Surface pollution is a major cause of partial discharges in high voltage insulators in coastal cities, leading to degradation of their surface and accelerating their aging process, which may cause visible arcing, flashovers and system faults. Thus, this work provides a methodology for the assessment of the condition of insulators based on an instrument which generates a severity degree to help the electric utility team schedule maintenance routines for the structures that really need it. The instrument uses a Raspberry Pi board as the processing core, a PicoScope oscilloscope for the data acquisition and an antenna as a partial discharge sensor. The algorithms are implemented in Python, and use artificial intelligence tools, such as a convolutional network and a fuzzy inference system. Laboratory test methods for the simulation of the field pollution conditions were successfully used for the validation of the instrument, which showed a good correlation between the pollution level and the severity degree generated. In addition to that, field collected data were also used for the evaluation of the proposed severity degree, which is demonstrated to be consistent when compared with the utility’s reports and the history of the selected areas from where data were collected.