We propose a methodology for proactive maintenance of Uninterruptible Power Supply (UPS) devices based on online health-status monitoring and application of data analytics for prediction of UPS components' failures. The goals of the work are (i) improvement of reliability of UPS devices by ensuring prompt action prior to a failure of a component, and (ii) provision of more cost-effective maintenance by servicing the device only when required instead of periodically. Improving reliability of UPS'es also enhances dependability of critical infrastructures that require high quality power supply.