPN)-based risk labeling to classify maintenance records into Low-, Medium-, and High-risk categories. Experimental results demonstrated that the proposed model achieved an accuracy of 97.6%, with an average precision, recall, and F1-score of 96.2%. Feature importance analysis identified Measurement, Tier 1 Findings, and Tier 1 Inspection as the most influential predictors of disruption risk. The proposed approach supports data-driven predictive maintenance by enabling early disruption risk identification, thereby improving maintenance planning and electricity distribution reliability.