Digital signal processing (DSP) methods have been used by many researchers for detection and classification of transient disturbances because of their fast and powerful abilities to recognise waveform distortions. However, some DSP methods such as the wavelet transformation (WT) show less accuracy when applied to noisy real data. In this study, disturbance features are extracted in the wavelet domain based on the WT levels. Moreover, a new feature extraction algorithm namely normalised Renyi entropy with the signal energy is applied. This algorithm has been proven to be effective and robust for noisy signals. However, their application in power systems has not yet been tested. Using a laboratory setup of an islanded micro-grid, experimental results validate the efficacy of the wavelet-based normalised Renyi entropy in the detection and classification of four disturbance types (voltage sag, interruption, harmonics, mixture of harmonics and sag).