Anomaly Detection and RUL determining in NPP Power Transformers

Solution by: waico

Challenge

Chromatography is a method of separating and analyzing mixtures of substances. Chromatography aims at studying the physical and chemical properties of substances. It is based on the distribution of substances between two phases — stationary (solid phase or liquid bound on an inert carrier) and mobile (gas or liquid phase).

A special role in monitoring the condition of oil-filled equipment is played by monitoring gases dissolved in oil. The appearance in the equipment of almost any type of defect is accompanied by the formation of gases that dissolve in oil. Specific types of defects generate their gases in different quantities (for example, when local overheating of insulation or discharges occur in the area of ​​contact between insulating paper and oil, decomposition products are released in the form of characteristic types of gases).

Having studied the processes of gas formation by various types of defects, it is possible to judge the types of defects that led to their appearance by the composition of gases dissolved in oil. This principle is the basis of the most powerful diagnostic method, Chromatographic Dissolved Gas Analysis (CDGA), which can be part of a complete analysis of the operating oil.

At the initial stage of the development of defects, the amount of evolved gases is small and does not exceed the level of their solubility in oil. They can be stored in oil for a long time. For the extraction of these gases and their identification, CDGA is used.

Today, a significant number of NPP power transformers (block transformers, auxiliary transformers, and communication autotransformers) are operated with an extended service life, meaning the assigned service life is 25 years.

Considering the extended service life of NPP power units, there is the need to monitor the technical condition of power transformers.

Obtaining a software product is necessary analyzing the current state of power transformers and predicting remaining useful life based on the concentrations of dissolved gases in transformer oil measured during monitoring by SITRAM gas analyzers.

Solution

A web application with machine learning models analyzing the concentrations of four gases measured in transformer oil (H2; CO; C2H4; C2H2) set as a function of time to analyze the current state of power transformers and predict remaining useful life before a possible failure based on the concentrations set as initial data dissolved gases in transformer oil.

Related articles

For introduction into Remaining Useful Life (RUL) determination, you can visit this link.

For introduction into Anomaly Detection, you can visit this link.