Higher penetration of renewable energy and increased electricity consumption makes delivering a reliable power supply a complex task. Power system operators are nevertheless required to provide a reliable power supply with a power quality within specified limits. Power quality events can cause malfunctioning of sensitive equipment for consumers in the distribution network. Communities in the periphery of radial distribution networks, such as rural and island communities, often experience more challenges with power quality than other areas. This study focuses on a small island community in the Norwegian Arctic that exemplifies the challenges of power reliability. It explores an alternative approach to utilise the existing electrical infrastructure. By analysing one year of measurements from an electrical substation near a large industry complex, combined with weather forecast data, machine learning models are employed to recognise conditions leading to low power quality. Although these models show satisfactory performance, some data is missing to fully understand the causes of such events. To address this, feature selection techniques are applied to identify the most significant data features and find new important features. Through this process, a refined dataset is obtained, enabling more accurate detection of power quality events in the Arctic community. This research not only enhances the detection capability but also provides valuable insights into the specific data features that contribute to identifying power quality events, offering opportunities for targeted mitigation strategies and generalisation to other locations.