Partial Discharge (PD) detection and measurement are considered as reliable source of information related to early signs of insulation degradation in order to avoid complete breakdown and longer power outages. During the last decade, the online PD measurements and self healing networks gained significant interest over the periodic maintenance. However, online measured signals are suppressed by high frequency noise, therefore, de-noising of measurements is of paramount importance to get reliable information about a fault. An adaptive de-noising technique based on discrete wavelet transform (DWT), capable of de-noising the measured signals automatically without any human intervention is presented in this paper. The major challenges in using DWT such as wavelet function selection and reconstruction of de-noised signals are addressed. A simple criteria about selection of decomposition coefficients based on dominant frequency and amplitude is used in the presented algorithm. The de-noising performance is evaluated by comparing with other techniques such as correlation based wavelet selection and energy based wavelet selection methods. The results prove that adaptive de-noising is simple, effective and reliable in terms of run time which makes it more useful for online monitoring application.