Implementation of Machine Learning to Predict Cable Failures in Electrical Networks

Electrical networks are critical infrastructures that power industries, businesses, and households. Among their key components, electrical cables play a vital role in ensuring uninterrupted power distribution. However, cable failures due to aging, environmental factors, mechanical stress, and load imbalances pose significant challenges, leading to outages, financial losses, and reputational damage. Traditional maintenance approaches, which rely on periodic inspections and reactive repairs, have proven inadequate in preventing unexpected failures. In response to this challenge, predictive maintenance using Machine Learning (ML) has emerged as an effective solution. This research focuses on developing an ML-based predictive model to forecast cable failures in Dubai Electricity and Water Authority (DEWA)’s electrical network. The study leverages historical failure data, environmental factors, operational parameters, and real-time sensor readings to identify patterns associated with cable degradation. Various data preprocessing techniques, including missing value imputation, outlier detection (using Mahalanobis distance), and feature selection, were applied to enhance data quality and model reliability. The research follows the CRISP-DM methodology, ensuring a structured approach to business understanding, data exploration, model development, and evaluation. Multiple machine learning algorithms, including Neural Networks, Support Vector Machines (SVM), Logistic Regression, and Random Trees, were explored for predictive modeling. The models were validated using partitioning techniques, data balancing strategies, and evaluation metrics such as AUC-ROC curves, confusion matrices, and predictor importance analysis. Experimental results demonstrate that ML-driven predictions can significantly improve maintenance planning, reducing unplanned outages and lowering maintenance costs. The best-performing model achieved an accuracy of 98.76% and a recall of 97.78%, highlighting its effectiveness in predicting cable failures. Furthermore, implementing ML-based predictive maintenance resulted in a cost reduction of 53.61%, reinforcing its financial and operational benefits. The findings highlight the potential of AI-driven predictive maintenance in enhancing the reliability and efficiency of power distribution networks. By adopting ML techniques, DEWA can proactively detect early signs of cable deterioration, optimize resource allocation, and improve service continuity. The study contributes to the broader field of smart grid management, offering a scalable solution that can be implemented in other electrical utilities worldwide. Future research may focus on integrating Internet of Things (IoT) sensors, deep learning architectures, and real-time adaptive models for even greater predictive accuracy.