Since many IoT devices are battery-powered, dealing with power consumption uncertainty is vital to ensure energy efficiency, as changing batteries is costly or sometimes even impossible. Self-adaptation is a major approach to handling uncertainties inherent to IoT applications, such as changes during operation (e.g., the sensor’s inaccurate collecting) and modifications in the availability of resources, e.g., decreasing energy battery level. Existing solutions commonly use data aggregation (processing demand), predefined duty cycling, or extra hardware elements for energy harvesting. In this context, this paper faces the energy uncertainties of IoT devices by proposing an adaptation mechanism that dynamically adjusts the device’s duty cycle, allowing applications to operate with different execution times for energy saving at runtime. Evaluation results show that dynamically changing the device’s duty cycle can better adjust the IoT application to fit its context while suggesting increased battery life.