Adaptive Weighting of Oil Quality Index on Power Transformers Using Particle Swarm Optimization

Power transformers play a significant role in the entire electrical network. However, local, or large-scale power failures may occur due to the deteriorating quality of the insulating oil due to electrical and thermal disturbances. Monitoring the condition of insulating oil regularly needs to be carried out to reduce the risk of failure of power transformers, involving determining the oil quality index in power transformers which includes five parameters, breakdown voltage, water, acidity, interfacial tension, and color. Unfortunately, the weight value to determine the oil quality index has depended on the subjectivity of experts. There is no definite standard regarding how to specify the weight. Therefore, in this research, Particle Swarm Optimization (PSO) is used to adaptively determine the weight value of each parameter determining the oil quality index. Each particle in the population represents the weight value of the five parameters, which is determined based on the results of the oil testing that has been carried out. The particles look for the best combination of parameter weights so that the oil quality index categories obtained are close to the original results. The test results that have been carried out using 504 data indicate that the optimal parameters of PSO consist of the number of iterations is 168 and the population size is 65. By using these optimal parameters in the PSO process, accuracy testing using Mean Absolute Percentage Error produces a minimum value of 2.48%, which is classified as a highly accurate prediction.