On prospects of development of automated complex systems of continuous condition monitoring of power transformers

The power transformer and its condition are one of the components of ensuring the reliability of the power grid complex. As power companies transition to actual maintenance and repair of equipment, various diagnostic techniques, especially under operating voltage, have become widely used. The demand for control is also due to the large percentage of operating electrical equipment over the standard period. To assess a power transformer, a technique of an integral indicator of technical condition is used, which includes a set of diagnostic parameters at different units of the transformer. Information on the condition of electrical equipment is traditionally obtained during testing of units during repair work, which requires disconnection of the line. Currently, the development has received automated monitoring and technical diagnostics of electrical equipment, which is confirmed by a number of regulatory and regulatory documents of electric grid companies. The monitoring systems proposed by the manufacturers are distinguished by a significant set of measured parameters, many of which are additional, not included in the list of NTDs. In Russia, such systems were used as pilot projects during the reconstruction and construction of facilities in PJSC «Rosseti Moscow Region». To improve methods for predicting and optimizing the operation of the electric power system, it is logical to use artificial intelligence methods that are critical for expanding a person's cognitive abilities in these tasks. In the world, the Support Vector Method (SVM), Extreme Learning Machines (ELM), fuzzy logic, and several others have been combined with methods of interpreting ARG to analyze nascent faults in transformers. Convolutional neural network models are actively used to classify electrical equipment faults during thermal imaging of transformer substations. The use of the YOLOv4 network allows identification of 4 types of power equipment at the substation with preliminary image processing (detection of background, interference caused by extreme weather conditions, noise and other factors). Insufficient data exchange between operating companies in Russia leads to the slow introduction of artificial intelligence methods.