Partial discharge (PD) is an abnormal activity that occurs in high-voltage components, such as power cables, switchgear, machines, and power transformers. Such activity needs to be diagnosed for the equipment to last longer as PD could harm the insulation and potentially lead to asset destruction from time to time. Moving one or more externally mounted acoustic sensors to different locations on the transformer tank is commonly used in order to detect and locate PD signal occurring in the power transformer. However, this procedure may lead to less accuracy in PD identification. Therefore, this research paper presents an analysis of acoustic sensor placement based on time of arrival (TOA) technique for PD location in a power transformer. The detection and location can be determined by permanently installing the acoustic sensor to provide valuable data in an early stage of occurrence for online condition PD monitoring. Several methods are available for the detection of PD signal, whereby one of the best choices is via acoustic emission (AE). PD creates an ultrasonic signal used for PD detection. This paper proposes the possible placement of AE sensors to be mounted on the power transformer wall based on ideal and static PD signals. The sensors were placed in order to capture the PD signal without any disturbance signal from inside or outside the tank. The time for the signal for the first approach for each sensor is recorded to estimate the PD location using the TOA technique. A comparison between the least square method (LSM) and Gauss-Jordan elimination (GJE) for the TOA technique was analyzed to differentiate the resulting performance. This research utilized three different PD sources to apply the performance analysis on PD locations, while five cases were proposed to represent the five different placements of four sensors for the analysis. This research ultimately suggests that sensors be placed and randomly mounted on the four sides of the transformer tank, with one sensor allocated to one side. Among all five cases, Case 1 and Case 5 yielded a displacement error (DE) less than others, while between these two cases, Case 5 gave the lowest DE. The findings were recorded based on LSM and GJE methods used to differentiate the resulting performance.
