A Comprehensive Database and Smart-Learning Framework for Monitoring Failure Risk Factors, Maintenance, and Protection in Electrical Networks

Abstract

Electrical power systems are exposed to interacting electrical, thermal, environmental, and resource-related faults such as leakage current, voltage and frequency deviations, overcurrent, harmonic distortion, phase-sequence error, humidity, fire, wind-speed variability, water insufficiency, and solar-resource loss. This study introduces a software-based database-generation and smart-learning framework that converts 22 candidate risk factors into six normalized severity levels and then maps the simultaneous system state to low-, medium-, and high-level protection decisions. The main novelty is that the software not only evaluates existing measurements; it also produces a literature- and standards-informed synthetic database when long-term real field measurements are not yet available. The database is generated by defining variable limits, sampling realistic operating states, computing severity labels, and storing input–output pairs that can later train or validate predictive maintenance models. The proposed framework therefore, links protection logic, database construction, and reusable training data in a single workflow. The results show how simulated annual operating scenarios can be transformed into structured risk records, warning classes, and shutdown decisions, supporting early fault detection, maintenance planning, and resilience improvement in renewable-integrated electrical networks.