Diagnosis and classification of internal defects of power transformers using a decision tree based on the simulation of the electrical model of the transformer

Electrical Engineering Master's Thesis

Power

Abstract

The extensive network of the power system has very expensive equipment, including generators, breakers, power cables and transformers. The power transformer is the beating heart of this network, which is always subject to various errors under the influence of operational and environmental conditions, and in some cases, it will cause the transformer to fail and go out of the circuit and be unavailable for a long time. As a result, maintenance programs should be based on operating and environmental conditions instead of time-based, which requires us to be aware of the current equipment conditions. Therefore, it will be very important to use monitoring and error detection methods that have the ability to evaluate the internal conditions of the equipment.

There are various methods to detect errors inside the transformer, among which we can mention the analysis of dissolved gases, partial discharge and frequency response analysis. Considering the limitations of the first two methods in detecting all types of faults, frequency response analysis is one of the best effective methods in the field of detecting electrical and mechanical faults inside the transformer. However, due to the reliance of this method on graphical comparison, the interpretation of the results obtained from the frequency response is very difficult and there is still no general and comprehensive method for classification. The aim of this thesis is to diagnose and classify the transformer fault with the help of frequency response and decision tree. Using the centralized electrical model of the power transformer, various errors have been simulated and their classification has been done using the decision tree. The results show that combining the frequency response with the decision tree has high accuracy and speed compared to other methods in fault classification in power transformers.  

Key words: power transformer; centralized electrical model; frequency response; decision tree; Error classification

1- Introduction

1-1- Introduction

One ??of the most important and complex systems that has been built so far is the power system. The electrical power system plays a key role in modern societies. Power transformers are one of the most important components in any power system. In fact, power transformers play the role of a communication link between the production and distribution sectors, and any unplanned outage causes power cuts and blackouts. Power transformers suffer different damages under different operational and environmental conditions. Some of these errors and damages are very severe and force the protection devices of the transformer to operate and take the transformer out of the circuit at once, while some of the errors are not so severe and the protection devices will not be able to detect them easily. This category of errors occurred in the insulation system, windings and core of power transformers, which are difficult to detect. Therefore, in order to evaluate the condition of power transformers, various tests and experiments are performed on them in a planned manner based on time. Most of these tests were performed in an untimely state, and this required the transformer to be removed from the circuit, which is not optimal and logical in terms of system reliability and costs related to power cuts and blackouts. Due to the importance of power transformers and the problems in offline tests, operators started to perform tests and diagnose errors in a timely manner in order to be constantly aware of the current status of the transformer and to prevent unplanned outages of the transformer and reduce outage costs. The transformer means the core and the windings. For example, by weakening the insulation system of the transformer, the pressure of the clamps decreases and as a result, it leads to a decrease in mechanical resistance.Many dielectric breakdowns inside the transformer are a direct result of the reduction of mechanical resistance due to deformation [1]. Therefore, it will be very significant and important to detect the deformation of the coil and core as soon as possible. 

There are various methods and tests to evaluate the condition of the transformer, including methods such as frequency response analysis, dissolved gas analysis, signal processing, leakage flux and negative sequence current. named [2]. Among them, the frequency response analysis method is a very popular and comprehensive method that has a high ability to detect errors and its implementation is simple and convenient. This stress will lead to changes in the windings and will lead to the potential failure of the transformer. These changes affect the capacitive and inductive values ??of the windings and as a result cause a change in the frequency response of the transformer and hence they can be easily recognized.

Analysis of the frequency response of the transformer, which has been presented since 1978, is a common tool for detecting changes in the transformer windings. Frequency response analysis will be implemented by injecting a signal between the terminals of the transformer and calculating the amplitude and phase of the received response against the frequency [3]. In general, this method is an industrial technique for skilled people in the field of fault finding to compare the frequency response with recorded historical data or with similar transformer information (so-called sister transformer) in terms of appearance. 

Minor shape changes in the transformer windings do not have any significant effect on the performance characteristics, but the mechanical properties of copper may change and also the impact resistance may be significantly reduced due to insulation damage and distance reduction. However, these deformations will be detectable after a long period of time through oil analysis or Buchholtz relay. 

This means that more advanced diagnostic methods are needed for the transformer using signal processing in order to detect the internal fault. Signal processing methods are used to extract useful information from the target signal. In this method, the signal can be a voltage waveform, tonal current or a combination of them. Because the existing methods for evaluating the internal conditions of the transformer cannot show all types of errors, intelligent methods are needed to be able to detect the error and its type. Different authorities have presented different methods to achieve this.

1-3- Review of articles

This part reviews the articles that have been researched and published in this field. In some of these articles, the modeling of the transformer is focused in order to determine the frequency response of the transformer, and in some others, the issue of detecting and classifying the transformer fault has been investigated. In reference [4], the electrical sequential network model for the high-voltage winding has been selected and its frequency response has been calculated. The frequency response is divided into three ranges: low, middle, and high, and the series capacitance in the low frequency range and inductance in the high frequency range are not taken into account, and the sensitivity of the frequency response to changes in parameters is investigated. In reference [5], a detailed model of the single-phase transformer is presented in order to detect the axial displacement error and deformation. In the mentioned model, parallel resistances (dielectric losses) and series resistances (copper losses) are considered dependent on frequency. Of course, the effect of the core and its related inductance at frequencies higher than 10 kHz has been neglected. Circuit parameters have been calculated from two analytical and finite element methods, and the sensitivity of different tests on frequency response has been evaluated. In references [6, 7], a laboratory method for detecting short-circuit faults in transformers using frequency response has been presented. The effect of measuring equipment (inductive and capacitive couplings) on the frequency response of the transformer has been researched.