Power transformer condition assessment based on online monitor with SOFC chromatographic detector

Highlights

• An online monitor based on SOFC gas sensor for DGA is designed and developed.

• The minimum detectability of the monitor to hydrocarbons lower than 0.1 ppm.

• The results of the online monitor can be calculated directly without calibration.

Abstract

Power transformer is vital for energy conversion in power system, the condition assessment of which based on dissolved gas analysis (DGA) is used worldwide. In this paper, an online monitor based on solid oxide fuel cell (SOFC) detector is designed to address the poor stability, low sensitivity of other online-monitoring system. The system is composed of a gas measurement subsystem, a central controlling unit, and an intelligent fault diagnosis subsystem. A portable gas chromatography module developed using SOFC detector. The quantificational algorithm is derived, on basis of which calibration can be skipped. Due to the essential role of the chromatography column as well as gas the detector’s temperature, proportional-integral-differential (PID) controller is implemented in the controlling unit. To evaluate the effectiveness of the developed online monitor, the quantitative analysis and engineering application have been performed; then, based on the measured results, the RVM-ANFIS algorithm is adopted to diagnosis the latent transformers faults. Using the developed online monitor, the results indicate that five characteristic gases, H₂, CH₄, C₂H₄, C₂H₆, and C₂H₂ can be measured accurately, and incipient faults can be detected, validate the effectiveness of this system.

Introduction

As a key component in power system, the oil-filled power transformer is used for electric energy conversion and transmission [1], [2], [3], [4]. Condition assessment of the power transformer continuously is significant to avoid collapse of the power system. For oil-filled transformers, mineral oil is used largely for insulation as well as heat dissipation [5], [6], [7]. Hence, when transformer failure occurs, the oil molecular is breakdown into combustible gases, e.g., H₂, CH₄, C₂H₄, C₂H₆, C₂H₂ and CO, CO₂. Accordingly, the dissolved gas analysis (DGA) is utilized worldwide to monitor the status of the power transformer.

In the past decade, to prevent failure of power transformer, the scheduled offline dissolved gas analysis is performed regularly. As the most accurate DGA measuring technique, the general procedures of DGA in most utilities are: (1) sample oil from power transformer in substation and transport to laboratory manually; (2) extract gas from oil; (3) perform gas chromatograph analysis. On one hand, this process is performed manually which is trivial and inefficient. On the other hand, due to the complexity of above operating process, for most of the power utilities, the regular interval of offline measurement is about two months ~ three months. However, the failure may happen in the interval and is not detected. Hence, with the development of smart grids, online monitor of power transformer is a promising method and have been extensively researched, which is able to measure other parameters, such as Furan, acidity, interfacial tension of transformer oil [8], [9], [10]. Moreover, the artificial intelligence algorithm, e.g., the fuzzy logic [11], artificial neural network (ANN) [12], relevance vector machine-adaptive neuro fuzzy inference system (RVM-ANFIS) [3] can be integrated into the online monitoring system to promote the status evaluation level of power transformers.

Presently, there are varieties of measurement techniques developed for online DGA monitors, e.g., gas sensor array [13], [14], infrared spectroscopy [15], photoacoustic spectroscopy [16], gas chromatograph (GC) techniques [17], et al. However, for online DGA monitoring, these techniques suffer from certain limitations. Given that the characteristic gases of transformer oil are mixtures, the sensitivity of gas sensor array is high while the selectivity is poor. Similarly, infrared spectroscopy (IS) can detect most of the DGA feature gas except H₂, which is essential in most of the diagnostic algorithms [18], [19], especially for partial discharge fault diagnosis. In terms of photoacoustic spectroscopy (PS), it may be affected easily by the electromagnetic interference and noises in substation environment. Moreover, the detection limit of IS and PS need to be promoted for high-voltage transformers monitoring.

In contrast, due to the great separation ability and high sensitivity of GC, it plays an important role for DGA application in the worldwide, i.e., offline or online monitoring [20], [21], [22], [23]. It should be noted that the performance of GC system is determined by its detectors, e.g., the most common GC detectors are the flame ionization detector (FID) [24] and the thermal conductivity detector (TCD) [25]. However, the detection limit of TCD does not meet the requirements of online monitoring, and H₂ is needed as auxiliary gas for FID, which is not allowed to be used in substation. To address this problem, we have developed SnO₂ detector for DGA application [17]. However, the limitation of SnO₂ to C₂H₂ is about 0.16 µL/L, hence, it is more proper to monitor the low voltage-level transformers, i.e., the required limitation to C₂H₂ is large than 0.5 µL/L. To promote the detection limitation of the characteristic gases, Fan et al. [23] developed the SOFC detector and applied it to DGA, and its corresponding properties are also studied, showing the advantage of high-sensitivity, good repeatability, fewer chromatographic columns and carrier gases. However, it is only used in laboratory, and the calibration process in [23] is complicated in actual application.

The contributions and novelty of this paper are as follows: 1.It is the first report that the online DGA monitoring system based SOFC detector is developed and applied to the substation. 2. According the principle of the SOFC, the qualification measurement without calibrating gases is firstly applied to the online DGA monitoring system. We extend the work presented in [23] about our newly developed high-sensitivity and low-cost SOFC chromatographic detector, and an online monitoring system are presented, in which the framework of the online DGA system, online GC instrument and digital signal (chromatogram) processing, mathematical model of SOFC detector without curve fitting, and power transformer condition assessment based on the measured DGA data are elaborated.

The developed online monitor can be used to detect five feature gases, e.g., H₂, CH₄, C₂H₄, C₂H₆, C₂H₂ dissolved in transformer oil, which are used largely for transformer faults diagnosis. Essentially, the purpose of the DGA is for power transformer condition assessment. In practice, the status evaluation of transformer is performed by the maintainers who are nonexperts, hence interpreting the DGA data obtained from online monitors is a challenging task since the relation between the faults type and the gas concentration is complex [26]. Hence, there is a necessity to integrate intelligent algorithms into the monitor for identifying faults type of power transformer automatically. Due to the excellent capability of distinguishing multiple faults and samples with ambiguous characteristic, the RVM-ANFIS algorithm developed in our previous work [3] is adopted in this online monitor. By using the developed online monitoring system, DGA can be carried out periodically in smart substation without human interference, providing status evaluation for transformers automatically. To verify effectiveness of the presented online DGA system and applicability of the diagnosis algorithms, case studies and results of applying the proposed system are provided.