THE INVESTIGATION OF MULTI-METHOD INTERPRETATION TO ENHANCE DISSOLVED GAS ANALYSIS FOR POWER TRANSFORMER DIAGNOSIS

• Main Author: Sutikno, Heri
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55323
• Institution: Institut Teknologi Bandung

Power transformers are critical components of the transmission grid, as they support the grid's ever-increasing energy demand. The liquid and paper insulation of the power transformer is critical to its performance. Compared to the other materials used in the transformer, the insulation system is the transformer's weakest point. Thermal and electrical stresses are applied to the insulating materials of power transformers during operation, resulting in accelerated ageing and failure. As a result, monitoring and diagnosing the insulation condition is critical to ensuring that the operating transformers operate at maximum efficiency. Dissolved Gas Analysis (DGA) is one of the most widely used diagnostic techniques for transformers used by various electrical utilities worldwide. The interpretation of DGA results is critical to the success of this diagnostic technique. Various techniques for interpretation have been proposed, including the Doernenburg Ratio Method (DRM), the Roger Ratio Method (RRM), the IEC Ratio Method (IRM), the Duval Triangle Method (DTM), and the Duval Pentagon Method (DPM). These are referred to as conventional techniques. However, interpretations are still based on the engineer's knowledge and experience, not on a mathematical calculation. Additionally, the ratio method's specific code may result in an out-of-code condition if the sample ratio is not present in the boundary value. There is a high probability that analyzing the same sample using multiple methods will yield inconsistent results. In extreme cases, incorrect interpretations may result in mismanagement and wasteful cost management. The purpose of this study is to develop new interpretation techniques for DGA in order to improve its accuracy and consistency. The developed method is multi-method, utilizing the scoring index and random forest machine learning principles to combine and integrate various existing methods. To validate the model's reliability, it was applied to three and six different types of faults. The final results demonstrate that the multi-method scoring index and random forest methods are more accurate and consistent than conventional interpretation methods. Additionally, the multi-method random forest demonstrated greater accuracy and consistency than the multi-method scoring index, with an accuracy value of 96 per cent and a consistency value of 93 per cent for the three types of faults and an accuracy value of 88.6 per cent and a consistency value of 87.4 per cent for the six types of faults. Additionally, the publication of IEEE C57.104 as the newest standard in 2019 resulted in significant revisions to the previous one. The most significant change in this standard is implementing a dissolved gas change rate as a reference value for detecting abnormalities caused by power transformer operation behaviour. The loading factor of a power transformer is highly correlated with its operation behaviour and is a significant factor in determining its insulation reliability. Managing proper loading for power transformers based on their DGA status may be another way to keep them healthy. As a result, it is necessary to conduct a field study to determine the effect of the loading factor on the growth rate of the dissolved gas. The final result demonstrates that the loading factor correlates with the change rate of dissolved gas at various concentrations. Additionally, the loading factor correlation value tends to be stronger in older transformers than in younger transformers