Monitoring and Diagnostic System for Dry-Type Transformers Using Machine Learning Techniques

Abstract

Power transformers are recognized as high-value assets in substation design, but their susceptibility to various failure modes poses a significant risk of damage and power supply disruptions. Consequently, extensive research has been conducted to develop diagnostic techniques and monitoring methodologies for these devices. This project aims to develop a comprehensive solution comprising hardware and software components for the online diagnosis of dry-type transformers, primarily focusing on the detection of Inter-Turn Short Circuits (ITSC) in conjunction with Partial Discharge (PD) signatures. Dry-type transformers utilize ambient air as both a cooling and insulating medium. Among its advantages, the most relevant for the oil and gas industry are the lower maintenance costs and the absence of flammable oil, ensuring lower fire risks, which is a critical factor in facilities. As offshore electrical plants grow in complexity, there is an increasing demand for dry-type transformers with higher power ratings. Effectively monitoring the operational condition of such transformers serves as a strategic tool to enhance the reliability, robustness and safety of the electrical system, while potentially reducing overall maintenance expenditures. Such transformers offer notable advantages in terms of safety and reliability [1]. However, they come with higher costs and have lower power and voltage limits. Similar to any other equipment, transformers experience aging as a natural consequence of their operation. Its most significant consequence is the gradual degradation of insulation due to thermal effects and mechanical stresses resulting from electromagnetic interactions between windings turns. Additionally, transformers are susceptible to short-circuits, which induce intense electromechanical stresses in the windings, as well as overvoltages stemming from maneuvers such as line energization or the presence of inductive or capacitive loads. These factors elevate the dielectric stress on insulating materials and connections, potentially surpassing their design limits.