How can we know the insulating paper state of the power transformer?
Nowadays, NATURE has published an article : Predicting the insulating paper state of the power transformer based on XGBoost/LightGBM models.
As we all konw,for electrical utilities, maintaining a power transformer's continuous operation is crucial since any transformer malfunction results in a loss of revenue. The majority of transformer failures stem from damage or failure of the insulation system, which consists of paper and insulating oil. The insulating paper is a crucial component of the transformer, and its condition serves as a key indicator of the transformer's overall health. The degree of polymerization (DP) measures the state of the insulating paper. A reduction in DP signifies paper aging and indicates increased degradation, leading to a higher risk of failure. As the insulating paper deteriorates, certain chemicals like Furfural (2-FAL), carbon monoxide (CO), and carbon dioxide (CO2) dissolve in the insulating oil; analyzing these compounds provides insights into the condition of the insulating paper. Other factors, such as moisture content and interfacial tension, can also contribute to the degradation of the insulating paper.
In fact,Power transformer plays a crucial role in the power networks. And in most of the situations, transformer malfunctions was due to failure in the insulating systems. Utilities are focused on the effective operation of the power network, emphasizing the early detection of transformer faults to prevent unwanted outages. The condition of the insulating paper serves as an indicator of transformer health, and its aging can lead to potential failures. Therefore, periodic and routine tests on the insulating oil are essential to assess the condition of the insulating paper.
The degree of polymerization (DP) is a crucial measure of the insulating paper's state. Various recommended tests, including dissolved gas analysis (DGA), breakdown voltage (BDV), oil interfacial tension (IF), oil acidity (ACI), moisture content (MC), oil color (OC), dielectric loss (Tan δ), and the concentration of furans, specifically 2-furfuraldehyde (2-FAL), were conducted to establish correlations between these parameters and the DP, and subsequently the state of the insulating oil. The data gathered from these tests were utilized to train XGBoost/LightGBM models, creating an artificial intelligence framework to predict the condition of the insulating paper. The results demonstrated the model's strong capability to accurately predict the insulating state.
