Metal-oxide surge arrester monitoring and diagnosis by self-organizing maps

• Arresters monitoring and diagnosis exclusively from analysis of the total current.
• Applied voltage measurements are unnecessary.
• Current harmonic components obtained from computationally efficient algorithm.
• Self-organizing maps clustered the patterns according to their similarities.
• High hit ratios on the determination of arrester operation condition were obtained.

The main metal-oxide surge arrester are usually monitored by measuring the total leakage current and decomposing it into a capacitive component and a resistive component. However, these techniques present some limitations which make their implementation in the field quite difficult. This paper describes a monitoring technique based on classifying the harmonic characteristics of the leakage current. Self-organizing maps have been used in the classification process. Several station classes arresters were tested in the laboratory and their leakage current signals were recorded. Six different types of artificial imperfections were created in the arresters in order to assess the technical capability to discriminate arresters operating under different field conditions. The self-organizing maps are able to identify defective arresters with certainty of almost 98%. The results show that the technique is feasible for monitoring the condition of metal-oxide surge arresters.