Accurate oscillation extraction in control systems is crucial for effective monitoring and root cause analysis. Recently, the ensemble empirical mode decomposition (EEMD) and its variants have garnered significant attention in the field of oscillation monitoring due to their effectiveness in analyzing and processing nonstationary data. However, the decomposition performance of these methods is highly sensitive to the calibration of noise amplitude. A mismatch between the assisted noise and the intermittency in target signal can exacerbate mode mixing, significantly compromising the accuracy of oscillation extraction. To address this issue, we propose a self-tuning EEMD (STEEMD) method, designed to minimize additional mode mixing by introducing an evaluation index of mode mixing for calibrating the optimal noise amplitude. The superiority of STEEMD, relative to established methods including EEMD, complete EEMD with adaptive noise, and median EEMD, is validated through comprehensive studies encompassing both simulated scenarios and industrial oscillation cases.