Abstract
Power transmission equipment condition is critical to power system stability. Detecting incipient faults in transformers as early as possible can significantly reduce losses from temporary outages and improve operational maintenance levels. This paper proposes a condition evaluation method for power equipment based on defect, testing, and operational data. By establishing intelligent evaluation models, consistent data extraction rules, and equipment specifications, this method accurately identifies equipment status in real-time. This approach supports scientific maintenance strategies and provides a multi-dimensional assessment of overall equipment health.
Introduction
Traditionally, power grid maintenance has relied on periodic, calendar-based schedules—a "good due to compulsory" philosophy. However, as grids grow in complexity, this approach has revealed significant flaws:
- Blindness: Maintenance is often performed on healthy equipment, leading to unnecessary downtime.
- High Costs: Manual periodic checks are prohibitively expensive at scale.
- Lagging Response: Manual intervention often fails to catch faults before they cause outages.
- Resource Waste: Functional components are replaced prematurely based solely on age.
To address these issues, this research proposes an "intelligent condition evaluation" system that shifts the paradigm toward "state overhaul"—maintenance driven by actual equipment health data.
Methodology
The core of the proposed system is a hierarchical evaluation model based on a weighted fault tree. Equipment, such as a transformer, is decomposed into four primary components: General, Ontology, Casing, and Non-electricity protection systems.
Scoring and Weighting
Each component is assigned a weight (1 to 4) based on its criticality. Degradation is categorized into four severity levels (I to IV), with basic scores ranging from 2 to 10 points. The final deduction score is calculated as the product of the degradation points and the weight coefficient.
State Classification
Equipment is classified into four states—Normal, Attention, Abnormal, and Severe—based on total and single-component scores. The system follows a "bottleneck principle," where the overall state of the transformer is dictated by its most severely degraded component.
Results and Case Study
The model was validated using operational data from the Yunnan Power Grid Corporation, specifically targeting substations in Fuyuan, Huize, and Dongshan.
110kV Huize Substation
The system flagged the #2 main transformer for "Attention." Specific issues identified included seal aging, slight oil leakage, and low oil levels in the conservator. Given the unit's 10-year operational history, the system recommended strengthened monitoring.
35kV Dongshan Substation
The evaluation identified damage to the #2 main transformer fan motor. The system provided a clear recommendation for immediate replacement to prevent overheating and further damage.
Conclusion
The intelligent condition evaluation method provides a robust framework for transitioning from periodic to condition-based maintenance. By utilizing weighted scoring and hierarchical decomposition, the model removes the "blindness" of traditional methods and provides actionable decision support for grid managers. Future work should focus on integrating more advanced machine learning techniques to move from reactive evaluation to predictive maintenance, further enhancing the reliability of the power transmission network.
Summary updated